A novel polymeric prosthetic heart
valve: design, manufacture, and
testing
Jacob Brubert
Department of Chemical Engineering and Biotechnology
and
Queens’ College,
University of Cambridge,
UK
September 2015
This dissertation is presented for the degree of
Doctor of Philosophy
In memory of my grandparents.
ii
iii
Declaration
This dissertation is the result of my own work and includes nothing which is the outcome
of work done in collaboration except as declared in the Preface and specified in the
text.
It is not substantially the same as any that I have submitted, or, is being concurrently
submitted for a degree or diploma or other qualification at the University of Cambridge or
any other University or similar institution except as declared in the Preface and specified
in the text. I further state that no substantial part of my dissertation has already been
submitted, or, is being concurrently submitted for any such degree, diploma or other
qualification at the University of Cambridge or any other University of similar institution
except as declared in the Preface and specified in the text.
This dissertation contains fewer than 65,000 words including appendices, bibliography,
footnotes, tables and equations and has fewer than 150 figures.
Jacob Brubert
iv
Abstract
In this thesis a flexible leaflet polymeric prosthetic aortic heart valve was designed, man-
ufactured, and tested. The prosthesis was designed with the aim of overcoming the need
for anticoagulant therapy, which is required for current mechanical prostheses; while also
having lifelong durability, which current bioprosthetic heart valves are not able to achieve.
Inspired by the anisotropic architecture of collagen in the natural valve, a shortlist of
polystyrene based block copolymers (BCPs), which can be processed to yield mechanically
anisotropic materials, was proposed. The shortlist was evaluated based upon processabil-
ity, biostability, ex vivo haemocompatibility, and a novel material performance index
comprising the flexural modulus and the cyclic fatigue stress predicted by fracture me-
chanics methods. Polystyrene-block -polyethylene-polypropylene-block -polystyrene with
22 mol% polystyrene (SEPS22) was selected for further testing and use in the design.
Haemocompatibility and calcification of the BCPs was assessed against reference ma-
terials. In measures of coagulation and thrombogenicity the BCPs were better than
polyester, but worse than expanded polytetrafluoroethylene and pericardium graft ma-
terials. In measures of inflammation, the BCPs and polytetrafluoroethylene were bet-
ter than polyester and pericardium. A durable heparin coating gave SEPS22 superior
haemocompatibility compared to all the reference materials. The BCPs calcified less
than pericardium, but calcification still accelerated failure.
The technique of injection moulding discs of the BCP from a point was used to create a
novel biaxial structure of cylindrical polystyrene domains. A combination of modelling
and bench-scale injection moulding was used to select a point from which the prosthetic
heart valve injection tool cavity should be filled. By simultaneously injecting at a point
at the centre of the free edge of each leaflet, a bioinspired orientation was produced.
Based upon hydrodynamic testing, a spherical form leaflet design was selected. The
hydrodynamic performance of the complied with the ISO 5840 standard for cardiac valve
prostheses, but the fatigue performance was inadequate due to the leaflets being thinner
than specified due to manufacturer error. Fatigue prediction and finite element analysis
were used to conjecture that correctly manufactured polymeric valves could theoretically
reach the ISO limit, indicating that there is potential for polymeric prostheses to overcome
the issues of durability and need for anticoagulation.
vSummary
In this thesis I designed, manufactured, and tested a flexible leaflet polymeric prosthetic
heart valve. The prosthesis was designed with the aim of overcoming the need for anti-
coagulant therapy, which is required for current mechanical prostheses; while also having
lifelong durability, which bioprosthetic heart valves are not able to achieve. No flexible
leaflet polymeric heart valve has yet reached the clinic successfully.
Inspired by the anisotropic architecture of collagen in the natural valve, I proposed a
shortlist of polystyrene based block copolymers (BCPs) which can be processed to yield
mechanically anisotropic materials. The BCPs self-assemble at the nano-scale to form stiff
polystyrene cylinders in a rubbery bulk. I formed a shortlist of cylinder-forming BCPs
which I evaluated based upon processability, biostability, and ex vivo haemocompatibility.
To select optimal polymeric materials for a prosthetic heart valve I proposed a mate-
rial performance index, in which leaflet thickness was the design variable, and flexural
modulus and fatigue stress were combined in the ratio
Eflex
σ3f
, such that superior mate-
rials had a lower performance index. To rapidly evaluate the durability new materials
I used fracture mechanics methods for the prediction of the cyclic fatigue stress, σf .
I selected polystyrene-block -polyethylene-polypropylene-block -polystyrene with 22 mol%
polystyrene (SEPS22) for further testing and use in the design.
Two potential leaflet designs were proposed. One was based on cylindrical surfaces (Bur-
riesci et al., 2010), the other on spherical surfaces (Zaffora, 2011). Prototype valves were
injection moulded from SEPS22 and their hydrodynamic performance tested in vitro.
The spherical form valve had a smaller effective orifice area, but lower regurgitation,
than the cylindrical form valve. Only the spherical form valve complied with the ISO
5840 standard, and so this design was selected. The spherical form valve design was
subsequently improved by increasing the thickness of the stent posts which reduced their
deflection during diastole.
I proposed injection moulding as a means to orient the stiff polystyrene domains and
thus manufacture prostheses with a bioinspired orientation. When the polymer melt was
isothermally injected from a point, into a thin disc, the (Giesekus and 3rd order fluid)
viscoelastic constitutive models provided the best description of the velocity and pres-
sure field. The (generalised Newtonian) Carreau-Bird model also provided an acceptable
description of the velocity field. The ratio of shear to extensional deformations was used
to predict the layered biaxial orientation of polystyrene cylinders which was found in the
point injected samples. To determine the optimal point at which to inject the valve,
vi
I used a combination of modelling (using the Carreau-Bird model and shear-extension
ratio) and bench-scale injection moulding. By simultaneously injecting at a point at the
centre of the free edge of each leaflet, a bioinspired orientation was produced.
The haemocompatibility of the shortlisted BCPs was tested with reference to expanded
polytetrafluoroethylene (ePTFE), polyester (PES), and glutaraldehyde-fixed bovine peri-
cardium (BP) graft materials using fresh human blood in modified Chandler loops. In
measures of coagulation and thrombogenicity the BCPs were better than PES, but worse
than ePTFE and BP. In measures of inflammation, the BCPs and ePTFE were better
than PES and BP. A heparin coating was applied to SEPS22 which made the BCP supe-
rior to all the other materials, in both categories. There was minimal loss of the coating
from the surface of SEPS22 after three weeks of dynamic straining in buffered solution.
Hence, heparin coated and uncoated SEPS22 would be suitable for use in cardiovascular
devices.
Degenerative aortic stenosis is caused by calcification. Calcification also threatens poly-
meric prostheses. The candidate BCPs were tested using a device which flexed the ma-
terials whilst they were immersed in calcifying solution. The BCP materials calcified
less than the glutaraldehyde-fixed pericardium reference, but calcification was shown to
increase the flexural stiffness and accelerate failure of the BCPs. Heparin coating of
SEPS22 improved its durability in both calcifying and non-calcifying solutions.
I tested the hydrodynamic performance and durability of the bioinspired spherical leaflet
form prostheses; the effective orifice area of the valve was 1.59 ± 0.02 cm2 (ISO 5840:
> 1 cm2); and regurgitation was 5.06 ± 0.61 ml/stroke (ISO 5840: < 10%). In terms
of hydrodynamics the polymeric prostheses were inferior to a reference bioprosthetic
valve. The polymeric prostheses survived 4× 106 cycles at 100 mmHg and 200 BPM
(ISO 5840: > 2× 108 cycles). The leaflets of the valve were thinner than specified (due
to inaccurate tooling), reaching 150 µm instead of 300 µm at crucial points. This was a
major contributor to the premature fatigue failure of the valve.
By extrapolating the fatigue prediction and using the results of finite element analysis,
I predicted that correctly manufactured polymeric valves could theoretically reach the
minimum ISO fatigue limit, indicating that there is potential for polymeric prostheses to
overcome the issues of durability and need for anticoagulation.
vii
Acknowledgements
I owe a great many thanks to a multitude of people. First and foremost, my supervisor Dr
Geoff Moggridge who has provided guidance and support throughout. I received funding
from the WD Armstrong fund, a travel grant from Queens’ College, and lab funding from
the British Heart Foundation (New Horizons grant NH/11/4/29059), without which this
project would not have been possible.
My thanks go to Drs Marta Serrani and Joanna Stasiak who have been invaluable col-
leagues, particularly with respect to their experience of finite element analysis and small
angle X-ray scattering. I am grateful to have had the polymer processing expertise of my
advisor, Dr Bart Hallmark.
I was fortunate to spend time at the laboratories of Prof. Maria Laura Costantino at
Politecnico di Milano, and Prof. Hans Pieter Wendel at Universita¨t Tu¨bingen, who were
generous with both their time and resources in support of my project.
In the lab, Francesco de Gaetano (Milano), Michaela Braun (Tu¨bingen), and Sergio
Mercado (Cambridge) all taught me a great deal and were unsparing with their time. I
constructed several items of equipment in this project which would not have been possible
without the patient help of the electrical and mechanical workshops, particularly Gary
Chapman.
I have had three fantastic years of research, which was very much aided by some great
colleagues and friends within the Department of Chemical Engineering and Biotechnology,
particularly those of the “M1 office”. The Sparrows, 308ers, sardines, Queensmen and
women, boys of Mussell Fort, and especially Cassi, you’ve made these years an awful lot
of fun.
I am indebted to my family for so much. Within the scope of this thesis, Simon’s scrupu-
lous proof reading, and Sarah and Zach’s delicious distractions.
Thank you everyone,
Jacob
viii
Preface
The results of Chapter 2 were presented at the Design of Medical Devices (DMD) con-
ference in Minneapolis, MN, April 2015, and subsequently published in the Journal of
Medical Devices, 2015; 9(2):020912-020912-2.
The results of Chapter 3 were presented at the European Society of Biomaterials (ESB)
conference in Liverpool, UK, September 2014; and published in the Journal of Materials
Science: Materials in Medicine, 27(2), 1-12: 10.1007/s10856-015-5628-7.
The results of Chapter 4 were presented, as a poster, at the Materials in Medicine Con-
ference, in Faenza, Italy, October 2013.
The results of Chapter 6 were published in Soft Matter journal, 2014, 10, 6077-6086. A
patent was also filed as a result of this work: Heart Valve, British Patent Application
1403454.0, filed February 2013.
Material from throughout this thesis was presented at PCR London Valves, London, UK,
October, 2014.
In this thesis I used, among many other techniques, small angle X-ray scattering (SAXS)
and finite element (FE) modelling. When explicitly noted, the SAXS measurements were
performed by Dr Joanna Stasiak, and the FE models were built and run by Dr Marta
Serrani, who both worked on the British Heart Foundation grant at the Department of
Chemical Engineering and Biotechnology, University of Cambridge, UK.
Contents
Declaration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii
Nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xix
1 Background 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Anatomical and physiological introduction . . . . . . . . . . . . . . . . . 2
1.3 Heart valve disease . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 Heart valve replacement . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.5 Cells, structure and function . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.6 From the inception of prosthetic heart valves to their current state . . . . 19
1.7 Design, development and testing methods . . . . . . . . . . . . . . . . . 28
1.8 Flexible leaflet polymer prosthetic heart valves . . . . . . . . . . . . . . . 29
1.9 Possible failure routes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
1.10 Bioinspiration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
1.11 Polymers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2 Polymer selection 47
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.2 Materials and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
2.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
2.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
2.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
3 Biocompatibility 87
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.2 Materials and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
3.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
ix
x Contents
3.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
4 Calcification 115
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
4.2 Materials and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
4.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
4.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
4.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
5 Valve design 133
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
5.2 Materials and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
5.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
5.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
5.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
6 Valve fabrication 167
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
6.2 Point injection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
6.3 Flow characterisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
6.4 Orientation modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
6.5 Controlling orientation in the injection moulded valve . . . . . . . . . . . 206
6.6 Final valve injection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
6.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
7 Valve testing 221
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
7.2 Hydrodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
7.3 Valve fatigue testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
7.4 Failure analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
7.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
8 Concluding remarks 241
8.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
8.2 Further work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
A Valve development flow chart 251
B Derivation of performance indices for material selection 255
Contents xi
C Construction of the final valve 257
D Cross model 259
E Mooney-Rivlin model 261
xii Contents
List of Figures
1.1 Functional anatomy of the heart. . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Cross sectional schematic of the aortic root. . . . . . . . . . . . . . . . . 3
1.3 Valve nomenclature. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Typical pressure and flow curves at the valves. . . . . . . . . . . . . . . . 6
1.5 Leonardo da Vinci’s sketches of flow in the sinuses of valsalva . . . . . . . 8
1.6 Leaflet cross section. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.7 Porcine valve stained for collagen. . . . . . . . . . . . . . . . . . . . . . . 17
1.8 Valve tissue mechanical properties. . . . . . . . . . . . . . . . . . . . . . 19
1.9 Bileaflet mechanical valve. . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.10 Bioprosthetic valve. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.11 Medtronic CoreValve transcatheter aortic valve implant. . . . . . . . . . 26
1.12 FE modelling of orientation in an anisotropic valve. . . . . . . . . . . . . 41
1.13 Change in strain energy density after the optimisation of orientation. . . 42
1.14 Chemical structures of common polymer groups. . . . . . . . . . . . . . . 44
2.1 Block copolymer phase diagram. . . . . . . . . . . . . . . . . . . . . . . . 49
2.2 Reversing the curvature of a leaflet. . . . . . . . . . . . . . . . . . . . . . 56
2.3 Dynamic mechanical analysis of SIBS30-L. . . . . . . . . . . . . . . . . . 59
2.4 Schematic and actual Ashby maps. . . . . . . . . . . . . . . . . . . . . . 60
2.5 Elastomer Ashby map. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
2.6 Typical σ-N curve. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2.7 Tensile testing samples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.8 Crack growth rate as a function of tearing energy (Lake, 1983). . . . . . 68
2.9 Compression moulding schematic. . . . . . . . . . . . . . . . . . . . . . . 70
2.10 Dimensions of samples used for mechanical testing. . . . . . . . . . . . . 70
2.11 Shink-induced cracking in SIBS30. . . . . . . . . . . . . . . . . . . . . . . 73
2.12 FTIR spectra of fresh and aged polymer samples. . . . . . . . . . . . . . 74
2.13 Mass loss after accelerated ageing. . . . . . . . . . . . . . . . . . . . . . . 74
2.14 Stress-strain plots of polymers in tension. . . . . . . . . . . . . . . . . . . 75
xiii
xiv List of Figures
2.15 Strain energy density of polymers in tension. . . . . . . . . . . . . . . . . 76
2.16 Crack tip position for SEBS20. . . . . . . . . . . . . . . . . . . . . . . . 78
2.17 Crack growth rate against strain energy density. . . . . . . . . . . . . . . 78
3.1 Schematic of the Chandler Loop. . . . . . . . . . . . . . . . . . . . . . . 92
3.2 SEM micrographs of block copolymers contacted with blood. . . . . . . . 100
3.3 Photograph and SEM micrograph of pericardium contacted with blood. . 101
3.4 Haemolysis and thrombin generation in blood. . . . . . . . . . . . . . . . 101
3.5 Platelet count, activation, and adsorption. . . . . . . . . . . . . . . . . . 103
3.6 Inflammatory markers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
3.7 Viability of L929 cells directly contacted with materials. . . . . . . . . . 106
3.8 Analysis of heparin coating before and after ageing. . . . . . . . . . . . . 108
4.1 Cross sectional detail of calcification device. . . . . . . . . . . . . . . . . 123
4.2 Surface calcium concentrations. . . . . . . . . . . . . . . . . . . . . . . . 126
4.3 SEM micographs of calcified samples. . . . . . . . . . . . . . . . . . . . . 127
4.4 Days to failure for samples in calcifying and non-calcifying solutions. . . 128
4.5 Change in flexural modulus after calcification. . . . . . . . . . . . . . . . 129
5.1 Axial view of leaflet edges of a tricuspid valve. . . . . . . . . . . . . . . . 137
5.2 Valve dimensions depend on positioning. . . . . . . . . . . . . . . . . . . 139
5.3 Open form of the valve of Thubrikar and Klemchuk (1990). . . . . . . . . 144
5.4 Cylindrical leaflet form valve. . . . . . . . . . . . . . . . . . . . . . . . . 146
5.5 Original and updated spherical form valves. . . . . . . . . . . . . . . . . 147
5.6 Cylindrical form valve surface construction dimensions. . . . . . . . . . . 148
5.7 Spherical form valve surface construction dimensions . . . . . . . . . . . 149
5.8 Schematic of pulse duplicator. . . . . . . . . . . . . . . . . . . . . . . . . 151
5.9 EOAG of prototype valves. . . . . . . . . . . . . . . . . . . . . . . . . . . 152
5.10 Regurgitation of prototype valves. . . . . . . . . . . . . . . . . . . . . . . 153
5.11 Energy loss over prototype valves. . . . . . . . . . . . . . . . . . . . . . . 154
5.12 Components of energy loss over prototype valves. . . . . . . . . . . . . . 154
5.13 Images during systole for cylindrical form valve. . . . . . . . . . . . . . . 155
5.14 Images during systole for spherical form valve. . . . . . . . . . . . . . . . 156
5.15 Images during diastole for spherical form valve. . . . . . . . . . . . . . . 157
5.16 Effect of leaflet thickness on opening of the spherical form valve. . . . . . 158
5.17 Spatial variation in leaflet thickness. . . . . . . . . . . . . . . . . . . . . 159
5.18 Photographs during diastole of spherical form valve with fixed posts. . . 160
5.19 EOAG when stent posts are fixed radially. . . . . . . . . . . . . . . . . . 160
List of Figures xv
5.20 Regurgitation when stent posts are fixed radially. . . . . . . . . . . . . . 161
5.21 Updated spherical form valve CAD. . . . . . . . . . . . . . . . . . . . . . 165
5.22 FE modelling of spherical form valve with two stent post designs. . . . . 165
6.1 Schematic of injection moulder. . . . . . . . . . . . . . . . . . . . . . . . 172
6.2 Schematic of SAXS setup. . . . . . . . . . . . . . . . . . . . . . . . . . . 174
6.3 SAXS pattern from a Z-axis beam and azimuthal integration for a point
injection moulded sample. . . . . . . . . . . . . . . . . . . . . . . . . . . 175
6.4 SAXS maps of three point injected discs. . . . . . . . . . . . . . . . . . . 175
6.5 SAXS pattern for an in-plane beam and schematic. . . . . . . . . . . . . 177
6.6 Schematic of radial flow problem. . . . . . . . . . . . . . . . . . . . . . . 178
6.7 BOY 22D injection moulder. . . . . . . . . . . . . . . . . . . . . . . . . . 181
6.8 Steady shear rate sweeps for SEPS22. . . . . . . . . . . . . . . . . . . . . 182
6.9 Steady shear sweeps at various temperatures. . . . . . . . . . . . . . . . 183
6.10 Pressure and screw position during injection moulding. . . . . . . . . . . 184
6.11 Experimental pressure vs. generalised Newtonian pressure predictions. . . 185
6.12 Oscillating rheometry sweep of SEPS22 melt. . . . . . . . . . . . . . . . . 186
6.13 Normal stresses during pulse shear. . . . . . . . . . . . . . . . . . . . . . 189
6.14 Normal stresses and viscosity during shear rate sweep. . . . . . . . . . . . 190
6.15 Experimental vs. Newtonian and 3rd order viscoelastic model pressures. . 191
6.16 Shear rate sweep data and results of the Giesekus model. . . . . . . . . . 193
6.17 Disc segment mesh. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
6.18 Velocity profiles for analytical and FE models. . . . . . . . . . . . . . . . 196
6.19 Shear and extensions rates in Giesekus and Carreau-Bird models. . . . . 202
6.20 Vectorial shear and extension predicted by the Giesekus and Carreau-Bird
models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
6.21 SAXS scan of PS cylinder orientation in a segment of a disc. . . . . . . . 204
6.22 Correctly predicted orientation. . . . . . . . . . . . . . . . . . . . . . . . 205
6.23 Schematic of stretch and shear at the critical value. . . . . . . . . . . . . 205
6.24 Correct prediction of orientation as a function of CsC , and the flow-thickness-
orientation space. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
6.25 Inflows and outflows for valve injection. . . . . . . . . . . . . . . . . . . . 208
6.26 FE valve mesh. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
6.27 Predicted and measured orientation at various depths in valves injected
from commissures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
6.28 Predicted and measured orientation at various depths in valves injected
from the free edge. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
xvi List of Figures
6.29 Predicted and measured orientation at various depths in valves injected
from the centre of the free edge. . . . . . . . . . . . . . . . . . . . . . . . 213
6.30 Modelling of final cavity filling. . . . . . . . . . . . . . . . . . . . . . . . 218
6.31 Final injection moulded valve. . . . . . . . . . . . . . . . . . . . . . . . . 219
6.32 SAXS mapping of leaflets of final valve. . . . . . . . . . . . . . . . . . . . 220
7.1 Images of failed valves. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
7.2 Mooney-Rivlin model and actual strain energy density. . . . . . . . . . . 227
7.3 von Mises stresses predicted by FE analysis during diastole. . . . . . . . 228
7.4 Maximum strain energy density in the valves as a function of applied pressure.229
7.5 Fracture mechanics and crack nucleation fatigue models. . . . . . . . . . 232
7.6 Crack growth rates for SEPS22 in solution. . . . . . . . . . . . . . . . . . 233
7.7 Predicted and measured valve fatigue life. . . . . . . . . . . . . . . . . . 234
7.8 Birefringence and SAXS of tensile bar samples. . . . . . . . . . . . . . . 237
7.9 Light microscopy of the surface of SEPS22 samples. . . . . . . . . . . . . 238
8.1 Bridging and looping in triblock copolymers. . . . . . . . . . . . . . . . . 249
A.1 Flow chart for the selection of a polymeric material for use in a prosthetic
heart valve. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
A.2 Flow chart for the design of a polymeric prosthetic heart valve. . . . . . . 253
B.1 Uniformly loaded beam schematic. . . . . . . . . . . . . . . . . . . . . . 255
D.1 Comparison of the Carreau-Bird and Cross model viscosities. . . . . . . . 260
List of Tables
1.1 Pressures upon and deformations of native and diseased valves. . . . . . . 9
1.2 Echocardiographic parameters for stenosis and regurgitation. . . . . . . . 10
1.3 Bioprosthetic heart valve deformations. . . . . . . . . . . . . . . . . . . . 23
1.4 Early attempts at flexible leaflet polymeric prostheses. . . . . . . . . . . 30
1.5 Later reports of polymeric prostheses. . . . . . . . . . . . . . . . . . . . . 33
1.6 Failure routes for prosthetic heart valves. . . . . . . . . . . . . . . . . . . 37
2.1 Cylinder forming triblock copolymers. . . . . . . . . . . . . . . . . . . . . 50
2.2 Shortlist of block copolymers and reference polyurethane. . . . . . . . . . 69
2.3 Tensile moduli of polymers. . . . . . . . . . . . . . . . . . . . . . . . . . 76
2.4 Summary of durability and flexural stiffness properties which yield the
performance indices. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
3.1 Haemocompatibility ranking of materials. . . . . . . . . . . . . . . . . . . 112
4.1 Ion concentrations in calcifying solution. . . . . . . . . . . . . . . . . . . 124
4.2 Survival of samples during immersed durability tests. . . . . . . . . . . . 129
5.1 Typical aortic root dimensions. . . . . . . . . . . . . . . . . . . . . . . . 138
5.3 Surfaces used to described trileafelet aortic valves and prostheses. . . . . 141
5.4 Design features for heart valve leaflets. . . . . . . . . . . . . . . . . . . . 145
5.5 Mooney-Rivlin parameters for strain energy density of SEPS22. . . . . . 152
5.6 Summary of hydrodynamic data for the prototype valves. . . . . . . . . . 161
6.1 Injection rates and sample thickness for point injected specimens of SIS30. 172
6.2 Giesekus model parameters. . . . . . . . . . . . . . . . . . . . . . . . . . 193
6.3 Bird-Carreau model parameters. . . . . . . . . . . . . . . . . . . . . . . . 194
6.4 Comparison of pressure drops recorded and predicted using analytical and
FE models. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
6.5 Processing conditions for injection of valves. . . . . . . . . . . . . . . . . 207
xvii
xviii List of Tables
7.1 Hydrodynamic valve performance and ISO standards. . . . . . . . . . . . 223
7.2 Results of quasi-real time durability testing of valves. . . . . . . . . . . . 225
7.3 Mooney-Rivlin parameters for anisotropic SEPS22. . . . . . . . . . . . . 226
7.4 Fracture mechanics growth parameters and effective flaw sizes. . . . . . . 231
C.1 Steps in the construction of the final spherical valve geometry . . . . . . 257
D.1 Cross model fitting parameters for SEPS22 at 150 ◦C . . . . . . . . . . . 259
E.1 Mooney-Rivlin parameters optimised for the tensile strain energy density
of SEPS22 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
Nomenclature
Acronyms and abbreviations
Symbol Description Units
(TPGd)10 Transvalvular pressure supported for 10% or more of
the cycle
mmHg
(TPGd)MAX Maximum diastolic transvalvular pressure difference mmHg
β-TG β-thromboglobulin
BSA Body surface area m2
FWHM Full width half maximum (of azimuthal integration of
SAXS data)
°
TPGd Transvalvular pressure gradient during diastole mmHg
b Block (as found in a block copolymer)
AAS Atomic Absorption Spectrometer
ADP Adenosine diphosphate
AFM Atomic force microscopy
ATR Attenuated Total Reflectance
BCP Block copolymer
BSA Body surface area m2
CH Corline Heparin
CNC Computerised numerical control
CTAD Citrate-Theophylline-Adenosine-Dipyridamole antico-
agulant
ECM Extracellular Matrix
EDX Energy dispersive X-ray
EOA Effective Orifice Area cm2
EPDM Ethylene propylene diene terpolymer
ePTFE Expanded polytretrafluoroethylene
ESC Environmental stress cracking
xix
xx Nomenclature
Symbol Description Units
ESRD External sewing ring diameter mm
EWF Essential work of fracture J m−2
FDA Food and Drug Administration
FE Finite Element
FSI Fluid structure interaction
FTIR-ATR Fourier Transform Infra-red spectroscopy with atten-
uated total reflectance mount
GAG Glycosaminoglycan molecules
GOA Geometric orifice area cm2
HAp Hydroxyapatite
IC Interstitial cells
IOD Internal orifice diameter mm
ISO International standards organisation
LCA Left coronary artery
LLDPE Linear low density polyethylene
LVOT Left ventricular outflow tract
MRI Magnetic resonance imaging
PDMS Polydimethylsiloxane
PES Polyester
PET Polyethylene terephthalate
PMN-Elastase Polymorphonuclear lastase
PS Polystyrene
PU Polyurethane
PVC Polyvinylchloride
QALY Quality Adjusted Life Year years
RCA Right coronary artery
SAXS Small angle x-ray scattering
SBS Polystyrene-block -polybutadiene-block -polystyrene
SC Solvent cast
SC5b-9 Serum complement 5b-9
SE Stent-leaflet boundary
SEBS Polystyrene-block -polyethylene/polybutylene-block -
polysyrene
SEM Scanning electron microscopy
SEPS (Poly)styrene-b-ethylene-propylene-b-styrene
Nomenclature xxi
Symbol Description Units
SI/BS (Poly)styrene-b-ethylene-butadiene-b-styrene
SIBS (Poly)styrene-b-isobutylene-b-styrene
SIS (Poly)styrene-b-isoprene-b-styrene
TA Texture analyser
TAD Tissue annulus diameter mm
TAT Thrombin-antithrombin complex
TAVI Transcatheter aortic valve implants
TE Tissue engineered
vWF von Willebrand Factor
XPS X-ray Photoelectron Spectroscopy
XRD X-ray powder diffraction
xSIBS crosslinked polystyrene-block -polyisobutylene-block -
polystyrene
.
Dimensionless groups
Symbol Description Units
Cs Ratio of shear rate to stretch rate
Cv Ratio of vector shear rate to vector stretch rate
CsC Critical ratio of shear rate to stretch rate
CvC Critical fatio of vectorial stretch rate to shear rate
De Deborah number
R Dimensionless radius
Re Reynolds number
Symbols - Greek
Symbol Description Units
σ Total stress tensor Pa
τ Shear stress tensor Pa
τE Viscoelastic component of stress tensor Pa
τ V Purely viscous component of stress tensor Pa
χ Polymer-solvent interaction, or Flory-Huggins param-
eter
xxii Nomenclature
Symbol Description Units
ε˙ Extension rate s−1
 Engineering strain
flex Flexural strain
η Viscosity Pa s
η∞ Infinite shear rate viscosity (Bird-Carreau model) Pa s
ηv White-Metzner viscosity Pa s
η0W Zero shear rate viscosity in Cross model
ηb0 Zero shear rate viscosity (Bird-Carreau model) Pa s
λ Wavelength m
λb Natural time (Bird-Carreau model) s
λr White-Metzner relaxation modulus s
λs Stretch ratio, where s corresponds to a vector direction
µ Dynamic viscosity Pa s
∇ Tensor gradient
∇· Tensor divergence
ρ Density kg m−3
σ Engineering stress Pa
σ∗ Tear rupture stress Pa
σc Commissural stress, per unit length Pa m
−1
σcov Covariance
σf‖ Fatigue stress with principal stress parallel to the fibre
direction
Pa
σflex Flexural stress Pa
τW Critical stress level for transition to shear thinning in
Cross model
Pa
θ Angle in radial coordinate system
θc Half angle between the leaflets at the commissure °
~γv Shear vector
~εv Stretch vector
Symbols - Latin
Symbol Description Units
I Identity matrix
∆Pm Mean pressure gradient mmHg
Nomenclature xxiii
Symbol Description Units
∆Pmax Peak pressure gradient mmHg
~n Surface normal
~v Velocity vector m s−1
A Crack area m2
a Mobility parameter in Giesekus equation
a0 Initial flaw size m
Ae Pre-exponential factor in Arrhenius rate equation mol s
−1
af Critical flaw size for failure m
am Length of monomer unit m
B Power-law crack growth model coefficient
b Half-thickness of the disc m
bf Flexural sample width m
c Crack length m
dcom Commissural spacing, normal to reflection lines mm
Ea Activation energy J mol
−1
Eflex‖ Modulus when the contours of constant displacement
are parallel to the stiff cylinders
Pa
Eflex⊥ Modulus when the contours of constant displacement
are perpendicular to the stiff cylinders
Pa
F Power-law crack growth model exponent
f Fraction of monomers A in a block copolymer chain
Frev Maximum force for curvature reversal N
h Unstrained length of crack growth sample m
I‖ Material performance index when leaflet opening is
improved by anisotropy
m4 N−2
IP Material performance index for polymers in prosthetic
heart valves
m4 N−2
IW Material performance index parameter for the design
of a uniform cross-section beam or wing
I⊥ Material performance index when leaflet opening is
not improved by anisotropy
m4 N−2
k Reaction rate constant moln s−1
kg Geometry related constant for tearing energy in a ten-
sile bar
L Strained length m
xxiv Nomenclature
Symbol Description Units
L0 Unstrained length m
lc Characteristic length m
Ls Length of s, where s is defined in the subscripts section mm
m Power law fluid consistency index
mλ Viscoelastic relaxation modulus power law consistency
factor
mv Viscoelastic power law consistency factor
N Polymerisation index
n Power law fluid power index
nλ Viscoelastic relaxation modulus power law index
nb Power law index (Bird-Carreau model)
nv Viscoelastic viscosity power law index
nW Cross model power law index
P Pressure Pa
PH Hydrostatic pressure Pa
Q Flow rate m3 s−1
R Gas Constant J K−1 mol−1
r Radius m
Rf Regurgitant fraction
rp Pressure measurement radius Pa
T Absolute temperature K
t (Leaflet) thickness m
Te Energy release rate or tearing energy N m
−1
Tm Melting temperature K
TODT Order-disorder transition temperature K
U Total internal energy J m−3
V Volume m3
Vc Closing volume ml
Vl Leakage volume ml
vm Mean velocity m s
−1
Vrev Retrograde flow in one cycle ml
Vsv Stroke volume ml
W Strain energy density in test piece Pa
z Axial position in radial coordinate system
Nomenclature xxv
Symbol Description Units
L Denotes stretching parallel to the PS domain orienta-
tion
R Denotes stretching perpendicular to the PS domain
orientation
Subscripts and superscripts
Symbol Description Units
A Aortic/distal side of the valve
c At the commissure
f At edge of disc
V Ventricular/proximal side of the valve
cc−aa From the coronary arteries to the aortic annulus
max Maximum
min Minimum
stj Sino-tubular junction
TAD (At the) tissue annulus (diameter)
xxvi Nomenclature
Chapter 1
Background
1.1 Introduction
The first heart valve prostheses were developed and implanted in humans during the
1950s (Bodnar and Frater, 1992): since then ten of millions of lives have been trans-
formed worldwide. Nonetheless, research and development of heart valve prostheses re-
mains an important public health endeavour. The current gold standards for heart valve
replacement are not a complete solution to heart valve disease: they leave patients with a
reduced lifespan or quality of life compared to healthy patients of the same age (D’Arcy
et al., 2011). There are two main types of prosthesis - biological and mechanical valves.
Those from biological materials allow patients to lead a relatively normal life, not requir-
ing anticoagulant therapy, but have a limited durability of up to approximately 15 years.
Mechanical valves rarely fail in a human lifetime, however they require lifelong anticoag-
ulant therapy which increases the risk of fatal bleeds. This thesis examines whether it
may be possible to overcome these 2 issues.
This introduction will first provide a general description of heart valve anatomy and
physiology before reviewing the pathologies which lead to native heart valve failure. I
will consider the extent of heart valve disease and replacement before turning to the
prosthetic implants which have been developed to date. I will perform a critical review of
the prosthetic valves that have achieved clinical success, as well as those that have fallen
by the wayside, setting a context for the research that is presented in this thesis.
1
2 Chapter 1. Background
1.2 Anatomical and physiological introduction
Valves are abundant throughout the cardiovascular system, and lead to unidirectional flow
(an observation that led William Harvey to propose the theory of circulation in 1617).
All valves in the circulation are passive; they do not require any muscular activation
to function. The four cardiac valves are located at the entry and exit of the cardiac
chambers, as illustrated in Figure 1.1.
Figure 1.1: Functional anatomy of the heart showing the four cardiac valves. Taken from
Medicore (2012).
The heart provides the pressure head to circulate blood throughout the body. The left side
of the heart receives oxygenated blood which is pumped to the body, while the right side
pumps the returning deoxygenated blood to the lungs. A cardiac cycle commences with
atrial contraction, forcing blood through the mitral and tricupsid valves, and contributing
around 10% to the final ventricular volume at resting heart rates, and up to 40% at high
heart rates. Blood is then forced from the ventricles when they contract during systole,
opening the aortic and pulmonary valves, and closing the mitral and tricuspid valves.
The ventricles passively dilate during diastole causing the aortic and pulmonary valves
to close, while the mitral and tricuspid valves open to allow the ventricles to refill. To
Chapter 1. Background 3
recommence the systolic phase, the atria contract.
LTAD
RCA
LCA
Lstj
LVOT
Sinuses
A
scen
d
in
g
aorta
Mitral valve
Lca−aa
Figure 1.2: Cross sectional schematic of the aortic root. RCA and LCA indicate the
right and left coronary arteries respectively, the non coronary leaflet is not shown. LTAD
indicates the diameter of the aortic annulus (TAD), which occurs at the narrowest point
in the root, near to the nadirs of the leaflets. Lstj is the diameter at the sino-tubular
junction, Lca−aa is the distance between the aortic annulus and the coronary arteries.
LVOT refers to the left ventricular outflow tract.
The work presented in this thesis concentrates on the development of aortic prostheses
and as such the discussion of anatomy and physiology is focussed on this valve. The
reasons behind this choice are expanded upon in Section 1.3 and 1.4. As depicted in
Figure 1.2, the aortic valve is located in the aortic root where the aorta joins the left
ventricle. The proximal end (nearest to the heart) of the root is the transition from
muscular cardiac tissue to fibrous arterial tissue. This join is termed the annulus, and
occurs near to the narrowest point in the root which is termed the true annulus. This
annulus has a diameter from 18 to 30 mm, which roughly scales with body size. The
aortic root then expands into 3 sinuses, before contracting to the sino-tubular junction
where it joins the aorta.
The nomenclature used to refer to locations on both the native and prosthetic valves is
displayed in Figure 1.3a. The aortic valve contains 3 crescent shaped leaflets, or cusps,
4 Chapter 1. Background
Sewing ring
Stent post
Nadir
Leaflet belly
C
R
Nodule of
arantius*
Free edg
e
Stent cylinder
Commissure
Stent-leaflet
boundary
Distal
Proximal
(a) (b)
Figure 1.3: (a) shows the nomenclature used to describe the parts of a prosthetic valve.
The nodule of Arantius (where the collagen bundles meet) is only found on the native
leaflet, not on a prosthetic valve. All positions on this prosthetic leaflet are also found
on the native leaflet. C and R refer to the circumferential and radial directions on the
leaflet. (b) is from (Ho, 2009), images A and B show a dissected human heart with the
left (L), right (R), and non-coronary (N) cusps of the aortic valve marked. In image C,
the aortic valve has been opened longitudinally. The open arrows indicate the location
of the sino-tubular junction, the small arrows indicate the coronary artery orifices and
the dotted line marks the fibrous tissue separating the mitral and aortic valve. With
permission from Oxford University Press.
which are joined to the sinuses along their proximal sides, and have a distal free edge.
The most proximal point on the leaflet-sinus boundary is the leaflet’s nadir, while the
central point of the free edge is the nodule of Arantius1. The vertices where the free edge
meets the aorta are called commissures.
The cusps are referred to as the right coronary, left coronary and non-coronary cusps;
named according to the artery to which they are nearest. The nadir of each leaflet of the
valve is at the level of the annulus, while the commissures of each leaflet are approximately
15 mm distally. The coronary arteries are found in the sinuses and are the first blood
1Name after the prominent 16th century anatomist Giulio Cesare Aranzi, from the University of
Padua.
Chapter 1. Background 5
vessels to depart from the artery. The position of the aortic valve leaflets within the aortic
sinus relative to the intricacies of the anatomy is shown in Figure 1.3b. In particular,
note the proximity of the left ventricular outflow tract (LVOT) to the coronary ostia in
Figure 1.2, a prosthetic valve must not obstruct either structure.
A typical resting heart rate is 70 bpm, with a cardiac output of 5 l min−1. This equates
to approximately, 40 million cycles each year, or 3 billion cycles in a lifetime: hence any
implant which dares to imitate the native organ must have remarkable resilience. At rest,
in general, the systolic phase lasts 1/3 of the cycle period and the diastolic phase for the
remaining 2/3. Roughly 70 ml are ejected from both sides of the heart during each stroke
(the stroke volume). The pressure in the aorta oscillates between 120 and 80 mmHg in
a normal, healthy heart. Typical pressure and flow rate profiles are presented in Figure
1.4. Qualitatively, the cardiac cycle consists of the following phases:
1. Atrial contraction: completes the filling of the ventricles, up to ∼ 120 ml each. The
mitral and tricuspid valves close.
2. Isovolumic contraction: the start of ventricular contraction causes the ventricular
pressure to increase. The phase finishes once the aortic and pulmonary valves are opened,
and blood starts to flow.
3. Systole: the ventricles continue to contract causing rapid ejection of blood into the
aorta and pulmonary artery. The maximum flow rate (∼ 20 l min−1) is rapidly reached,
and the diminishing flow continues until the minimum ventricular volume (∼ 50 ml each)
is reached, and the aortic and pulmonary valves close.
4. Isovolumic relaxation: All 4 valves are closed. The ventricles relax, and the ventricular
pressure falls.
5. Diastole: The mitral and tricuspid valves open, and the relaxed ventricles are filled
with blood. Ventricular filling is completed by the contraction of the atria as step 1
resumes.
1.2.1 Valvular biomechanics and haemodynamics
The Reynolds number, Re, is used to describe fluid flow around an object or through a
vessel. Re is the dimensionless ratio of inertial to viscous forces in a flow:
Re =
ρvmlc
µ
(1.1)
6 Chapter 1. Background
Figure 1.4: Typical pressure and flow curves for aortic (AV) and mitral (MV) valves.
ECG is the trace from the electrocardiogram, upon which the QRS complex, T, and P
waves are indicated. Taken from Bronzino (2000).
where ρ is the density, vm is the mean velocity, µ is the dynamic viscosity of the fluid, and
lc is the characteristics length within the flow, which, in this case may refer to the aortic
diameter. The systolic flow through the open aortic valve has a Reynolds number between
1500 and 5000. A Reynolds number up to 2100 describes laminar flow, while above 4000
the flow is turbulent. Turbulent flow is characterised by small random variations in
velocity and pressure, rapid mixing, and the presence of vortices and high shear forces.
In the transition region, for flow in a smooth pipe, elements of both flow types may be
present.
High shear forces in blood lead to the release of adensoine diphosphate (ADP) from
platelets, causing them to aggregate and blood clots to form; a process which receives
further attention in Chapter 3. Importantly, as the flow through the valve may be
turbulent, small changes to the flow field and geometry can trigger clot-forming turbulence
(Yoganathan et al., 2005).
Chapter 1. Background 7
The time averaged 5 l min−1 blood flow through the aortic root is far from having a
steady laminar profile. The fluid is initially stationary in the ventricle, so flow is far
from developed. Also, blood does not behave as a Newtonian fluid, so a parabolic flow
profile is unlikely. Considering the anatomy, ventricular contraction extends to the aortic
annulus, while the sinuses dilate under pressure, even before reaching the arch of the
aorta. The geometry is not a uniform cylinder and is highly dynamic. Overall, these
factors result in a relatively flat flow profile upstream of the valve (observed in vivo using
3D magnetic resonance imaging (MRI)), which is slightly skewed (10%) towards the septal
side (Balachandran et al., 2011). In healthy individuals the maximum velocity through
the aortic valve at rest is approximately 1 m s−1, as measured by pulsed Doppler during
cardiac surgery (Mathison et al., 1988).
The opening of the heart valve is driven by the pressure difference between the ventricle
and aorta. The curvature of each leaflet must reverse in order to open. This reversal
starts at the central ‘belly’ of each leaflet, and all 3 leaflets generally open simultaneously
(Thubrikar and Klemchuk, 1990). Closure of the heart valve is more complex. The
valve sits at the base of the aortic sinuses, the expansion in diameter at the sinuses leads
to interesting haemodynamics. Investigation into flow in the aortic sinuses has a long
history; between 1508 and 1513 Leonardo da Vinci completed accurate drawings of a
pig’s aortic root, recorded in Quadratica Anatomica, an example of which is shown in
Figure 1.5. Noticing the peculiar form of the sinuses, da Vinci experimented further by
constructing a glass model of the root using a plaster cast. Then, he “let the water that
strikes there have millet or fragments of papyrus mixed with it so that one can see the
course of the water better from their movements”. da Vinci observed that the swirling
vortices that are generated in the sinuses behind the valve’s open leaflets contribute to
the closing of the valve. This observation - which is probably the first recorded example
of particle image velocimetry - has been corroborated using modern in vitro and in silico
methods (Gharib et al., 2002; Robicsek, 1991).
When the valve leaflets are closed they support a large pressure differential. This causes
stresses and strains within the leaflets, which are supported by the aortic wall. All living
cells respond to mechanical stimuli in some way, and those cells that make up cardiac
valves are no different. Cell deformation - and any changes which result - may be causative
of, or associated with, disease. Thubrikar et al. (1980) made in vivo measurements of
leaflet strain (mostly on dogs using fluoroscopic markers). Based upon excised valves, they
assumed that the leaflets were cylindrical, with a thickness of 0.4 to 0.5 mm, from which
they estimated the tissue modulus in vivo. The results were verified with measurements
of the moduli on the excised valves. These results are shown in Table 1.1. It should be
8 Chapter 1. Background
Figure 1.5: Detail of Leonardo da Vinci’s sketches of flow in the sinuses of valsalva. From
RL19117 verso, with permission from Royal Collection Trust / ©Her Majesty Queen
Elizabeth II 2016.(Gharib et al., 2002).
noted that the valve leaflet is a composite structure (discussed further in Section 1.5) and
so the listed stress is not homogeneous throughout the leaflet.
1.3 Heart valve disease
Valves are responsible for the unidirectional flow of blood and so there are two ways in
which this can be compromised. If a valve is unsuccessful at restricting reverse flow, it
is described as regurgitant. This may be caused by failure to maintain a tight seal when
closed during diastole, or by being too slow to close. Alternatively, the valve may become
stenotic, meaning the open orifice area has decreased which disrupts forward flow. This
disruption is generally caused by stiffening or thickening, referred to as sclerosis, of the
valve’s leaflets which stop it from opening fully. Heart valve disease is defined as the
presence of either regurgitation or stenosis in any of the valves which compromises the
cardiovascular system.
It is quite possible to live a perfectly normal life and be totally unaware of the first signs of
heart valve disease. In fact, the prevalence of valve disease based upon clinical suspicion
is considerably lower than the prevalence based upon echocardiographic measurements
(Iung and Vahanian, 2011). Normally, the first signs that a valve is defective will be a
heart murmur, which can be diagnosed via auscultation. However, a heart murmur may
be present in the absence of any other symptoms, and may not always indicate heart
Chapter 1. Background 9
Table 1.1: Summary of the pressures exerted on, and the deformations found in native
and diseased heart valves, adapted from Balachandran et al. (2011) and ISO 5840. The
radial and circumferential directions are declared in Figure 1.3a.
Mechanical parameter Normal
Diseased
(very severe)
Notes
Resting aortic pressure
during diastole (max.)
65 to 85
mmHg
≥120
Resting aortic pressure
during systole (max.)
100 to 130
mmHg
≥210
Differential pressure
across closed valve (max.
during diastole)
95 mmHg 185
Geometric area 3 to 4 cm ≤2 (Garcia and
Kadem, 2006)
Diastolic tensile strain
(circumferential)
9-11% ≥15%
Measured in
Porcine aortic
valve (Yap et al.,
2010)
Diastolic tensile strain
(radial)
13-25% 15-31%
Measured in
Porcine aortic
valve (Yap et al.,
2010)
Diastolic membrane stress 2.4 kPa
Measured in dogs
(Thubrikar et al.,
1980)
Systolic membrane stress 0.167 kPa
Measured in dogs
(Thubrikar et al.,
1980)
valve disease.
Unfortunately, many effects of heart valve disease are not as benign as a heart murmur.
Heart valve disease can lead to heart failure, where death is normally a result of stroke,
blood clots, or sudden cardiac arrest. A typical heart failure patient will present with
shortness of breath, unusual fatigue, and swelling of the ankles, feet, legs, abdomen, and
veins in the neck. Heart valve disease may also lead to chest pain upon exertion, irregular
heart beats, or dizziness and fainting. Surgical intervention is the only therapy which can
restore a normal way of life. No other medical treatments currently exist which prevent or
reverse the physical reality of heart valve disease, however the effects and symptoms can
often be mediated by lowering the load on the heart, preventing arrhythmias, and taking
anti-clotting therapies (National Heart Lung and Blood Institute, 2011). A definitive
diagnosis of heart valve disease is normally based upon an echocardiogram which indicates
10 Chapter 1. Background
the incomplete opening or closing of the valve. Table 1.2 shows the clinical definition
of severe aortic valve stenosis and aortic regurgitation based upon this data. If left
untreated, aortic stenosis has a 5 year mortality of 88.4% with a mean survival of 1.8
years (Clark et al., 2012).
Table 1.2: European Society of Cardiology (ESC) and the European Association for
Cardio-Thoracic Surgery (EACTS) guidelines for echocardiographic evaluation of severe
aortic stenosis and aortic regurgitation (Vahanian et al., 2012)
Severe aortic stenosis
Valve area (cm2) ≤ 1.0
Indexed valve area (cm2 m−2 body surface area
(BSA))
≤ 0.6
Mean gradient (mmHg) ≥ 40
Maximum jet velocity (m s−1) ≥ 4.0
Velocity ratio (= VLV OT
VAV
) ≤ 0.25
Aortic regurgitation
Effective regurgitant orifice area (mm2) ≥ 30
Regurgitant volume (ml/beat) ≥ 60
Vena contracta width (mm) ≤ 6
1.3.1 Aetiology and epidemiology
Heart valve disease can manifest itself from birth through to old age, though it has
differing aetiologies depending on age. Congenital deformities associated with heart valves
typically present with biscuspid aortic valves (just 2 leaflets), which occurs in 0.6-0.8%
of males and 0.2% of females. While a bicuspid valve will function with little cause for
alarm, the deformity results in a highly elevated risk of aortic stenosis, which manifests
itself with age (Iung and Vahanian, 2011). More rarely there are cases of unicuspid valves,
which also result in stenosis. Quadricuspid valves are found even more infrequently, and
result in aortic insufficiency (Thubrikar and Klemchuk, 1990). Occasionally, heart valve
disease may be caused by infective endocarditis (inflammation of inner lining of the heart),
generally from Staphylococcus, or other inflammatory causes. This accounts for 0.9% of
aortic stenoses and 11.6% of aortic regurgitation cases, corresponding to 31 cases per
million population (Iung and Vahanian, 2011).
For most of the 20th century rheumatic fever was the principle cause of heart valve failure.
Rheumatic fever is caused by a group A streptococcal infection which results in a chronic
and degenerative disease whose greatest effect is the growth of excess connective tissue
Chapter 1. Background 11
on the heart valves (fibrosis). Currently, over 15 million people are estimated to be
affected by rheumatic heart disease, the vast majority in developing countries. A large
proportion will require heart valve surgery which is unlikely to be available to them.
The near eradication of rheumatic fever in developed countries can be attributed to
improvements in hygiene and availability of antibiotics (World Heart Federation, 2012),
and hence strategies for eradication of this form of heart valve disease in developing
countries are predominantly political and economic issues. For those who have suffered
from rheumatic fever, and will develop heart valve disease, heart valve replacement is
the only solution. Although the opportunities for cardiac surgery are limited, it is not
impossible to have a heart valve replaced in developing countries. Cardiac centres can be
found in 18 African countries, and in 2005, over 16,000 cardiac operations took place in
Africa. As cardiac surgery is so closely tied to economic growth, the number of operations
is expected to increase (Edwin et al., 2011; Pezzella, 2005). Currently, mechanical valves
are more commonly implanted due to their durability and affordability. However, the lack
of availability of anticoagulants and poor patient compliance means that the implantation
of mechanical valves carries a significant risk of a thrombotic event occurring. As such
there exists a need for heart valve prostheses which do not require anticoagulation, while
being more durable and more affordable than bioprostheses (Zilla et al., 2008).
Despite the decrease in heart valve disease resulting from rheumatic fever, the overall
incidence of heart valve disease continues to rise. This rise can be attributed to the
degenerative form of heart valve disease, which is now the most prevalent type. The
incidence of valvular disease is ≤ 2% before the age of 65, but rapidly rises to 8.5%
between 65 and 75, and further increases to 13.2% above the age of 75. There is an age
adjusted prevalence of 2.5%, corresponding to approximately 1.6 million people suffering
from valvular heart disease in the UK (Iung and Vahanian, 2011). In the United States,
of the 2.4 million deaths in 2006, 20,260 were attributed to valvular diseases, with aor-
tic disease accounting for 62.5%, and mitral valve disorders 12.6% (Heron et al., 2009;
Rosamond et al., 2008). The burden of valvular disease is projected to grow. In the UK
for example, the number of people older than 75 years is projected to increase by 50%
by 2025 (Malhotra, 2012), leading to the pronouncement by D’Arcy et al. (2011) that
“valvular disease is the next cardiac epidemic”.
Aortic stenosis is often referred to as a “degenerative” pathology as it shows a strong
correlation with age, however the mechanism associated with valvular diseases is poorly
understood. It is clear that smoking, hypertension, hyperlipidaemia, hypercholestraemia
and diabetes are the greatest risk factors (D’Arcy et al., 2011).
The distribution of valvular disease type is dominated by pathologies of valves of the
12 Chapter 1. Background
left side, which could be attributed to the higher pressures found on this side. The Euro
Heart Survey - which is dominated by degenerative cases - reported that 44.3% of valvular
disease was aortic, 34.3% mitral and 20.2% mixed. And within this breakdown, three
quarters of disease, for both valves, is associated with the systolic phase - aortic stenosis
and mitral regurgitation (Iung, 2003).
1.3.2 Pathological mechanisms and biomechanics
Biomechanics is a major player in the initiation and progression of aortic valve patholo-
gies. Aortic stenosis results from the incomplete opening of the heart valve leaflets which
is caused by a mechanical stiffening of the leaflets. Most typically, this stiffness is caused
by the deposition of calcium phosphate crystals in a process termed calcification. Specif-
ically, this pathology is termed calcific aortic valve disease. It is tempting to draw paral-
lels between stiffening and blocking of arteries in atherosclerosis, and valve calcification;
both display similar inflammation and lead to the release of cytokines which trigger bone
morphogenic proteins and cause osteogenic cell phenotypes to develop. However, statin
therapy does not halt calcification of the valves and may in fact accelerate it; the evidence
is unclear as to whether it is at all preventative (Goldbarg et al., 2007).
Aortic stenosis leads to a significant change in haemodynamics, which is manifested in 2
ways; firstly, an increased resistance to flow, and secondly a change in flow profile. While
a healthy valve may have a peak axial velocity of 1.20 m s−1, a severely diseased valve
may reach 7 m s−1. This elevation is sufficient to raise shear stresses, causing rupture of
blood cells and damage to the endothelial cell layer which coats the aorta (Balachandran
et al., 2011).
The mechanisms that lead to aortic regurgitation are less clear. While the majority of
cases are due to rheumatic fever, a study of 756 patients in the developed world found
that 40% of cases had a mixed or unclear cause, 29% were degenerative, 10% occurred in
patients with a bicuspid valve, 11% in patients with an aortic root pathology and 10%
due to endocarditis (Varadarajan et al., 2013). Inflammation and scarring is present in all
these pathologies, which lead to abnormal valve mechanics. For example, dilation of the
aortic root may reduce the contact area between the leaflets during diastole (coaption),
or scar tissue may lead to stress concentrations and subsequent tears in the valve.
Degeneration and ageing are a common cause of both stenosis and regurgitation, but what
does ageing do to heart valve leaflets? Christie and Barratt-Boyes (1995) reported that the
strain, at fixed stress, in leaflets collected from allograft donors decreased approximately
4%/year from age 15 to 25, and at 1%/year until the age of 58 - and presumably thereafter.
Chapter 1. Background 13
The authors concluded that stiffening of the tissue reduced the performance, due to a
reduced coaption area, however they were unable to say whether the change was due to
calcification, or if other physiological factors were at play.
1.4 Heart valve replacement
In any patient receiving a diagnosis of severe aortic stenosis (likely via echocardiogram)
and who is also symptomatic, aortic valve replacement is indicated. In asymptomatic
cases, aortic valve replacement may be indicated by the presence of a low systolic ejec-
tion fraction, abnormal exercise test, or when undergoing a coronary artery bypass graft
(Vahanian et al., 2012).
Worldwide, between 275,000 and 370,000 valve replacements take place annually (Zilla
et al., 2008). In the UK in 2010 there was an incidence of 240 valvular procedures
per million population (Society for Cardiothoracic Surgery in GB and Ireland, 2011).
Considering the observed prevalence of heart valve disease, it is estimated that roughly
1% of those who suffer from valvular heart disease actually receive a replacement valve.
This low fraction is supported by results from the Helsinki ageing study (Lindroos et al.,
1993), 476 were aged 75 to 85 years, 14 of whom were diagnosed with severe aortic valve
disease (3%). Although 8 of these were symptomatic and eligible for valve replacement,
just 2 opted for valve surgery. This opt-for fraction seem to be relatively constant in
studies worldwide, and is likely to be because many patients are otherwise healthy and
surgery is traumatic.
Although the worldwide number of heart valve procedures has remained relatively stag-
nant for the past decade, the number of procedures is expected to increase dramatically
in the coming decades; this rise will be predominantly driven by an increase in access
to cardiac surgery. Cardiac surgery is available to the majority of European and North
American patients, but just 10% of the Indian and Chinese populations have had access
over the past decade. Africa is still poorly served: there are 112 times as many cardiotho-
racic surgeons per capita in Europe as Africa (Zilla et al., 2008). Considering the overall
picture, Yacoub and Takkenberg (2005) predicted that worldwide, valve replacements
would triple from 290,000 in 2003 to over 850,000 by 2050.
Over 85% of valve replacements are performed using open heart surgery. First time
surgery for an aortic valve replacement carries a mortality risk of 2% (Society for Cardio-
thoracic Surgery in GB and Ireland, 2011). First reoperation increases that risk to 10%
(Jones et al., 2001), and further reoperations carry even higher risks. This is a highly
14 Chapter 1. Background
traumatic surgery. Under general anaesthesia, an incision is made down the chest, the
sternum is sawn open, the pericardium opened and the patient is placed upon a heart-lung
bypass machine. The heart is stopped, the aorta opened, the native valve excised, and
the prosthetic valve implanted. The incisions through the aorta and pericardium must
be closed, the patient brought off the bypass and the heart restarted. In an octogenarian
population heart valve replacement may require approximately 5 days in the intensive
care unit, and a 2 week hospital stay (Kolh et al., 2007). However, we should recall the
benefits: once a patient presents with angina, syncope or dyspnea, only 50% survive to 2
years. Without valve replacement, 75% of patients will be killed by aortic valve disease.
Survival after replacement is nearly equal to average survival rates when adjusted for age,
until valve degradation becomes an issue after 10 years (Carabello, 2002).
1.4.1 Selection of valve
For several reasons, the focus of this thesis is on the aortic valve. Firstly, it is the valvular
disorder with the highest mortality risk. Furthermore, age-related valvular disease is often
caused by calcification which leads to stenosis, while the lingering effects of rheumatic
fever (which causes regurgitation) become rare; the ratio of aortic stenosis to mitral
regurgitation is expected to increase in an ageing population. Secondly, aortic prostheses
are occasionally used as mitral valve replacements, however the converse does not occur.
Finally, mitral disease is dominated by regurgitation, for which some interventions do not
use complete prosthetic valves. Instead, a surgical or transcatheter repair is performed
in which the native leaflets are tightened.
1.5 Cells, structure and function
The aortic valve is passive - opening and closing is driven purely by fluid forces - and
hence it contains no muscular tissue. The principal structures of the valve are the fibrosa
and ventricularis, which are separated by the spongiosa layer, as shown in Figure 1.6.
These three layers have five principle constituents: collagen, elastin, glycosaminoglycans
(GAGs), endothelial cells and interstitial cells. All structural and constituent types play
a specific role and can shed light on ways in which synthetic implants might cope with
the biological, chemical, and mechanical environment.
Proceeding from the ventricular side of the leaflet, the first structure is appropriately
named ventricularis. This layer contains both elastin fibres and collagen bundles, however
Chapter 1. Background 15
Figure 1.6: Cross section through a leaflet of the aortic valve. From USC Cardiovascular
Thoracic Institute.
this surface of the leaflet is smooth, unlike the fibrosa, which promotes laminar flow over
the open valve.
The central layer is the spongiosa. This consists primarily of glycosaminoglycan molecules
(GAG). Various functions have been proposed for this layer. It is widely accepted that it
provides a form of lubrication between the ventricularis and the fibrosa allowing them to
move past each other with ease. The spongiosa contains many, well ordered fibroblastic
cells which produce collagen for the valve. The fibrosa is found on the aortic side, which is
the thickest layer in the valve. This layer is predominantly composed of collagen bundles.
The fibrils of collagen are neatly arranged, and are lubricated by proteoglycans which
contain one of more covalently bonded GAG chains.
The blood contacting surfaces of the valve and blood vessel are coated with endothelial
cells. These cells line the cardiovascular system and are the barrier between blood and the
extracellular matrix (ECM). The endothelial cell has multiple roles; as the barrier between
the blood and the rest of the body it holds the keys to permeation, while taking part
in the immune response, coagulation, growth regulation, production of ECM (Sumpio
et al., 2002), and the modulation of blood flow and blood vessel tone. In fact, no other
material or tissue is known to interact with blood in the same way as endothelial cells do.
Other cell types, tissues, natural and synthetic materials, including of course polymers,
will lead to a changes in the blood, generally for the worse.
Cardiac valves contain cells which are phenotypically unique and with a structure not
found elsewhere in the body (Butcher et al., 2007). Interstitial cells (IC) are the most
abundant cell type present in heart valves. IC are a type of myofibroblast which are
16 Chapter 1. Background
concentrated in the spongiosa but may be found throughout the valve. The ICs also tend
to be oriented circumferentially across the leaflet. ICs in the spongiosa are anchored to
collagen fibres in the ventricularis and fibrosa layers of the valve. This anchoring is likely
to act as a “strain sensor” which can sense when the valve is straining more than usual,
so signalling that more collagen should be synthesized. As well as producing collagen,
the fibroblasts also break down collagen. Compared to other tissues, the valve has a
particularly high tissue turnover which may contribute to the durability of the healthy
valve (Thubrikar and Klemchuk, 1990), and so one cannot compare the durability of
polymers with that of regenerative tissues.
It is no surprise that valve disease is often associated with IC dysfunction. In calcific
valve disease, IC differentiate to become osteoblast valvular interstitial cells which can
lead to the formation of calcific nodules. However the picture is not entirely clear: IC
seem to have a complex and symbiotic relationship with EC, making it hard to develop
targeted therapies (Liu et al., 2007a).
1.5.1 Collagen
The fibrous protein, collagen is found throughout the body. In some tissues, such as
tendons, it is the primary component. Collagen forms most of the organic matrix in
bone and is the predominant component of the ECM throughout the body. Amino
acids are arranged in a triple helices (∼ 300 nm), which are crosslinked to form collagen
fibrils (∼ 1 µm), that bundle to form collagen fascicles. The collagen fibrils found in
soft tissues (such as tendons) have a modulus of approximately 1.2 GPa (Fratzl, 2008),
which is similar to many common thermoplastics such as acrylonitrile butadiene styrene
(ABS) and thermoplastic polyurethanes (Granta Design Ltd., 2003). Collagen also has
an ultimate tensile stress of 120 MPa, which is double that of the equivalent synthetic
plastics. These mechanical properties go some way to explaining the ubiquity of collagen
throughout the body.
The mechanical properties of collagen are non-linear and dependent upon environment.
Alongside the anisotropy of the fibres; water content, crosslink density, mineralization,
strain, and strain rate must be considered. Single collagen fibres have a non-linear stress-
strain response - initially they are soft, but stiffen significantly once they reach about
10% strain (Shen et al., 2008). This quirk might be important in the cusps of the heart
valve as it allows the valve to be opened easily, while having strength when stresses are
high.
In the aortic valve, bundles of collagen are located in the fibrosa. These bundles commence
Chapter 1. Background 17
at the commissures of the valve and run roughly parallel to the free edge of each cusp in
the circumferential direction. At the centre of each cusp the bundles become connected
resulting in a dense reinforcement which runs radially from the centre of the free edge to
the nadir. This thickening is referred to as Arantius’ Nodule (Sauren, 1981).
To investigate the orientation of collagen across the valve, I explanted and dissected a
healthy porcine valve which I stained for collagen with Van Gieson’s Picrofuschin solution
(VWR International Ltd). I mounted one cusp on a metal frame to simulate a loaded
conformation. Figure 1.7 shows the prepared cusp. The circumferential collagen structure
is visible, with radial collagen interconnections between the large bundles. The thickening
at the centre of the leaflet is also clear.
Nadir
Free edge
Commissure
Com
missu
re
Figure 1.7: Leaflet from a porcine valve suspended between stretched threads and stained
for collagen using Van Gieson’s Picrofuschin. The fibres of collagen clearly extend from
the commissures and aortic wall circumferentially across the cusp. The white ellipse
marks the central area where the bundles branch and combine. The leaflet is 20 mm
wide.
I observed that the main collagen fibres were oriented circumferentially (Figure 1.7),
but it is established that older, more developed leaflets have greater bundle branching
and connectivity (Sauren, 1981). In a study, using high resolution digital imaging with
transmitted elliptically polarized light, Doehring et al. (2005) reported that collagen
structure is more complex than previously thought. The fibres show significant branching
and are also arranged in multiple layers which is thought to aid the sliding of fibres over
one another. Furthermore, there is significant variation between leaflets with the right
coronary cusp being self symmetric while the remaining cusps are mirror images of each
18 Chapter 1. Background
other. The branched collagen structure in an unpressurised leaflet becomes deformed as
the valve is pressurised. Once the transvalvular pressure reaches 20 mmHg (approximately
20% of the resting maximum) the collagen bundles reach their maximum alignment.
1.5.2 Elastin
Tropoelastin protein molecules associate and crosslink to form insoluble elastin fibres.
The production of elastin diminishes rapidly with age. However, as elastin has a half
life of 74 years we rely upon the fibres that were produced during the first years of life
(Weiss, 2010). Although the valve contains a relatively low proportion of elastin (13%,
compared to 50% collagen (Vesely, 1998)), it has specific structural roles. Compared to
collagen, elastin is soft and much less abundant, indicating that it has a small role in
directly supporting large loads. However, the ventricularis can undergo strains of 60%
during opening (in the radial direction), and elastin is the only natural material capable
of withstanding such strains. Any collagen which is radially oriented in the ventricularis
is wavy when unloaded, and so only becomes loaded at high strains. Thus, during small
strains the mechanical properties are dominated by elastin.
Valvular elastin is found in sheet structures which are mechanically isotropic and so do not
directly contribute to the anisotropic properties of the leaflets. The leaflets undergo large
deformations, and yet they are capable of returning to their unloaded starting position.
Elastin is also responsible for this property as its preloaded fibres can return collagen
fibres to their original positions (Vesely, 1998).
1.5.3 Resultant mechanical properties
The combination of the fibrosa, spongiosa and ventricularis layers result in a heart valve
leaflet with a stiffness between 4 and 8 MPa (Weinberg and King, 2004), however, the two
most prominent properties of the combined heart valve tissue are the highly non-linear
and anisotropic response.
The non-linear mechanical properties of the valve can be attributed to the collagen ar-
rangement. At rest the collagen is crimped such that the fibres are wavy. In this state
only a small amount of force is required to stretch the composite material. As the strain
increases, the fibres begin to uncrimp and align, resulting in a stiffer material. Finally,
once all of the fibres are straightened the maximum stiffness is exhibited (Holzapfel et al.,
2000).
Chapter 1. Background 19
The stress-strain properties of frozen and fresh human aortic valve leaflet tissue are
shown in Figure 1.8. The data shown here illustrates the non-linear anisotropic response
of the tissue. In particular, leaflets can be strained up to 7% in the circumferential
direction with a modulus of 0.02 MPa, which then rapidly increases to 6 MPa. In the
radial direction, fresh leaflets stiffen at approximately 15% strain with a post-transition
modulus of 1.7 MPa.
Radial - Frozen
Circumferential - Fresh
Circumferential -
Frozen
Radial - Fresh
-20
-15
-10
-5
S
tr
es
s
(k
P
a)
5 10 15 20 25 30 35 40
Strain (%)
Figure 1.8: Plot of the unaxial stress strain relationship for fresh and frozen aortic valve
leaflets in the circumferential and radial directions measured in vitro. Adapted from
Thubrikar and Klemchuk (1990).
1.6 From the inception of prosthetic heart valves to their current
state
Since the first valve replacement, in 1952, a wide variety of prostheses have had some
success in humans. The range of prosthetic valves may be described as: autographic,
homographic, mechanical, or xenographic. Autographic prostheses involve the trans-
plantation of a valve from elsewhere in a patient’s body, typically the pulmonary valve.
Homographic replacements involve the use of tissue from the same species, generally
cadaveric valves which are often cryogenically preserved before use in a patient (Goffin
et al., 2000). Xenographic, hereafter referred to as bioprosthetic, valves utilise the tissue
of another species. Mechanical prostheses consist of non-biological materials.
20 Chapter 1. Background
1.6.1 Autographic prostheses
Autographic replacement refers to the use of an existing valve in the patient’s body to
replace a diseased valve. Most often, a pulmonary autograph is used to replace the aortic
valve and an alternate prosthesis is implanted in place of the pulmonary valve. This is
termed the Ross procedure. A high quality of life, and good survival rates make the Ross
procedure attractive, but the high surgical complexity makes the procedure only sensible
in centres of excellence for aortic root surgery (El-Hamamsy and Poirier, 2013).
1.6.2 Homographic prostheses
Prostheses from matched donors perform very well, with excellent haemodynamics and
good durability. As donation from a living donor is not possible, only cadaveric valves
can be used. This operation was first performed in 1962, however the use of cadaveric
implants has now waned due to the difficulty of supply (Dasi et al., 2009).
1.6.3 Mechanical valves
Mechanical heart valves were the first type of prosthesis to be developed and implanted.
There is an extensive variety of designs of which the most notable valves are described
here in chronological order.
The first valve type to be developed and implanted was the caged ball valve, the Starr-
Edwards valve being the most successful in this category. The caged ball design functioned
adequately, though improvements that provided a larger effective orifice area, reduced
shear, and required less anticoagulant therapy were in quick succession. The monoleaflet
valves, for example Bjork-Shiley, Medtronic-Hall and Omnicarbon valves, improved the
effective orifice area of the valve, though suffered from reduced durability to such an extent
that some models were recalled. The occasional catastrophic failure of monoleaflet valves
has been attributed to fluid cavitation as the valve closes, which can lead to surface
pitting (Kini et al., 2000). Bileaflet valves, such as the example shown in Figure 1.9,
produced by St. Jude Medical, Carbomedics or Edwards-Duromedics, are now seen as
the gold standard in mechanical heart valve design (Vongapatanisin et al., 1996).
Nowadays mechanical valves possess excellent durability properties. They are capable of
withstanding well over 20 years of operation and possess a desirable effective orifice area
(≥2 cm2 for a 23 mm valve) and minimal regurgitation. However, despite their admirable
haemodynamic parameters they suffer from serious flow field problems which necessitate
Chapter 1. Background 21
the use of anticoagulant therapy. For example, the bileaflet heart valve possesses several
flow channels whose jets have peak velocities of at least 2 m s−1 during systole (Pibarot
and Dumesnil, 2009) which lead to thrombus forming vortices. In the turbulent vor-
tices encountered downstream from a mechanical valve, shear stresses between 120 and
200 N m−2 have been measured using Laser Doppler Velocimetry, which are significantly
greater than the threshold shear stress at which thrombus generation occurs. A critical
parameter in the design of the bileaflet valve is the angle between the open leaflet and the
flow axis; too great and the systolic flow field is severely disrupted, too low and closing
of the valve becomes difficult (Yoganathan et al., 2005).
The use of rigid materials in mechanical heart valves requires hinges to achieve rotation.
However, over time, they have a tendency to be fouled and clog up. To counter this, a
small gap is left between leaflets and the supporting structure that allows a small amount
of retrograde flow which helps to keep the hinge clean. Unfortunately this reverse flow
can also generate very high, thrombogenic shear stresses (Yoganathan et al., 2005).
The most promising next improvement in mechanical valves is the Lapeyre-Triflo FURTIVA
valve. This has three rigid polyether ether ketone leaflets which open to produce a central
flow orifice, like the native anatomy. The valve has been tested in a sheep for regulatory
approval and performed safely, in line with the reference St. Jude bileaflet mechanical
valve (Gallegos et al., 2006). However at the time of writing the valve has not reached
clinical trials.
Figure 1.9: Sorin Carbomedics bileaflet mechanical valve with a pyrolitic carbon coating,
titanium stiffening ring. The external diameter is approximately 30 mm.
1.6.4 Bioprostheses
In 1969 Alain Carpentier developed a method of fixing biological tissues using glutaralde-
hyde that improved their in vivo stability and reduced degradation (Dasi et al., 2009).
Using his technique bioprosthetic valves were first made commercially available in 1972
22 Chapter 1. Background
Figure 1.10: Edwards PERIMOUNT aortic valve prosthesis. The external diameter is
approximately 30 mm.
and marketed as the Hancock valve (Cohn et al., 1989). Their implantation is a phe-
nomenal success story: implanted valves rarely require severe anticoagulant therapy and
allow patients to rapidly return to their quality of life pre-disease.
Among bioprosthetic valves, a wide variety of designs have evolved in response to failures
and shortcomings in design. Practically all bioprosthetic valves are made from either
bovine pericardium or porcine aortic root tissue that is fixed with glutaraldehyde. The
valves must first be decellularised to reduce immunogenicity. Glutaraldehyde is used to
crosslink the proteins in the tissue, which improves durability (Ma et al., 2014).
There are a multitude of bioprosthetic heart valve manufacturers and designs. All aortic
prostheses have 3 leaflets which open to form a central orifice, giving them flow dynamics
similar to the native valve. Although there are notable differences between fundamentally
different designs, there is little to choose between manufacturers (Dasi et al., 2009). There
is also a wide variation in patient anatomy, meaning that no single valve is optimal for all
patients. It should also be noted that haemodynamic improvements may be counteracted
by reduced durability, and improvements in opening can come at the expense of greater
regurgitation.
Valves can be classified as either stented or stentless bioprosthesis. A stent refers to any
rigid structure placed inside a blood vessel. In a stented bioprosthesis, biological tissue
is mounted upon a rigid metallic or polymeric stent. The tissue may be inside or outside
of the stent and the leaflets mimic the native cuspal form. The most common stented
prostheses are the Carpentier Edwards PERIMOUNT and PERIMOUNT MAGNA, and
the Sorin Mitroflow which are composed of bovine pericardial tissue, while the Medtronic
Mosaic and Hancock valves use porcine leaflets. Stented valves may be placed intra-
or supra-annularly. Supra-annular placement allows valves with a larger diameter to be
used, that have a lower resistance to flow. The valves are sutured within the sinuses,
however this carries a small risk of occluding the coronary arteries (Ruzicka et al., 2009).
Chapter 1. Background 23
Table 1.3: Typical deformations found in bioprosthetic heart valves, adapted from Bal-
achandran et al. (2011)
.
Parameter Normal Notes
Aortic pressure during di-
astole (max. at rest)
65 to 85 mmHg
Aortic pressure during sys-
tole (max. at rest)
100 to 130 mmHg
Differential pressure across
closed valve (max. at rest)
95 mmHg
Peak flow velocities 3.7 m s−1 (Dasi et al., 2009)
Turbulent shear stresses Mean 200, max 450 N m−2 (Dasi et al., 2009)
Systolic bending strain 14.5% (Corden et al., 1995)
Diastolic bending strain 8.3% (Corden et al., 1995)
Stentless bioprosthetic valves do not have a rigid frame upon which the leaflets are
mounted. In a stentless valve, the valve suture line may follow the edge of the leaflet,
or they may be fabricated from a complete porcine aortic root, and use the natural at-
tachment between leaflets and the aorta (Pibarot and Dumesnil, 2009). Stentless valves
are either implanted in place of the resected aortic root or as a tube within the aortic
root. They initially fell out of fashion as they were trickier to implant. However, their
popularity resurged in the 1990s as they can have better haemodynamics and reduced
leaflet stresses. Among stentless valves, the most common is the St. Jude Medical Quat-
tro Valve. Also available are the Edwards Lifesciences Prima Valve, Biocor Valve and the
Medtronic Freestyle Valve, and the third generation of valves include the Sorin Pericarbon
Freedom and the 3F valve. Although stentless valves may be superior in some patients,
there is little conclusive evidence that they should be used in preference to stented valves
(Kobayashi, 2011; Siddiqui et al., 2009).
The deformations found in bioprosthetic heart valves are similar to those of native valves.
Table 1.3 shows the deformations and haemodynamics of bioprostheses. This can be
compared with Tables 1.1 and 1.2, which present structural and haemodynamic data for
the native valve. Further to the data listed, Subhani et al. (2013) observed that stented
porcine prostheses have inferior valve dynamics relative to the native valve, as they may
stay open for 25% less time during a full cardiac cycle.
1.6.5 Failure of bioprostheses
In the first years after implantation, bioprostheses are reliable: <1% of valves suffer struc-
tural dysfunction within 5 years. Among valves implanted in the general elderly popu-
24 Chapter 1. Background
lation in the 1980s and 1990s, 20-30% become dysfunctional within 10 years and ≥50%
failed via structural dysfunction within 12-15 years of implantation (Siddiqui et al., 2009).
Newer valve designs, such as the Hancock II valve, reduced structural valve deterioration
to approximately 20% over 15 years of implantation (Rizzoli et al., 2003).
In contrast, the failure rate has always been much higher in younger patients: nearly all
valves fail in less than 5 years in patients under 35 years old. The higher failure rate in
younger people may be attributed to their more active immune system which promotes
a greater inflammatory reaction, and higher circulating calcium concentration (Siddiqui
et al., 2009).
Structural valve deterioration was tackled by reducing stresses, using anticalcification
treatments, and alternative tissue cross-linking methods (Park et al., 1997; Tam et al.,
2015; Vyavahare et al., 1997b). The prospect of mitigating the immune response to
xenographic tissues through the use of genetically modified pigs is on the horizon, though
social and ethical barriers to acceptance do exist (Manji et al., 2012).
A large number of bioprosthetic explants are recovered after premature failure or at
patient post-mortem, providing a significant body of evidence for the mechanisms of
failure. This knowledge helps to inform the design of polymeric valves. The most likely
reason for rapid valve failure (within 2 years) is due to infective endocarditis. This has
an incidence of 1-6%, and is most likely to occur in patients with a previous history
of endocarditis; poor valve sterilisation is rarely an issue. Thrombosis as a cause of
bioprosthetic valve failure is relatively rare, occurring in just 0.1 to 5.7% of cases (Siddiqui
et al., 2009).
Calcification and mechanical damage are the two main mechanisms of valve failure. Ini-
tially it was thought that calcification led to mechanical damage, but this was unfounded
and the two processes can initiate separately (Vesely, 2003). There is a synergy between
the two processes - calcification leads to the stiffening of cusps, rupture of tissue and
subsequent valve failure. Calcification also occurs at a higher rate in regions of highly
localized mechanical forces, be they tensile or compressive (Vesely et al., 1998). The ex-
tent of knowledge surrounding calcification mechanisms is described further in Chapter
4.
When calcification is not present in failed bioprostheses, cuspal tearing is most common,
particularly in porcine prostheses. These tears are typically located either at the com-
missures or at the free edge where stresses are highest (Vesely, 2003). The use of flexible
stents to reduce mechanical pressures during diastole is presumed to reduce the likelihood
of tearing, but has not been proven (Siddiqui et al., 2009). Vesely et al. (1998) proposed
Chapter 1. Background 25
that fixed biological material does not fail due to tensile stresses; instead it is buckling
of tissue in compression which leads to valve failure. Compressed tissue can be found at
the centre of the free edge of each leaflet on the aortic side. Pannus overgrowth is also
recorded as a pathway to failure. The physiological wound healing process can sometimes
overshoot, resulting in the growth of collagen rich tissue onto the cusps or into the flow
which then leads to valve dysfunction (Siddiqui et al., 2009).
1.6.6 Valve selection
Given the duopoly of bio- and mechanical prostheses, it is appropriate to consider the
factors which determine the use of either valve when a patient requires a replacement.
The selection of a prosthesis type is made using a multitude of factors: their indication or
contra-indication for anticoagulant therapy, their age and life expectancy, and the desires
of the patient. A patient fulfilling any of the following criteria will generally receive
a bioprosthetic valve: age ≥65, limited life expectancy, contra-indication for warfarin
therapy or female of child-bearing age (Pibarot and Dumesnil, 2009).
1.6.7 Transcatheter valve replacement
In 2002, Alain Cribier implanted a prosthetic heart valve via a catheter with an ante-
grade transeptal approach on a 57-year-old patient who had been deemed inoperable and
on whom balloon valvuloplasty had been previously unsuccessful (Cribier et al., 2002).
Cribier triggered a disruptive change which has revolutionised the treatment of valvular
diseases (Berlin et al., 2014). The Edwards Sapien valve was first to reach the Euro-
pean market, receiving a CE mark in 2007. By 2013, an estimated 90,000 transcatheter
aortic valve implants (TAVI) had been implanted worldwide (Bourantas et al., 2013).
In the USA, the Food and Drug Administration (FDA) approved the Sapien valve in
November 2011, initially restricting implantation to inoperable patents and only using
the transfemoral route.
The implantation of prosthetic heart valves using a catheter avoids the need for heart-
lung bypass and open heart surgery, thus opening up valve replacement to the cohort
of patients whose co-morbidities left them at an unacceptably high risk of death with
traditional procedures. The proportion of elderly patients who are refused surgical valve
replacement was recorded as 33% by Iung et al. (2005) in the European population; this
cohort can now be treated, thus increasing the total number of heart valve implantations
(Rode´s-Cabau, 2012).
26 Chapter 1. Background
Figure 1.11: The Medtronic Corevalve transcatheter aortic valve implant consists of 3
porcine pericardial leaflets sutured onto a self expanding Nitinol stent. The valve sits in
the supra annular position and can be delivered using an 18Fr catheter. Image from of
USC Clinical Trials.
In the early 2000s clinicians and valve companies were slow to adopt TAVIs. A decade
later they have become ubiquitous in developed countries and it seems that their use is
set to increase further. Although TAVI may be clinically justified for inoperable and high
risk patients, it still comes under intense scrutiny when cost implications are considered.
A typical valve may cost $40,000, compared to $5000 for a surgically implanted valve
(Neyt and Van Brabandt, 2013). A significantly reduced fee has led to a small number of
implantations in developing countries. The standard pricing of TAVI is yet to decrease,
although it has been forecast to do so as competition increases.
While there is a large difference in price between surgical and transcatheter valves, it is the
overall cost of each procedure that appears to be driving the disparity. The transcatheter
procedure - which permits a shorter hospital stay than the surgical procedure - enables
the transcatheter valves to have a price premium. In the PARTNER trial there was
a cost saving of $1250 and a modest increase in Quality Adjusted Life Year (QALY)
when TAVI was used, where one QALY indicates one year of perfect health, and years
spent at below perfect health correspond to values between zero and one. TAVI was
economically attractive in 70% of procedures at an incremental cost effectiveness ratio of
$50,000/QALY, compared to surgical heart valve replacement. This was not the case for
the transapical approach where surgical valve replacement was more attractive.
Chapter 1. Background 27
1.6.8 Tissue engineering
Tissue engineering refers to the artificially induced growth of a natural biological struc-
ture. Tissue engineering (TE) can be autographic if the body’s own cells are grown to
replace a tissue, or it can use non-autographic stem cells which do not elicit an immune
response.
TE heart valves are still under development. One method involves the seeding of a scaf-
fold (mimicking the extracellular components of the heart valve) in an in vitro bioreactor
before implantation. The scaffold is degraded within the bioreactor and the living valve
would be implanted. Alternatively, autologous cells may be recruited to an implanted de-
vice - in situ TE. This concept is supported by the observation that endothelial progenitor
cells are able to adsorb to the surface of certain ventricular assist devices. Furthermore,
the coating of stents with CD34 (an antibody found on the surface of endothelial progen-
itor cells) further encourages adhesion and proliferation (Mendelson and Schoen, 2006).
Researchers pursuing both methods continue to record promising results and there is
little doubt that, given time, the successes already seen in other organs (kidney and bone
regeneration, for example) can be repeated for heart valves.
Optimism was high in 2000 when autologous heart valves were grown on a polyglycolic-
acid mesh in an in vitro pulse duplicator before implantation and 8 months of successful
function in sheep. In vitro and in vivo, the tissues remodelled to form a trilaminar dis-
tribution of cells and ECM, much like the native valve. Unfortunately, the challenges to
translate this to human patients have proven to be insurmountable thus far, due to poor
reproducibility (Hoerstrup et al., 2000a,b). More recently, with a focus on replicating the
functional anatomic structure of the native heart valve, Masoumi et al. (2014) developed
a tri-layered elastomer scaffold for TE heart valves. This biologically inspired structure
consists of poly(caprolactone) microfibres separated by a poly(glcyerol sebacate) sheet
whose microstructure is engineered to have anisotropic mechanical properties. The scaf-
fold was seeded with valvular interstitial cells and mesenchymal stem cells, and cultivated
for 4 weeks. The valve tissue matched a variety of mechanical and biological parameters
and performed adequately in an isolated ovine heart model (Masoumi et al., 2014).
Despite the promise of developing a TE valve, continued development may not be fully
justified given the relatively reliable current solutions for adults. From a regulatory
perspective, TE implants may be classified as a drug as opposed to a device, consequently
requiring a tougher, longer, period of certification. The in vitro culture of valves is
unattractive from a scale-up perspective, while the in situ recruitment of stem cells
must be able to cope with patient variability. As such, TE valves are probably best
28 Chapter 1. Background
suited to paediatric patients who, currently, must have frequent surgeries, with increasing
mortality, to replace prostheses which do not grow, and have poor durability. As such, a
TE valve - possibly grown from prenatal foetal cells harvested during pregnancy - could be
of tremendous benefit for children born with congenital defects (Yacoub and Takkenberg,
2005).
An innovative halfway house between the classical bioprosthesis and a TE valve was pro-
posed by Syedain et al. (2012). Species specific fibroblasts are seeded upon a cylindrical
fibrin gel. A pulsatile flow of culture medium is flowed through the gel, leading to the
circumferential alignment of the collagen which is synthesised by the fibroblasts. The
scaffold is then decellularised, using 1% sodium dodecyl sulfate instead of glutaralde-
hyde, and the tube is manipulated to form 3 leaflets which are mounted upon a custom
frame. The decellularised valves are shelf stable and have good hydrodynamic properties,
though no long term fatigue tests have been performed yet. The valves can then be im-
planted in the same species that the fibroblasts came from, making them allografts. Early
studies with an ovine model (cells and valve recipient) have shown promising clinical and
recellularisation results.
1.7 Design, development and testing methods
In this thesis, I will describe the design of a novel polymeric heart valve with the aim of
overcoming the limited durability of bioprostheses without the requirements for anticoag-
ulant medication that are associated with mechanical valves. The medical device design
process starts with the statement of a set of requirements; these may be driven by both
regulation and functional analysis (Alexander et al., 2001). The fundamental functional
objective of the device is to only permit flow in one direction, and the design outcomes for
heart valve prostheses are well defined by the ISO 5840 standard which must be satisfied
in order for a valve to gain a CE mark.
This project aims to achieve the following goals:
1. Determination of parameters for the selection of a polymer,
2. Selection of a polymer,
3. Design of the prototype valve,
4. Fabrication of the prototype valve,
5. Evaluation of the biocompatibility of the selected polymer,
Chapter 1. Background 29
6. Evaluation of the in vitro performance of the prototype valve,
7. Prediction of the lifetime of the valve,
8. Investigation into the likelihood of calcification, possible effects and causes,
The following sections of this introductory chapter first review the attempts by other
researchers to make polymeric prosthetic heart valves, which are used to inform a set
of criteria for the design and testing of the valve. I will then introduce the basis from
which the novel polymeric prosthetic heart valve will be designed, and review the range
of polymeric materials that may be considered.
1.8 Flexible leaflet polymer prosthetic heart valves
1.8.1 Early attempts
The mechanical properties of polymers have long placed polymer prostheses as prime
candidates to overcome the failings of mechanical and biological prostheses, but none
have yet succeeded. A collection of early polymer prostheses attempts is shown in Table
1.4. The first polymer prostheses were proposed independently by Berge and Lillehei,
who fabricated polyurethane and single cusp silicone (Silastic) valves, respectively. A
flurry of designs emerged in the following years. The surgeon Hufnagel produced well
over 10 different designs of polymer valve with flexible leaflets, but it was his caged-
ball mechanical valve was most successful clinically. Polyurethane prostheses were first
implanted into dogs in 1959 by Nina Starr Braunwald. The following year she implanted
the valve into a 44-year-old women in the mitral position, citing the operation as successful
as the patient survived 4 months before death due to arrhythmia (Roe, 1992; Ten Berge,
1958). The valve, however, was not adopted by other surgeons, nor was the procedure
repeated.
Further designs were tested throughout the 1960s. From Table 1.4 it seems that polymer
heart valve development has taken a path of trial, error, and little improvement. Indeed,
Roe (1992) describes the first attempts at valve replacement in the 1960s as operating “in
a sea of ignorance”. Designs lacked a scientific basis, and valid testing outside of humans
was notably absent. Regardless, the Roe-Moore valve (1960) functioned for 13 years in
a single patient without complications, and is a fine example of the potential of polymer
prosthetic valves. Despite the early promise of flexible leaflet polymer valves, Roe at-
tributed their demise to marketing factors and fear of manufacturer liability. In the early
days of valve development, both flexible leaflet and rigid valves were under consideration,
30 Chapter 1. Background
the Starr Caged Ball Valve (1960) gained commercial backing, mass production, quality
control, and significant marketing at the time, and subsequently became the dominant
valve (Roe, 1992).
Table 1.4: Early attempts (until 1965) at flexible leaflet polymeric prosthetic heart valves.
Polytetrafluoroethylene is abbreviated to PTFE.
Year
Human /
animal (#
implanta-
tions)
Material Failure mode
Valve name /
source
1958 Animal Polyurethane Unknown
(Ten Berge,
1958)
1958
Human
(50)
Single flap of silicone
elastomer (Silastic)
Material fatigue
Lillehei Valve
(Roe, 1992)
1959
Human
(47)
Dacron mesh
impregnated with
Silastic monomer
Successful when one
cusp was replaced; poor
results when 2 or 3
cusps were replaced.
Bahnson cusp
(Roe, 1992)
1959
Human
(350)
Teflon cloth coated
with polyurethane
foam
Limited information,
but no anticoagulation
used with disappointing
long term results
Earle B Kay
(Roe, 1992)
1960
Human
(76)
Dacron
Variable results - poor
durability
Muller (Roe,
1992)
1960
Human
(minimal)
Polyurethane foam
with thin knit Dacron
Attributed to complex
implantation technique
Braunwald-
Morrow (Roe,
1992)
1960
Human
(98)
200 denier PTFE,
impregnated with
polyurethane and
knitted into a round
tube
Stiffening and fracture
of Teflon cloth
McGoon (Roe,
1992)
1960 Human (8)
Elastomer on PTFE
mesh
Longest survival of
flexible leaflet valve
(one patient for 13
years). Unknown
failure.
Roe-Moore
(Roe, 1992)
Chapter 1. Background 31
Table 1.4 continued.
Year
Human /
animal (#
implanta-
tions)
Material Failure mode
Valve name /
source
1965 Human
Multiple flexible
leaflet models,
generally based on
Silastic
Unknown
Hufnagel (Roe,
1992)
1.8.2 Modern polymer prostheses development
The dominance of mechanical valves in the 1960s drew attention away from polymer
valves. But the arrival of bioprostheses in the 1970s led to a resurgence in interest for
polymeric prosthetic heart valves with the aim of achieving durability without the need
for anticoagulant therapy. Recently, two further applications of polymer heart valves
have become established. The first is the use of polymer valves in artificial hearts and
ventricular assist devices. Such devices have gained prominence in surgical applications
for their palliative and bridge-to-transplant roles. In these instances the required duration
of implantation is limited and polymer heart valves are favoured for their reproducibility
and affordability.
The second instance where polymeric valves might have an opportunity to succeed is in
the new breed of transcatheter devices. In a TAVI, the biological tissue is mounted on an
expandable stent, which must be compressed to one third its diameter to pass through the
blood vessels and reach the aortic valve. This crimping may lead to damage of the fixed
tissue, reducing the durability of the valve (Zegdi et al., 2011). As of September 2015, very
little data has been collected on TAVI durability due to the low life expectancy of those
patients who have thus far received them. Nevertheless, it is expected that polymeric
materials will cope with the crimping process better than fixed, biological materials. The
Triskele valve is the most developed TAVI with polymeric leaflets, though a first-in-man
implantation has not yet been performed (Mullen et al., 2013).
Modern polymer prostheses attempts are shown in Table 1.5. The 1990s saw most efforts
channelled towards polyurethane based valves, with various combinations of soft and hard
32 Chapter 1. Background
segments used to overcome issues of stability, mechanical durability and calcification. The
Aortech valve, fabricated from Elasteon, seemed to be the most promising product of this
era with the granting of several patents, though legal and business difficulties seems to
have occurred en route to market (though not covered in the academic literature).
A particularly promising attempt at a prosthetic valve was published by Kiraly and Hil-
legass (1980) who identified ethylene propylene diene terpolymer (EPDM)-type polymer,
Hexsyn (fabricated by the Goodyear Tyre Company), as holding potential for a poly-
mer prosthesis. Its biocompatibility was reported to be adequate, and Hexsyn displayed
remarkable flex life of over 400 million cycles to failure, as compared to 18 million for
Biomer and 4 million for both natural rubber and styrene-butadiene rubber. However, in
an accelerated tester the valves only survived the equivalent of 2 years - approximately
70 million cycles - which highlights the shortcomings of their fatigue prediction. As such,
the valve was only suitable for application in temporary blood pumps, where it performed
satisfactorily at a fraction of the cost of bioprostheses.
Supplementary to the attempts tabulated in Table 1.5, a multitude of alternative poly-
mers have been proposed for prosthetic heart valves, though few have followed the
route to complete valve fabrication. The majority are commercial polymers based on
polyurethane structures, for example, Bionate (biostable polycarbonate urethane), Elas-
teon (polyurethane with poly(dimethylsiloxane) soft segment), Pellethane (polytetram-
ethylene glycol based PU elastomer), Carbothane, Biomer, Corethane and Chronoflex
(Parfeev et al., 1983; Pinchuk, 1994; Pinchuk et al., 2008; Tanzi et al., 1997).
C
h
ap
ter
1.
B
ack
grou
n
d
33
Table 1.5: Later reports (1960 to 2013) of design and testing of flexible leaflet polymeric prosthetic heart valves.
Year Trial stage Material Failure Mode Location / Name Source Notes
1974 Unknown Unknown Unknown
University of Notre
Dame, USA
Chetta et al 1974
in (Ku¨tting et al.,
2011)
1978
University of Bern,
Switzerland
Vucic et al 1978 in
(Ku¨tting et al.,
2011)
1982 In vivo Unknown Unknown
Pennsylvania State
University, USA
Wisman et al, 1982
in (Ku¨tting et al.,
2011)
1982
Ex vivo and
in vivo
temporary
blood pumps
Hexsyn, an EPDM-type
polymer
Commissural tears after
107 million cycles in
accelerated tester. 3 weeks
success in vivo.
Cleveland Clinic
Foundation, Ohio,
USA
(Kiraly and
Hillegass, 1980)
Only financially
viable when used
in temporary blood
pumps
1988
Single
animal study
(goat)
Polyurethane copolymer (KIII
from Nippon Zeon Co. Ltd)
polyether-polyurethane with
polydimethylsiloxane under
silane coupling agent. Similar
composition to Cardiothane
Calcification of valve and
heart pump.
University of Tokyo,
Japan
(Imachi et al.,
2001, 1989)
1990 In vivo Polyurethane
Repeatedly reached 550
million cycles
Helmholtz Institute,
Aachen, Germany /
HIA
(Ku¨tting et al.,
2011)
Matured into
ADIAM valve.
1991
In vivo
(sheep)
Polyurethane (Pellethane 75A
2363)
Animals survived. Gross
calcification of valves.
University of Utah,
USA
(Hoffman et al.,
1991)
34
C
h
ap
ter
1.
B
ack
grou
n
d
Table 1.5 continued.
Year Trial stage Material Failure Mode Location / Name Source Notes
1995
In vivo
(Calf)
ADIAMat (proprietary
polyurethane)
Superior to bioprosthesis
with respect to
calcification and durability.
Adiam
(Daebritz et al.,
2004b)
Valve never
certified.
1996 In vitro Polyurethane (Estane 58201)
6/6 survived 10
years-equivalent
accelerated fatigue test
Glasgow, UK
(Mackay et al.,
1996)
2002 In vitro Polyurethane Calcification Glasgow
(Bernacca et al.,
2002)
Last published
result. Precursor
to proprietary
“Elasteon” and the
Aortech valve.
Patents (US
6171335 and
20060241744) filed
in 2006.
2004
In vivo
(growing
calf)
Polycarbonate Urethane
Superior to bioprostheses.
Death due to pannus
overgrowth. Moderate
calcification of aortic
valves.
Aachen / ADIAM
(Daebritz et al.,
2004a,b)
Valve deemed
superior to
bioprostheses
2010
In vivo
(Calf)
SIBS with embedded PET
fabric and dimyristoyl
phosphatidylcholine coating
Material failure and
calcification
University of Texas,
USA
(Wang et al., 2010)
2012 In vitro POSS-PCU
Improved performance
currently predicted
UCL, UK
(Rahmani et al.,
2012)
Now incorporated
into the Triskele
TAVI valve
C
h
ap
ter
1.
B
ack
grou
n
d
35
Table 1.5 continued.
Year Trial stage Material Failure Mode Location / Name Source Notes
2013 In vitro xSIBS
Improved performance
currently predicted
Stony Brook
University, USA
(Claiborne et al.,
2013b)
Successor to Wang
et al. (2010).
Stresses reduced
(in silico) and
platelet activation
rate not
significantly
different from
Magna tissue valve.
36 Chapter 1. Background
1.8.3 Current polymer prostheses development
Current research into flexible leaflet polymeric heart valves is concentrated in four groups:
Stony Brook (NY, USA); Institute of Biomedical Engineering, University College Lon-
don (UK); Cardiovascular and Biofluid Mechanics Laboratory, Colorado State University
(CO, USA); and the Department of Chemical Engineering and Biotechnology, University
of Cambridge (UK) in collaboration with Politecnico di Milano (Italy).
The Stony Brook valve is fabricated from a crosslinked block copolymer polystyrene-block -
polyisobutylene-block -polystyrene (xSIBS)The major published work from this effort in-
volved the optimisation of shape for reduced haemodynamic thrombogenicity (Claiborne
et al., 2013a). The University College London valve is fabricated from a polycarbonate
urethane with polyhedral oligomeric silsesquioxanes nanoparticles, with shape optimised
for minimum energy for opening/closing (Burriesci et al., 2010). No fatigue testing or in
vivo studies are reported for either valve. The Colorado State group proposed the use of
interpenetrating networks of hyaluronan and linear low density polyethylene (LLDPE)
for prosthetic heart valve leaflets (Prawel et al., 2014). The hydrophilicity of hyaluronan
improves biocompatibility, and the addition of LLDPE allows the mechanical properties
to be tuned to be similar to the native heart valve tissue. The prototype valves, whose
leaflets are made from the novel material, had adequate hydrodynamic results, though
have not yet been tested for calcification, durability, or in vivo (Kheradvar et al., 2015).
PECA labs are a small spinout company from Carnegie Mellon University (PA, USA)
which is developing the “MASA” valve. The device consists of an ePTFE conduit with
an integrated valve for paediatric patients who require right ventricular outflow tract
reconstruction. An early design of the valve was implanted in 16 patients, with a median
age 1.7 years and with successful 6 month follow up (Yoshida et al., 2011). The group
are now working towards FDA approval under a humanitarian device exemption as no
other therapies are available for the target patient group (PECA Labs, 2015).
1.8.4 Textile prosthetic heart valves
The use of textiles for the design of leaflets for prosthetic heart valves appears to be an
attractive solution. From a mechanical perspective, textiles are fabricated from fibres
which are renowned for their tensile strength. When incorporated into a fabric, yarns
can move past one another and so flexibility can be achieved with minimal sacrifice of
strength.
The porosity of textile valves allows the ingrowth of tissue, but this can lead to stiffening
Chapter 1. Background 37
of the leaflets. The use of textiles for transcatheter valves is particularly interesting
as the fabric may be very thin which would allow for even smaller delivery sizes. Heim
et al. (2014) describe a textile leaflet valve made from polyester which reached the animal
trial stage. Although minimal tissue ingrowth was observed on the valve at autopsy, the
leaflets were found to have stiffened which was attributed to calcification, and led to the
animal’s premature death (Heim et al., 2014).
Durability has been the greatest hurdle for the bioprosthetic and polymeric valves re-
viewed here and thus it is not unsurprising that textile valves suffer from similar issues.
To achieve adequate durability, the yarns which make up the mesh must have tensile
durability and also be able to resist frictional degradation as they slide past one another
(Vaesken et al., 2014). The action of cycling results in rearrangement of the fibres which
further complicates the long term prediction of stress distribution in the valve.
1.9 Possible failure routes
To steer the design of a novel valve, I collated the known failure routes for the valve
based upon the criteria of ISO 5480, which regulates the design and manufacture of
cardiac valve implants; the failings of the valves cited in Section 1.8.2; the routes to
failure for bioprosthetic valves in Section 1.6.5; and the requirements stated by Burriesci
et al. (2010); these characteristics are shown in Table 1.6.
Table 1.6: Failure routes for prosthetic heart valves.
Clinical Hazard Device failure route Notes
Inability to complete
implantation
Improper sizing Poor labelling for size and location
Asymmetric valve does not
rotate
Blocks coronary arteries
Valve damaged During handling
Poor transportation / storage
Valve not compatible with
standard surgical tools
Sewing ring Non-compliant
Inaccessible
Short term issues:
endocarditis
Non sterile device
Non-sterile accessories
Poor packaging
38 Chapter 1. Background
Table 1.6 continued.
Clinical Hazard Device failure route Notes
Other short term
contamination
Materials contaminated Material synthesis
Material processing
In vivo diagnostic
incompatibility
Device migrates due to poor
anchoring
Device heating
Image distortion
Haemodynamic issues Paravalvular leak Poor suturing
Inflexible cuff / sewing ring
Suture breakout / tearing
Separation of cuff from frame
Increased ventricular load Transvalvular energy loss too high
Effective orifice area too small (patient
prosthesis mismatch)
Regurgitation too high
Aortic wall injury Flow jet
Thromboembolic event Material Thrombogenic
Activates coagulation cascade
Valve dynamics Shear vortices
Catastrophic valve failure Loss of a leaflet
Embolisation of whole vale
Long term haemodynamic
issue
Pannus overgrowth
Stiff fibrous capsule develops
around cusps
Calcification of leaflets
Deterioration of coating or
material results in
thrombogenic material
Long term valve failure
(durability)
Cuspal tearing
Stent post cracking
Cracks in joints
Creep of stent post / leaflets
Calcification - mechanical
strain synergistic failure
Chemical degradation -
mechanical strain synergistic
failure
Bleeding event
Improper anticoagulation
regime
Chapter 1. Background 39
It is clear from Table 1.6 that there are a wide variety of failure routes. Some, such
as hydrodynamic performance and thrombogenicity, may be tested with appropriate in
vitro models. Others, such as long term haemodynamic issues and fibrous encapsulation,
require in vivo animal models in order to be tested.
From the significant number of failed polymer prosthetic heart valves noted in Section 1.8,
it is appropriate to consider the most prevalent route by which they failed. 23 polymeric
valves designs are listed in Section 1.8; of the twelve that reported an outcome, one
was reported as successful, seven suffered from fatigue failure, four reported calcification
with or without failure, one failed with poor biocompatibility. While the fundamental
trileaflet design was used for nearly all of these valves, the material was varied. As such,
it is evident that material selection is exceedingly important. In his treatise on design,
Ashby et al. (2004) noted that the materials determine both function and form, and as
such their selection is absolutely critical to successful design. The first requirement to
satisfy in the design of a novel prosthetic heart valve is suitable material selection.
From my analysis of 50 years of polymer prosthetic heart valve failure, I drew the conclu-
sion that the most common modes of failure are through poor durability and calcification.
Calcification, which will be explored further in Chapter 4, shows many synergies with
mechanical fatigue failure, though it is unclear which is causative. Bernacca et al. (1995)
noted that many in vivo explants have reported that regions of calcification concentrate
at points of high stress in the valve’s leaflets. However, calcification-related fatigue failure
is likely to occur earlier before the deposition of calcium phosphate leads to stenosis. As
such, selection of materials on a mechanical basis becomes my first design requirement.
1.10 Bioinspiration
Research into prosthetic heart valves space has explored an extensive range of avenues.
A wide variety of biological tissues have been tested, a multitude of mechanical designs
developed, and the application of many polymers has been explored. Given this context,
I, in the Cambridge group, sought a new source of inspiration - nature. The art of
imitating nature, formalised in 1960 by the US air force with the advent of bionics, could
be applied to prosthetic heart valve design. This field represents the study and imitation
of nature’s methods, designs and processes. In the 1960s it became clear that while some
designs can be copied from nature, there are many ideas from nature which are best
adapted or used to serve as inspiration for human-made capabilities (Bar-Cohen, 2005).
40 Chapter 1. Background
The term biomimicry denotes the reproduction of a biological process using synthetic
methods. The development of a TE valve (Section 1.6.8) is a form of biomimicry. In this
thesis I will look towards reproducing function rather than form, and so when I draw
upon elements of the natural design to produce a better valve - without mimicking them
- I am bioinspired. Among the classical prosthetic valves, the implementation of three
leaflets in aortic prostheses is a fine example of successful bioinspired device design. The
Cambridge group has now drawn upon the anisotropy of the native leaflets to further
inspire the design.
1.10.1 Leaflet anisotropy
The structure of leaflets in the native heart valve was examined in Section 1.5, with
particular focus on the distribution of collagen in the aortic valve. It is generally accepted
that the arrangement of collagen in the valve plays a significant role in the durability of
the valve. Various approaches have been taken to demonstrate the importance of fibres
and the effects of orientation.
Liu et al. (2007b) carried out an in silico analysis of the effect of fibres on a polymer
prosthetic heart valve under diastolic loading. Using leaflets with a spherical form, Liu
et al parametrically varied the angular distribution of fibres in a single- and double-ply
model. 0.1 and 0.2 vol.% of fibres were tested. In the single-ply model, fibres oriented
perpendicular to the stent leaflet boundary resulted in a 63% reduction in stress relative to
the no fibre case. The double-ply model was used to mimic the native collagen structure
more accurately as a rhombic form of fibres is created, somewhat mimicking the branching
of collagen bundles, as seen in Section 1.5.1. Comparing the maximum principal stresses
in the designs, the authors determined that the double ply arrangement resulted in a
minimum of stress. This model was not tested for its opening so there is no way of
determining the overall effectiveness of the design.
An analysis performed by Zaffora (2011) commenced with a random orientation of fibres
in the leaflet and evolved the model to optimise orientation. The valve was modeled using
the mechanical properties of an anisotropic styrenic block copolymer (polystyrene-block -
polyisoprene-polybutadiene-block -polystyrene, 19wt% polystyrene) and the leaflet form
was based upon an unknown bioprosthesis. A selection of leaflet thicknesses were exam-
ined: 0.15, 0.3 and 0.45 mm. After three iterations the distribution of fibre orientation
became stable, and was broadly horizontal or circumferential, as shown in Figure 1.12a.
This orientation was very similar to the native valve’s collagen orientation (Marom et al.,
2013). As shown in Figure 1.13, the maximum strain energy density was reduced from
Chapter 1. Background 41
(a) (b)
Figure 1.12: Anisotropy was optimised in a polymeric prosthetic heart valve by orienting
fibres in the principal strain direction. The leaflet-stent boundary is constrained against
translation and a symmetric constraint is applied to the leaflet-leaflet contact. (a) shows
the orientation in the undeformed shape. (b) von Mises stresses in the leaflet when
loaded to 180 mmHg. The maximum stress is located at the commissure in this instance.
Courtesy of M. Serrani and Zaffora (2011).
0.44 MPa to 0.3 MPa. The peak stress locations for these in silico valves were found near
the commissures. This result is exacerbated by the stent-leaflet edge being fully con-
strained against translation. The investigators did not model the opening behaviour of
the anisotropic valve. Nevertheless, based upon the amount of coaption between leaflets
during diastole, a leaflet thickness of 0.3 mm was proposed for manufacture.
It has also been shown that collagen fibres play a role during systole; De Hart et al.
(2004) performed in silico studies with a bilayered fibrous structure in a 3 dimensional
fluid-structure interaction model. In particular, their model showed fluttering of the
non-reinforced leaflet during opening which is believed to accelerate fatigue. The fibres
seemed to stabilise the leaflet during opening without affecting the transvalvular pressure
drop.
Cacciola et al. (2000) created prototype stented and stentless valves with ethylene-
propylene-diene-monomer rubber matrix reinforced with high-performance polyethylene.
With the leaflets open to form a cylinder, the fibres were either arranged in a unidirec-
tional circumferential manner, or in a periodic sinusoid such that the fibres formed a
diamond lattice structure. This structure had a height of 4 diamond units and circumfer-
ence of 30 diamond units, where each diamond unit had a vertical long axis, with a 60°
acute angle. The valves opened and closed with a low transvalvular pressure drop, but
had an unacceptable level of regurgitation, which was attributed to the low coaption area.
The group at the Eindhoven University of Technology are now pursuing TE alternatives
(Kheradvar et al., 2015).
42 Chapter 1. Background
0 0.2 0.4 0.6 0.8 1
0
0.1
0.2
0.3
0.4
Normalised pressure
M
ax
im
u
m
st
ra
in
en
er
gy
d
en
si
ty
(M
P
a)
Isotropic
Baseline
Optimised
0 0.1 0.2 0.3 0.4 0.5
0
1,000
2,000
3,000
4,000
Strain energy density (MPa)
n
u
m
b
er
of
el
em
en
ts
Isotropic
Baseline
Optimised
Figure 1.13: Strain energy density in the hyperelastic anisotropic finite element (FE)
model of Zaffora (2011) from 0 to 180 mmHg. The data was extracted from the model by
Dr Serrani. The isotropic data refers to a random orientation of fibres, baseline data refers
to the initial guess in the iteration process where the fibres are all oriented perpendicular
to the surface, and the optimised data is obtained after three iterations.
1.10.2 Stent post flexibility
A marked difference between the prosthetic and native heart valves reviewed here is the
nature of the mechanical constraint at the commissure. The aortic root, which supports
the native leaflets, is relatively compliant as it dilates during systole and contracts during
diastole. This movement improves coaption between the leaflets and helps to redistribute
stresses. The standard bioprosthetic valve has a relatively stiff stent. The benefit of
commissure movement was highlighted by Krucinski et al. (1993) who modelled the cases
of rigid, flexible, and flexible and rotatable stents. Flexibility of the stent in the radial
direction was noted to reduce shock loading during diastole as the commissure moves
inwards. During systole, the sharp bend at the commissure results in compressive and
tensile stresses. These can be relieved by allowing the stent post to flex and the leaflet
to rotate around the commissure.
In summary, anisotropy and stent flexibility play an important role in the native leaflet.
Computational models have indicated that reproducing these characteristics in prosthetic
valves would also be beneficial - most significantly by reducing the maximum stress during
diastole. One aim of this thesis is to see whether these in silico results could be translated
to in vitro experiments.
Chapter 1. Background 43
1.11 Polymers
1.11.1 Polymers for medical applications
Since the first man-made plastics were fabricated at the turn of the 20th century, polymers
have brought material-based revolutions to many fields. Our understanding of polymer
architecture has improved throughout the 20th century, to the extent that during the
latter decades the concept of custom polymers - designed from molecular structure to
final product - has become a reality (Ebewele, 2000). The use of polymers in the medical
industry is driven by several motivations. Polymers can be very cost effective, easily
processed in a multitude of ways, fit into quality control systems with ease and can have
their properties tuned over a continuous spectrum.
Polymers are generally divided into thermosets and thermoplastic types. The former
irreversibly cures under a chemical or thermal stimulus while the latter can transition
from solid to liquid multiple times. Polymer behaviour can generally be attributed to
chain-chain interactions. Vulcanisation and crosslinking involve the formation of strong
intermolecular bonds which prevent polymeric chains from moving independently of one
another; this improves the elasticity of the polymer. Vulcanisation generally requires a
chemical curative, such as sulfur, peroxide, platinum, a urethane crosslinker, metal oxide
or acetoxysilane.
Silicones
Silicones are the most commonly found polymers among medical devices (Yoda, 1998).
They have relatively good biocompatibility, adequate biostability, can be fabricated in a
variety of forms and are easily sterilised. Silicones have a polymeric backbone consisting
of alternating silicon and oxygen atoms (Figure 1.14), and they can be crosslinked using
either peroxide (which leaves a surface bloom) or the preferred platinum systems. Sili-
cones can be moulded using liquid injection moulding, transfer moulding or compression
moulding. In liquid injection moulding, the catalyst (typically a platinum curing agent
which remains in the finished product) and crosslinker are injected through a mixer before
entering the mould at which point the curing process takes place. This is the preferred
method for medical devices as the system can be fully contained to reduce contamination,
fully automated to reduce the possibility of human error, and cycle times can be faster
(LeFan, 2011; MDDI Admin, 1999).
44 Chapter 1. Background
C
H
C
H
H


n
CH2
C
H3C
C
H
CH2
 
n
Polystyrene Polybutadiene
CH2
C
H
C
H
CH2
 
n
C
H
H
C
CH3
CH3


n
Polyisoprene Polyisobutylene
R N
H
C
O
O R2


n
Si
R
R
O Si
R
R
O


n
Urethane linkage, where R1 and R2 are
aliphatic or aromatic sections.
Silicone, where R is an organic group,
which would be CH3 in polydimethylsilox-
ane (PDMS).
Figure 1.14: Chemical structures of common polymer groups used in this thesis.
Polyester
Polyester (PES) is formed by the condensation reaction between a diacid and a dialcohol.
Through varying the chemical structure of the reactants, the mechanical properties of the
resultant PES can be tuned to be brittle and hard, tough and resilient, or soft and flexible
(Nawrat, 2009). The most commonly found PES is polyethylene terephlate (PET) which
was developed independently by Imperial Chemical Industries (ICI) and DuPont in the
1940s. While the most common uses of PES are in clothing and bottles, it can also be
found in biostable and biodegradable medical devices. In its flexible state, the polymer
can be drawn into yarns, which in turn can be manipulated to form textiles. These textiles
have found uses as bandages, sutures, and artificial tendons and ligaments, and have
been proposed for artifical heart valves and kidneys. Unfortunately, PES is susceptible to
Chapter 1. Background 45
hydrolytic degradation which can cause fibres to rupture, making it unsuitable for long
term in vivo implants (Rahman, 2012). Furthermore, PES is quite thrombogenic. This
response diminishes over time, and is also mediated by using antithrombotic therapy, but
means that PES materials (e.g. Dacron) are rarely used in cardiovascular applications
(Kester, 1984). PES meshes do find a use in the sewing rings of prosthetic heart valves
as they are rapidly infiltrated by tissue, without a long term inflammatory reaction.
Polyurethanes
Polyurethanes (PU) consist of a chain of organic units joined by urethane links (Figure
1.14). They are formed from a di- or polyisocyanate stream and a polyol stream which are
reacted together with chain extenders, catalysts and surfactants. Polyurethane elastomers
generally contain “soft” polyether or polyester segments, which phase separate from the
harder polyurethane segments. The hard segments form randomly located regions of
crystallinity within an amorphous matrix. The phase separation of PU, along with inter-
and intra- molecular hydrogen bonding, results in “virtual crosslinking” which results in
a high and flat plateau in modulus vs. temperature curves, as also seen for cross-linked
rubbers. The size of the phase separated regions, their crystallinity, and the degree of
crosslinking in either phase can also be controlled which allows a vast array of properties
to be achieved. A variety of processes can be used to manipulate PU into different forms:
extrusion, moulding, casting and dip coating, electrostatic and wet spinning of filaments
have all been used for medical devices. PU, particularly those with a silicone-containing
soft block, have good haemocompatibility. As such they have been used extensively
in cardiovascular devices such as blood bags, vascular catheters, vascular grafts, artery
bypass, and bladders in left ventricular assist devices, and total artificial hearts (Zdrahala
and Zdrahala, 1999).
Early polyurethanes suffered from poor biostability, but they rapidly became the polymer
of choice due to their mechanical properties which can mimic native tissues, and their
relatively good biocompatibility. Since their inception, polyurethanes have proceeded
from hydrolytically unstable polyester types, to hydrolytically stable but oxidatively un-
stable polyether polyol based PU. Finally, polycarbonate-based PU were developed with
adequate hydrolytic and oxidative stability to withstand long term implantation, up to
approximately 5 years. Although biostability was a prerequisite of early implants, more
modern devices can harness biodegradation and allow native tissues to regenerate in their
place (Zdrahala and Zdrahala, 1999).
46 Chapter 1. Background
Thermoplastic elastomers
There are 6 major classes of thermoplastic elastomer: polyamide, copolyester, olefinic,
styrenic (Figure 1.14), urethane thermoplastic elastomers, and thermoplastic rubber vul-
canizate. Thermoplastic elastomers are attractive from a manufacturing standpoint as
they are relatively easy to process (speedy, lower energy requirements and scrap can be
recycled). The polymer molecules generally comprise “hard” and “soft” blocks which can
be varied in type and length allowing the properties of the polymers to be tuned.
Yoda (1998) claimed that styrenic thermoplastic elastomers would eventually replace sil-
icone in a wide of range of medical applications due to their superior fatigue life and
ease of processing. During the 1990s, the use of polystyrene-block -polybutadiene-block -
polystyrene (SBS) and styrene-ethylene/butylene-styrene (SEBS) (see Figure 1.14) be-
came common in catheters, medical tubing, intravenous solution containers, syringe tips,
stoppers, dropper bulbs, baby bottle nipples, flexible caps, medical seals, and eye dropper
bulbs (Yoda, 1998).
Styrenic block copolymers make up a major class of styrenic thermoplastic elastomers.
As they are the focus of this project, I will present a more detailed discussion in Section
2.1.1.
Chapter 2
Polymer selection
2.1 Introduction
A range of polymers have been considered for use in prosthetic heart valves. As seen
in Chapter 1, polyurethanes have received the most attention due to their tunable, soft
tissue-like mechanical properties, and good biocompatibility. Unfortunately, some valves
have failed due to poor biostability, and others due to poor biocompatibility. Silicones,
polytetrafluoroethylene, and polyester have also been considered, but failed due to poor
durability, calcification, and thrombogenicity, respectively.
As discussed in Section 1.10.1, the collagen of the native heart valve is arranged into bun-
dles which result in an anisotropic material. The fibres extend circumferentially across
the leaflet and serve as a means of mechanically supporting the pressure load during
diastole, contributing to the durability of healthy heart valves. If this effect can be repro-
duced in a flexible leaflet prosthetic heart valve it may be possible to overcome the issues
of durability that have plagued polymer prostheses thus far. Several researchers have
proposed that fibres could be embedded into polymeric matrices to mimic the anisotropic
mechanical properties of the native valve, but these attempts failed due to dehiscence of
the fibres from the bulk. Another way of achieving anisotropy could be to use a block
copolymer.
Block copolymers (BCPs) might be thought of as the “alloys” of the polymer world
(Ruzette and Leibler, 2005). While a micro- or nanoblend of a rubbery and glassy polymer
may be tuned to like an alloy, these are lost on heating as enthalpic forces cause separation.
A BCP containing a rubbery and a glassy polymer can harness the elastomeric properties
of polymers at room temperature, combined with the ability to repeatedly mould the
47
48 Chapter 2. Polymer selection
polymer into complex forms at high temperatures, a key characteristic of thermoplastic
elastomers.
The molecular architecture of BCPs may also be tuned to produce materials which are
mechanically anisotropic on the macroscale, and could thus mimic the anisotropy of
heart valve leaflets (Stasiak et al., 2010). In this thesis I propose the use of a BCP with
anisotropic mechanical properties as a material for a prosthetic heart valve.
2.1.1 Block copolymer fundamentals
A BCP molecule consists of 2 or more different polymer chains, or blocks, which can be
synthesised using living anionic polymerisation. In the simplest cases BCPs consist of 2
polymer chains joined end-to-end, referred to as a linear BCP and denoted by A-block -
B or A-b-B. More complex molecules can be formed with 3 or more chains which can
be arranged linearly, into stars, miktoarm stars, cyclically, or grafted branches (Amado
and Kressler, 2011; Hamley, 2004). For these blocks to join, their monomers must be
chemically compatible such that they will bond to each other. However, monomeric
compatibility does not confer block compatibility, and so the most prominent feature of
many BCP at the polymer chain scale is the strong repulsion between unlike sequences,
despite only weak repulsion between their unlike monomers (Leiblert and Leibler, 1980).
If the forces of chain repulsion are strong enough, then microphase separation may occur
(at an order-disorder transition) leading to spatial domains into which each polymer
block will concentrate. The polymer seeks to minimise the total area of the A-B interface
despite this carrying an entropic penalty from the segregation and concentration of the
domains. This leads to several order-order transitions as enthalpic and entropic forces
are balanced. A full prediction of this transition behaviour for a linear diblock copolymer
(A-b-B) requires 7 parameters, of varying importance, to describe the blocks and their
interactions.
The polymerisation index N , temperature T , and the polymer-solvent interaction, or
Flory-Huggins parameter χ, have some importance in determining the structural transi-
tions. However the dominant parameter is the proportion of each block, described by f .
Self-consistent mean field theory can then be used to simplify the problem and predict the
morphology of the polymer domains based on the aforementioned parameters (Leiblert
and Leibler, 1980).
For a linear di- or triblock copolymer 4 types of phase separated structure are predicted.
When the fraction of block A (fA) is low, A concentrates into spheres. As fA increases,
Chapter 2. Polymer selection 49
the domains of A form cylinders, double gyroids and then lamellae as fA approaches 0.5.
These structures are illustrated in Figure 2.1. For phase separation to occur χ must be
Figure 2.1: Predicted phase diagram for A-B symmetric diblock of Flory-Huggins param-
eter and polymerisation index, χN , versus the proportion of each block, f . L = lamellae,
G and G′ = bicontinuous gyroids, C and C ′ = hexagonally packed cylinders, and S and
S ′ = body-centered-cubic spheres. A narrow region where the spheres take on a closely
packed structure (CPS and CPS ′) is also predicted between the disordered region and
S. For χN > 40 the phase diagram boundary lines tend to be vertical. From Mai and
Eisenberg (2012).
positive. χ also influences the sharpness of the transition between domains. However,
the order-order transitions have remarkably little dependence on χ, and so there is very
little variation in the value of f at which the order-order transitions occur in different
phase separating BCPs. The microdomain dimensions do however scale as D ∼ amN 23χ 16 ,
where D is the characteristic length of a repeating structural unit, am is the length of a
monomer unit, and N is the polymerisation index (Bates and Fredrickson, 1990).
The cylindrical morphology is of most interest because if the cylinders can be aligned,
it may be possible to produce mechanically anisotropic materials. Cylinders1 are formed
when the mole fraction of one block is in the region of 18 to 30%. These limits have a
small dependence on polymer chain type, but these bounds are relatively unchanged for
1A note on nomenclature of cylinders: I have introduced the concept of cylindrical domains into which
the polystyrene (PS) blocks are concentrated. As well as cylinders, these domains may also be referred
to as (PS) nanodomains, (PS) cylinders, or (PS) fibres. Unless otherwise stated, the terms are essentially
interchangeable, and are used according to the context.
50 Chapter 2. Polymer selection
symmetric, 2-component, triblock architectures (A-b-B-b-A) (Castelletto and Hamley,
2004).
In order to produce a shortlist of polymers which might be suitable for a prosthetic
heart valve, I formed a set of criteria; the BCP must exhibit phase separation at room
temperature, the cylindrical morphology must have been reported, the polymer should
not contain heavy metal ions, and neither block should be obviously unstable in the
presence of water, weak acids, or weak alkalis. Finally, only commercially available BCP
were considered at this stage.
I performed a literature search for phase separating BCP, and identified a list of these
that exhibit a cylindrical morphology. These are listed in Table 2.1. Among this set of
polymers there are a wide variety of architectures - from linear chains, to aromatics, to
those containing metal ions. A limited number are stable and commercially available,
which are marked in the table.
Table 2.1: List of cylinder forming triblock copolymers using A, B, and C to denote each
block (Mark, 2007). Commercial availability of the polymers was determined based upon
the Polymer Source™ catalogue and Google searches.
A B C
Archit-
ecture
Commer-
cially
available?
2-exo-3-endo-
bis(diphenyl-
phosphino)-
bicyclo[2,2,1]
heptene
Methyltetracyclo-
dodecene
A-B
Butadiene Dimethylsiloxane A-B X
Ethylene Ethylenepropylene A-B
Ethylene Ethylethylene A-B
Ethylene Tert-butylmethacrylate A-B
Methyltetracyclo-
dodecene
Organometallic
derivative of
norbornene
A-B
Organometallic
derivative of
norbornene
Acetylene A-B
Lactide Isoprene A-B-A X
Styrene
2-(3-cholesteryl-
oxycarbonyloxy)ethyl-
methacrylate
A-B-A
Styrene Butadiene A-B-A X
Chapter 2. Polymer selection 51
Table 2.1 continued.
A B C
Archit-
ecture
Commer-
cially
available?
Styrene Ethylene Butadiene A-B-C-A X
Styrene Isobutylene A-B-A X
Styrene Isoprene A-B-A X
Butadiene Styrene Isoprene A-B-C X
Ethylenepropylene Ethylenebutene Styrene A-B-C X
Styrene Ethylene Propylene A-B-C-A X
Styrene Ethylene Butene A-B-C-A X
Isoprene Butadiene Styrene A-B-C X
Isoprene Styrene 2-vinylpyridine A-B-C X
Styrene Butadiene 2-vinylpyridine A-B-C X
Styrene Butadiene Methylmethacrylate A-B-C X
Styrene Ethylenebutene Methylmethacrylate A-B-C X
Styrene Isoprene 2-vinylpyridine A-B-C
Isoprene Styrene
Miktoarm
star
Styrene
2-methyl-1,3-
pentadiene
Miktoarm
star
Styrene Isoprene Butadiene
Miktoarm
star
Styrene Isoprene Dimethylsiloxane
Miktoarm
star
Styrene Isoprene Methylmethacrylate
Miktoarm
star
Isoprene Styrene
Super-H
star
A3-B-A3
Polymer shortlist
From Table 2.1 I selected the following shortlist of polymers based upon commercial
availability and stability. PS refers to polystyrene and b represents a polymer block.
• PS-b-polyisoprene-b-PS (SIS),
• PS-b-polyisoprene-polybutadiene-b-PS (SI/BS),
52 Chapter 2. Polymer selection
• PS-b-polyethylene-polypropylene-b-PS (SEPS),
• PS-b-polyisobutylene-b-PS (SIBS),
• PS-b-polyethylene-polybutene-b-PS (SEBS),
From the table, polyvinylpyridine and polylactide were considered, but not included in
the shortlist due to their known toxicity (at all except low concentrations (Allison et al.,
2007)) and biodegradability, respectively.
If the PS molar content is between 18 and 30%, PS would be concentrated within the
cylinders. PS has a glass transition of ∼90 ◦C whereas the rubbery blocks have glass
transition ∼−100 ◦C, resulting in cylinders which are stiffer than the bulk. Having gen-
erated a shortlist of seven BCP, I sought further criteria to select one polymer for use in
a prosthetic heart valve.
2.1.2 Material selection parameters
From the history of polymeric prosthetic heart valve design in Section 1.8, over 30 can-
didate polymers are found, yet none have succeeded in clinical use. A large proportion
of polymer flexible leaflet heart valves failed, typically during in vitro fatigue testing, or
in a chronic animal model, with poor durability listed as a key cause. There are various
causes of inadequate durability. In general terms, the following environmental and use
factors can cause the failure of polymers (Hulme and Cooper, 2012):
• Mechanical (creep, fatigue, stress relaxation, impact, environmental stress cracking
(ESC) and wear);
• Thermal (degradation and fire);
• Chemical:
– oxidation (oxygen, ozone, and metal catalysed),
– biological,
– other chemicals encountered during use or processing;
• Radiation (sunlight, ultraviolet light and atomic radiation);
• Electrical (arcing and electrostatic build-up).
Within the body the dominant modes of failure are mechanical and chemical. Mechanical
forces in the cardiovascular system arise from fluid shear and pressure forces. Chemical
forces may arise from hydrolysis or oxidation in aqueous conditions. Or, there may be
Chapter 2. Polymer selection 53
chemical species of biological origin; inflammatory cells will secrete enzymes and oxygen
radical species with the purpose of breaking down the foreign materials. The proposed
implant is passive and permanent, and so must withstand these forces. Each of these
forces can cause failure individually, or in combination .
Radiation, extreme temperatures and electrical hazards are unlikely to be encountered
in the body to the degree which they would degrade a polymeric material, but resis-
tance to these conditions is still necessary as medical devices must be sterilised prior to
implantation.
Combining biological, chemical, and mechanical forces of degradation increases the com-
plexity and expense of testing, and in vitro testing systems are unable to simultaneously
impart all three forces. The only pre-clinical model which realistically imparts all the
forces upon the valve is a chronic, large animal model. In the selection process, the
candidate materials were subjected to individual forces in vitro.
2.1.3 Stability selection
In the absence of inflammation, the host environment is an aqueous solution containing
lipids and proteins, at pH 7.4 and 37 ◦C. Unfortunately for any implantable material,
the body is also capable of producing harsher chemical conditions which are intended to
break down foreign species. Most typically degradation is achieved through hydrolytic,
oxidative or enzymatic conditions (ISO, 2010).
The rate at which biomaterials may degrade is a function of the material, local reactant
concentrations, and a set of rate constants (Maxwell et al., 2005). To accelerate ageing
of the material, the rate constant or reactant concentration can be increased. The exact
mechanisms of degradation are rarely known, so an increase in temperature is used to
alter the rate constant. In general, the rate constant of degradation is assumed to be of
an Arrhenius form:
kr = Aee
Ea/RT (2.1)
where kr is the reaction rate constant, Ae is the pre-exponential factor, Ea is the activation
energy, R is the Gas Constant and T is the temperature. This method clearly relies on
the assumption that the same reactions and mechanisms for degradation occur at the
reference and accelerated temperatures2.
2In the testing of polymeric materials an alternative expression where the rate constant is increased
using temperature is commonly used. The accelerated rate at T is related to a reference rate at Tr, by
2(T−Tr)/10. This relation essentially assumes that the reaction is first order and has a rate constant of
10R ln 2, and the same degradation mechanisms occur at the two temperatures (Hukins et al., 2008).
54 Chapter 2. Polymer selection
In reality, quantitative lifetime predictions are impossible as chemical degradation cannot
be regarded as a linear process, and harsher conditions can promote different degradation
mechanisms. The only effective method of assessing chemical durability for material
selection is qualitatively and comparatively (Bruck, 1991; Pinchuk, 1994; Vondracek and
Dolezel, 1984).
In a simple accelerated degradation test, Pinchuk et al. (2008) demonstrated the stability
of SIBS by submerging the materials in boiling concentrated (65%) nitric acid. These
conditions resulted in the destruction of silicone and polycarbonate urethane, whereas
PTFE and SIBS survived. To test hydrolytic and oxidative stability ISO 10993:13 (for
the identification and quantification of degradation products from polymeric medical
devices) recommends the use of buffered water and oxidative solutions, respectively, under
accelerated conditions of 70 ◦C. Degradation can then be evaluated by considering any
of mass loss, solution viscometry, swellability, rheology, chromatographic methods (e.g.
gel permeation chromatography), spectroscopic methods (e.g. Fourier transform infrared
spectroscopy) or thermal analysis (e.g. differential scanning calorimetry).
2.1.4 Biocompatibility selection
In this thesis, I propose a regular or classical implant, as opposed to a TE implant.
Classical implants rely upon materials which are chemically and biologically inert, and
success is defined by the implant having “no clinically relevant deleterious effect on those
tissues or components”. This can be contrasted with TE implants, whose materials are
used to deliberately target and activate certain cells in order to create new tissue for the
therapeutic reconstruction of the human body (Williams, 2015).
Adequate biocompatibility without the need for anticoagulation is a criteria for the se-
lection of a material for the prosthetic heart valve. Biocompatibility was set as a sec-
ondary criteria for selection as large differences in biocompatibility were not anticipated.
Previously, the indirect cytotoxicity of SIBS, SIS, SIS, and SI/BS, was assessed by my
collaborators in Politecnico di Milano, with no polymers eliciting a positive response. Fur-
thermore, one of the shortlisted BCPs, SIBS, is already in clinical use as a drug eluting
coating for the Taxus stent (Boston Scientific), indicating its biocompatibility. Finally,
the molecular architecture of the various BCP are not so different, suggesting that they
are unlikely to behave markedly differently. An analysis of biocompatibility is presented
in Chapter 3.
Chapter 2. Polymer selection 55
2.1.5 Material selection function
Once the unstable polymers had been eliminated from the shortlist, another criterion
was required to select the most suitable material for my application. In engineering
design, selection strategies can be used to choose a material with the most potential
for fulfilling a certain purpose. Examples of selection strategies are: questionnaire based
which requires some expert input and returns “known” solutions, analogy based on past
cases of material choice, or free search with quantitative analysis. The free search method
leaves the greatest scope for innovation (Ashby et al., 2004). Elements of this method
can be used in the selection of a material for a novel polymeric prosthetic heart valve.
In Appendix B I show a detailed example of the derivation of a material selection index.
Here, I derive a heart valve performance index.
Material selection should stem from an understanding of the fundamental function of the
heart valve; it must open for systole, and resist reverse flow during diastole. Furthermore,
a valve should have minimal resistance to flow during systole, while resisting reverse flow
indefinitely.
During the opening of a flexible leaflet valve, the curved leaflet must reverse its curvature.
The force required to do this, Frev, should be minimised. Once open, the pressure drop
over the valve is dominated by the valve’s geometry. In the case of valve closing during
diastole the situation is more complex. The forces on the valve leaflet that lead to closure
are dominated by dynamic fluid effects; for example, the sinuses of Valsalva promote eddy
formation resulting in the initiation of closure. As such, parameters such as valve height
and leaflet angle are used to affect the open area and closing dynamics (Burriesci et al.,
2010). Neither the open area nor the closing dynamics have a first order dependence on
material type, and so I neglected it from the performance index.
Bernacca et al. (2002) performed hydrodynamic studies on valves whose leaflets had
various thicknesses. They reported that a reduction in leaflet thickness led to a lower
mean pressure gradient over a full cycle. The force which is required to reverse the
curvature of a curved surface, Frev, is related to the leaflet’s thickness and the material’s
stiffness, and is driven by the transvalvular pressure difference, PV − PA, as shown in
Figure 2.2. Assuming that stress and strain are linearly related, the force required to
bend a leaflet of uniform cross section is proportional to the second moment of area
of the leaflet, which is proportional to the cube of the leaflet’s thickness, t3, (Gupta,
2001; Haworth, 1978). Leaflet opening requires the leaflet to bend, for which the flexural
56 Chapter 2. Polymer selection
modulus, Eflex, is a good measure of the material’s stiffness:
Frev ∝ Eflext3 (2.2)
Thicker or stiffer leaflets result in valves which perform less well (Bernacca et al., 2002).
The pressure required to reverse the curvature of a pure hemisphere is a function of both
t3 and t4, and so for a leaflet with spherical and cylindrical structures the power of t
should lie between 3 and 4 (Ghista and Reul, 1983). I used t3 in the following analyses.
To optimise haemodynamics, Frev should be minimised, making Equation (2.2) an objec-
tive function which contains one material property Eflex, and one design variable t.
The durability of the valve is also a function of material and valve design. The first order
determinant of stresses within the leaflet is thickness; thicker leaflets have lower stresses.
In a uniformly thick leaflet maximum stresses occur at the leaflet-stent boundary. It
2θ
PA
PV
PV PV
FL Rv
t
1. 2.
3. 4.
Figure 2.2: 2-dimensional cross-sectional schematics of a leaflet with either a purely
cylindrical or spherical form, with characteristic radius, Rv, when closed, and thickness,
t. 1 to 4 are forms of the leaflet from closed to open.
1. Aortic pressure, PA, is greater than ventricular pressure PV during diastole. The
pressure load is supported by the stent. The maximum leaflet stress is found at the
centre of the leaflet.
2. Once PV rises at the start of systole, the curvature of the leaflet can start to reverse.
The force required to do this is a function of R, t and the material’s modulus.
3. PV continues to exceed PA, causing the leaflet to finish opening.
4. The leaflet is fully open during systole. The geometric orifice area is determine by the
shape of the leaflet.
The force required to reverse the curvature is the sum of the bending moments along the
curved leaflet.
Chapter 2. Polymer selection 57
is possible to estimate the stresses within the leaflet if a simple leaflet form is assumed
(Figure 2.2 (stage 1)), for a cylinder segment of radius R and arc 2θ, supporting PA. The
stress per unit length, σc, at the commissure of the leaflet is:
σc =
δPR tan θ
t
(2.3)
where δP is the pressure difference (PA-PV ) that must be supported during diastole. The
radius of the cylindrical section would be roughly equivalent to the valve annulus radius.
For the valve to function properly the stress must be less than some fatigue limit, σf :
σf >
δPR tan θ
t
(2.4)
Thicker or stiffer leaflets lead to lower stresses during diastole making them more durable.
Thinner leaflets require a lower pressure to reverse their curvature, leading to better
haemodynamics. Given these conflicting criteria, I need to identify the optimal compro-
mise solution.
By combining Equations (2.2) and (2.4), the design variable, leaflet thickness t, can be
eliminated, leaving only the material properties and the constants. This quotient becomes
a material performance index, IP , to be minimised:
min (IP ) ∝
(
Eflex
σf 3
)
︸ ︷︷ ︸
Material properties
(δPR tan θ)3︸ ︷︷ ︸
Constants
(2.5)
The performance index approach to selecting a material for prosthetic heart valve mate-
rials was previously considered by Haworth (1978) who used a similar selection function:
IP =
E
σ∗3
(2.6)
This method was unsuccessful and, crucially, was based upon the tear rupture stress σ∗,
which is an unrealistic model for failure in polymer prostheses as it occurs at strains
several orders of magnitude greater than those seen in actual prostheses.
Parfeev et al. (1983) took a different approach. They established stress wear limits for
various natural, silicone and fluorosilicone rubbers after 107 cycles. They also compared
the stress wear limits to the materials’ moduli to produce a selection index. They went
one step further by making estimates of the forces which are generated in heart valve
leaflets. Under biaxial tension the maximum stress in a hydrostatically loaded hemi-
58 Chapter 2. Polymer selection
spherical membrane is,
σ =
Pls
2t
(2.7)
where P is the pressure, ls the radius of curvature and t the leaflet thickness, which were
taken to be 120 mmHg, 10 to 20 mm, and 0.25 mm respectively. Under these conditions
the aortic valve leaflets must be able to withstand 0.32 to 0.64 MPa. Combined with the
wear testing they determined that none of the isotropic polymers which were tested would
be suitable for an aortic polymer prosthesis, although a reinforced rubber or polyurethane
may be suitable. This assertion lends support to my proposal of using an anisotropic
material.
Parfeev et al. (1983) performed their durability testing at 22 Hz, which was a higher
frequency than normal loading conditions. The properties of polymers are dependent
upon frequency and strain rate, and this testing may not have been representative of the
actual performance.
In order to check whether it might be possible to accelerate the onset of cyclic failure
by increasing the cycle frequency I performed dynamic mechanical analysis of candidate
BCP samples (using a TA Instruments Q500). I located the limit of the linear region
using a strain sweep before performing a frequency sweep. These results are illustrated in
Figure 2.3, as a plot of damping factor, δ = viscous modulus
elastic modulus
, versus frequency. From these
data I observed that, not only is there a large transition in the damping factor at 80 Hz,
but also a smaller, β transition at 20 Hz. These transitions arise when deformations
occur faster than molecules can rearrange themselves indicating that the material has a
different behaviour under these conditions.
The linear region at which these measurements were taken extends up to 0.1% strain.
Given the larger strains that occur during leaflet deformations, the transition may occur
earlier than 20 Hz, making accelerated testing much trickier. Even at 10 Hz, 107 cycles
requires one month of testing for just one sample. Thus, a faster method of evaluating
fatigue life was sought. This will be discussed in Section 2.1.6.
Material performance indices can also be displayed graphically using Ashby maps, as
shown in Figure 2.4a. The material properties are plotted on logarithmic scales, and
material performance indices appears as straight lines. Materials lying along the line have
an equal performance index, while shifting the line corresponds to different indices. A
maximum or minimum leaflet thickness (due to manufacturing constraints, for example)
corresponds to a modulus constraint, which is also shown in Figure 2.4a
The world of materials can be plotted on the Ashby map using the CES selector software
(CES Selector, Granta Design®, Cambridge, UK), which can be used to produce results
Chapter 2. Polymer selection 59
0 10 20 30 40 50 60 70 80 90 100
0
1
2
3
4
5
Frequency (Hz)
ta
n
δ
Figure 2.3: Dynamic mechanical frequency sweep of SIBS30 material (n=3) at 0.1%
strain. A spline is fitted to the data. Note the peak in the damping factor, tan δ, which is
also accompanied by an increase in the elastic modulus which indicates a glass transition
occurring at 80 Hz. The peak seen at 20 Hz corresponds to a smaller, β transition.
such as those shown in Figure 2.4b and 2.5. The CES software has the capability to add
further selection criteria: in Figure 2.5 a maximum is placed upon the Young’s Modulus
(0.5 GPa), basic chemical durability criteria are added (excellent performance in fresh
and salt water, and excellent or acceptable in weak acids and alkalis), and ISO 10993-1
acceptance, indicating that it would at least be acceptable for use in non-implantable
medical devices. The “optimal” materials are marked in this diagram with  (Natural
rubber) and F (polyurethanes - though as some degrade this is marked in grey.). The
software contains information on SEBS, a BCP, which is indicated by ♣.
For the purposes of prosthetic heart valve design, the data is somewhat limited as fatigue
strength is listed for only 106, and the fatigue data is based on an R = −1, where R
is the ratio of minimum to maximum stress, indicating equal compression and tension,
which does not occur in a valve. Polymers rarely have a linear stress-strain relationship,
so Young’s Modulus is only a good comparison of stiffness for small deformations. The
property bubbles in the plot are also large, and often cover whole orders of magnitude for
a single material type - indicating fatigue and modulus vary significantly among types
and grade of material and are also dependent on processing.
60 Chapter 2. Polymer selection
100 101 102 103
10−2
10−1
100
101
102
103
Better materials
Young’s Modulus (MPa)
F
at
ig
u
e
st
re
ss
(M
P
a)
n = 3
n = 3± δ
(a)
Young’s Modulus (MPa)
F
at
ig
u
e
st
re
n
gt
h
10
6
cy
cl
es
(M
P
a)
(b)
Figure 2.4: Ashby map schematic (a) containing constraints (hashed areas) and material
performance index lines (blue). On a log-log plot, the performance index (Equation (2.5))
appears as a straight line, where the gradient corresponds to the index. The heart valve
index (with respect to fatigue stress) was assumed to be 3, but it may deviate, altering
the optimal material. Superior materials have a lower performance index and are found
in the upper left region of the diagram. The red hashed area indicates a theoretical
maximum value for the performance index, above which a valve would not open, or
would not be sufficiently durable. The hashed yellow area corresponds to maximum
modulus constraint. In (b), the “complete material world” is plotted. An inset of the
map is shown in Figure 2.5.
Chapter 2. Polymer selection 61
Young’s Modulus (MPa)
F
at
ig
u
e
st
re
n
gt
h
10
6
cy
cl
es
(M
P
a) 
F ♣
(a)
Figure 2.5: Ashby map with Young’s Modulus and fatigue stress plotted on logarith-
mic axes, showing the elastomeric materials. Chemical durability and medical grade
criteria were applied (grey indicates that the material failed). Natural rubber () and
polyurethanes (F) are the optimal materials, nearest to the top left. A block copolymer,
SEBS, is marked with ♣.
2.1.6 Elastomer failure
To select a polymer using the material performance index, Equation (2.5), the material’s
stiffness Eflex and fatigue stress limit σf must be determined. The former is relatively
rapid and reliable, while the latter is much trickier. Haworth (1978) used large strains
and Parfeev et al. (1983) used high frequency to predict the lifetime of elastomeric mate-
rials, but these means of extrapolation are not a representative assessment of long term
behaviour of polymers.
Within the engineering design community the method of stress vs number of cycle lifetime
is commonly used (σ-N). This method requires a large number of samples to be tested at
variety of stresses until they fail. When the failure stress is plotted against the logarithm
of number of cycles to failure, as shown in Figure 2.6, the curve features low and high
cycle limits, and a transition region.
Curves of σ-N were developed for use with polycrystalline engineering materials in which
dislocations and slips between shear planes propagate to form crack many times larger
than the grain size. σ-N curves can be used to estimate the lifetime of a whole part,
but the curves cannot describe the time to failure for a sample which already contains
62 Chapter 2. Polymer selection
a crack (Teoh, 2000), or be readily applied to alternative geometries or stress situations
(Papadopoulos, 2005). As such, I needed to identify an alternative method.
102 103 104 105 106 107 108
0
0.5
1
1.5
2
2.5
Number of cycles to failure
S
tr
es
s
Figure 2.6: Typical σ-N curve for a cyclically loaded sample exemplifying the pertinent
features. During experimentation the limiting plateau which is found in high cycle fatigue
may occur much later than indicated, and thus not be reached.
When a polymer is cyclically strained the chains will rearrange, a process which is referred
to as dynamic stress relaxation. It is reasonable to believe that at low strains these
rearrangements are reversible and the polymer behaves elastically, however at high strains
these rearrangement can eventually lead to sample failure. El Fray and Altsta¨dt (2003a,b,
2004) made extensive use of a test using stepwise increases in load or strain. Their samples
underwent a fixed number of cycles to a particular strain, that was increased stepwise
every 10,000 cycles. By analysing the change in elastic and viscous moduli, and the
damping factor, it might be possible to spot when a polymer moves from one stretching
regime to another, and thus determine a fatigue limit, however no clear changes were
observed in their data (Puskas et al., 2009b).
The failure of polymers can also be predicted by examining the nucleation of cracks and
then their growth until failure of the material. This method has been used extensively
by the automobile industry (Mars and Fatemi, 2002). Fracture mechanics can be used to
model the behaviour of the crack. The initial crack may be present due to an inherent
flaw in the material, as a result of chemical action, or spontaneous nucleation when the
material is strained. If the material is then strained the crack may grow. In this thesis
I use the following conventions: extension or stretch ratio, λs =
L
L0
, where L is length
and L0 is the unstrained length; and (engineering) strain,  = λ − 1. I will also assume
incompressibility, ∆V = 0, so λxλyλz = 1.
Chapter 2. Polymer selection 63
At the tip of a crack the stress becomes concentrated. If the stress exceeds a critical stress,
the crack will grow. This requires the stress distribution around the every crack tip to be
known, as well as the inherent critical stress property for the material. The work carried
out by Thomas (1994) indicated that finding a relationship between a “critical stress
criterion” and a measurable tearing force would be “impossible”. As a result, subsequent
analyses have been based upon an energy criterion, which can be applied over a volume
around the tip, rather than a single point maximum.
The use of an energy criterion stems from work by (Griffith, 1921) who applied the
concept of a minimum energy criterion to the problem of brittle material fracture. Rivlin
and Thomas (1953) then extended this energy criterion method to be used with rubbery
materials. This subsequently spawned the development of the J-integral concept which
can be used to define the stress and strain fields around the crack tip (Cherepanov, 1967;
Rice, 1968). The minimum energy criterion concept assumes that a crack will grow if
the elastic stored energy, U , released is greater than the free energy required to create a
new surface. For a growing crack, the formation of new area, ∂A, causes the specimen to
relax, leading to a decrease in stored elastic energy, ∂U . This decrease in stored energy
is referred to as the strain energy release rate or tearing energy, Te,
Te = −
(
∂U
∂A
)
l
(2.8)
where ∂A refers to one face of a crack and l signifies that the change occurs at constant
displacement. Whereas a polycrystalline material will fail due to rapid crack growth once
the tearing energy rate at the crack tip exceeds a critical value, exceeding this critical
value in an elastomer simply leads to slow, incremental crack growth. One reason for
this lies in the hysteresis of elastomers. Once a critical stress is exceeded, new surface is
formed which dissipates energy and reduces the strain in the surrounding material. In
a hysteretic material, a small reduction in strain results in a much greater reduction in
stress than for an elastic material, hence reducing the strain energy density available to
form new surfaces (Lake and Thomas, 1967).
In the following analyses, the fastest form of crack propagation is considered: mode 1
tensile fracture where the crack grows normal to the principal tensile stress direction. If
the crack extends through the thickness of the material then the area, A, of one face of
the crack is a product of the crack’s length, c, and the sample’s thickness, t. If, according
to Equation (2.8), the static strain is large enough, then it is favourable for a new surface
to form and the crack to grow, as opposed to stretch further. A similar relation holds
for cyclically strained sample, where the growth of the crack depends on the maximum
64 Chapter 2. Polymer selection
value of Te which is reached during a cycle (Gent et al., 1965). The cyclic crack growth
rate can be related to the energy release rate as,
dc
dN
= f(Te) (2.9)
where c is the length of the crack, N is the number of cycles and Te is the maximum
tearing energy in a cycle. Rivlin and Thomas (1953) demonstrated that the relation
between crack growth rate and Te, as shown in Equation (2.9), is independent of sample
geometry, allowing various materials to be compared. Over a large range of strain values,
the relationship between energy release rate and crack growth rate can be fitted to a
power law model (Mars and Fatemi, 2003):
dc
dN
= BT Fe (2.10)
where B and F are the coefficient and exponent, respectively, in this power-law model.
Te itself is a function of the type of strain and the strain energy density, W .
The most simple sample geometry in which a crack can be grown is the planar tension
crack growth geometry (also known as single edge cut or pure shear geometry) as shown in
Figure 2.7a, which produces a plane stress state (λx =
1
λz
, λy = 1, where the x-component
of λ is λx, and so on). The energy release rate is calculated as:
Te = WL0 (2.11)
where W is the strain energy density in the sample and L0 is the unstrained length of
the sample. In this geometry the crack grows at a constant rate independent of the crack
length (Papadopoulos, 2005).
By contrast, tensile bar (dogbone or dumbbell) geometries produce an uniaxial strain,
where λx =
1√
λy
= 1√
λz
. This strain state is commonly found in crack nucleation test
pieces without a pre-formed crack. This state also has an additional factor in its energy
release rate formulation:
Te = 2kgWc (2.12)
where c is crack length, W is the strain energy density, and kg is a function of strain
and arises due to the change in aspect ratio of cracks when stretched. It can be de-
rived analytically, though an experimental fit is generally used, which has the following
Chapter 2. Polymer selection 65
(a)
x
y
z
(b)
Figure 2.7: Tensile testing samples: (a) planar tension, pre-cracked specimen schematic,
and (b) is a uniaxial tension specimen schematic with side crack. The principal strain is
in the x-direction for both samples.
dependence on strain,  (Mars and Fatemi, 2002):
kg =
2.95− 0.08
(1 + )
1/2
(2.13)
Alternatively, trouser, angle or split crack test pieces can be used, though they also have
a tearing energy dependent on crack length and strain (Mars and Fatemi, 2002).
2.1.7 Lifetime parameter
Equations (2.10), (2.12), and (2.13), can be combined and integrated to give the number
of cycles required to grow a crack from an initial flaw, to the length required for failure
during unaxial cyclic strain:
Nf =
1
F − 1
1
B(2kgW )F
[
1
a0F−1
− 1
afF−1
]
(2.14)
where kg is the parameter dependent on strain geometry, W is the strain energy density,
a0 is the initial flaw size, and af is the critical flaw size for failure. In general, af  a0,
and so Equation (2.14) can be simplified to:
Nf =
1
F − 1
1
B(2kgW )F
1
a0F−1
(2.15)
66 Chapter 2. Polymer selection
2.1.8 Other factors
Although once Te is calculated it is independent of sample geometry, there are several
other factors that can play a part in fatigue life. Radir and Thomas (1981) measured
crack growth rates for samples with thicknesses between 0.1 and 10 mm. For samples
either below 0.5 mm or above 10 mm, crack growth rates were independent of thickness
and the crack surface was smooth. Between these bounds, rough or stick-slip crack growth
occurred and the rate changed by more than one order of magnitude. They attributed
this dependence to the crack tip roughness being of the same order of magnitude as the
sample thickness. I tested samples with a thickness of ≤ 0.5 mm, similar to the expected
heart valve leaflet thickness, which is in this regime.
Although it is true that the crack growth rate during cyclic loading is a function of the
maximum tearing energy, more factors must also be considered. Crack growth rate is
also a function of the strain ratio; R is defined as the ratio of the minimum strain in a
cycle to the maximum,
R =
min
max
(2.16)
where  is engineering strain. For purely tensile testing R is between 0 and 1. For a
sample that is fully relaxed when unloaded, R = 0. If R < 0 then the sample undergoes
compression as well as tension during a cycle. The value of R is particularly important
for rubbers (such as natural rubber or polyisoprene) which undergo stress crystallisation.
In these materials, for a given Te, the fatigue life during cyclic straining will be increased
when R > 0. Heart valve leaflets become fully relaxed and not compressed at their
minimum strain so testing should be performed at R = 0.
Anisotropy
I propose to take advantage of anisotropy to reduce the strain energy density in the
leaflets of the valve. Anisotropic effects were not considered in the theory outlined in
Section 2.1.6. To apply the isotropic methods to anisotropic materials requires that the
nature of the stress intensification around the crack tip be the same. Sih et al. (1965)
demonstrated in both the isotropic and anisotropic cases that stress intensity field which
is found around the crack tip diminishes as 1√
r
, where r is the distance from the crack tip.
Hence, the assumption that crack growth occurs as a function of Te is valid (Equations
(2.8) and (2.9)) for the anisotropic case.
Chapter 2. Polymer selection 67
The phases of failure
I will now explain the risks associated with extrapolating fatigue models. In a polycrys-
talline σ-N curve shown in Figure 2.8, there are 3 regimes of failure corresponding to low
cycle fatigue, a transition regime, and high cycle fatigue. For crack growth rates derived
from macro crack growth curves, as described in Section 2.1.6, it is normally possible
to fit a power law model to the majority of crack growth rates. In the study by Lake
and Thomas (1967) on crack growth rates in rubber, they noticed that the power law
relationship between crack growth rate and energy release rate did not hold at low values.
They identified four regimes of differing behaviour, as shown in Figure 2.8. Below the
threshold tearing energy, T0, the crack growth rate was independent of Te. Immediately
above T0 crack growth had a linear dependence on Te. In regime 3, at a higher values
of Te, crack growth rate had a power law dependency. Finally, the crack growth rate
approached the velocity of elastic waves in rubber, resulting in catastrophic tearing in
regime 4. To safely predict lifetime using Equation (2.14), testing and device operation
must occur in regimes 2 or 3.
Intrinsic Flaws
It is generally assumed that the failure of a material is dominated by the growth of cracks
from some initial size, to a critical size. Even if a sample appears to be macroscopically
perfect, with no sharp cracks or flaws there will be intrinsic internal flaws. These could
originate from any of the following causes: filler particles, voids in the polymer, imperfect
mixing or dispersion of polymer ingredients, imperfections in the mould tool surface,
mould lubricants, impurities gained during the moulding process or stress concentrations
during cooling (Mars and Fatemi, 2002). It is conventional when predicting elastomer
lifetimes to use an effective initial flaw size of 0.02 to 0.06 mm (Lake and Thomas, 1967;
Zarrin-Ghalami and Fatemi, 2012).
In the fracture mechanic analyses for lifetime predictions that I have used here, I have
assumed continuity and homogeneity of the material. If initial flaws are too small, such
as those found in metals, then this may not be true. Fortunately, in polymeric mate-
rials the initial flaws are several orders of magnitude greater than polymer chain sizes
and filler particles. The domain spacing of cylindrical PS domains in a typical BCP is
approximately 30 nm (Stasiak et al., 2010). As flaws are generally much larger, on the
order of 40µm (Choi and Roland, 1996), continuity can be assumed.
68 Chapter 2. Polymer selection
Figure 2.8: Crack growth rate as a function of tearing energy for unfilled natural rubber
(NR O) and styrene-butadiene rubber (SBR ×). Four regimes are shown: 1 for Te < T0 in
which dc
dN
is a constant finite value attributed to environmental conditions (ozone attack);
2 above T0 where
dc
dN
is linear; 3 where dc
dN
has a power law dependency; and 4 where
rapid tearing occurred. From Mars and Fatemi (2002).
2.2 Materials and methods
The shortlist of materials and their specific details is shown in Table 2.2.
PU2363 is a thermoplastic polyurethane based upon methylene diphenyl diisocyanate
(Zdrahala and Zdrahala, 1999), which was used as a reference material. Results re-
garding biostability and mechanical properties of PU2363 are reported in the literature,
though when used in pacemaker leads, the combination of adverse chemical conditions
and stresses caused failure (known as environmental stress cracking) (Pinchuk et al.,
2008; Simmons et al., 2004; Thomas and Jayabalan, 2009). Polyurethanes are the “most
optimal” material as shown on the Ashby map in Figure 2.5, and demonstrated by their
frequent appearance in polymeric prosthetic heart valves. Sheets of PU2363 were fabri-
cated by solvent casting. The pellets were dissolved in toluene at 10w/v%, then dried on
Chapter 2. Polymer selection 69
Table 2.2: Block copolymers and reference polyurethane used in selection procedure.
Polymer
Styrene
content
(mol%)
Code Form Supplier
PS-b-polyisoprene-b-PS 18 D1163P Pellet
Kraton,
Belgium
PS-b-polyisoprene-b-PS 30 D1164P Pellet
Kraton,
Belgium
PS-b-polyisoprene-polybutadiene-b-PS 19 D1170B Pellet
Kraton,
Belgium
PS-b-polyethylene-polypropylene-b-PS 22 G1730M Pellet
Kraton,
Belgium
PS-b-polyisobutylene-b-PS 30 n/a Pellet
Innovia
LLC, FL,
USA
PS-b-polyethylene-polybutylene-b-PS 20 G1642H
Fluffy
crumb
Kraton,
Belgium
PS-b-polyethylene-polybutylene-b-PS 30 G1650E
Fluffy
crumb
Kraton,
Belgium
Polytetramethylene glycol based PU n/a
PU2363-
80AE
Pellet Velox, UK
a Teflon coated sheet, in a fume cupboard for a minimum of 24 hours.
Pellets and fluffy crumb BCPs were compacted between aluminium plates at 160 ◦C using
a heated press. Mechanically anisotropic samples with oriented PS domains were fab-
ricated using the method previously described by Stasiak et al. (2009). Silicone mould
release was applied before each moulding. A thick strip of polymer is placed in the centre
of the aluminium channel mould, as shown in Figure 2.9, and heated to 160℃. The plate
is slowly compressed within the channel resulting in planar growth of the sample. The
mould is cooled with water for about 2 minutes, to room temperature, before the sample
is removed. The sample thickness was controlled using shims placed in the mould. Sam-
ples with overall dimensions of 170× 170× 0.3 mm were produced. The extension ratio
during moulding was ≥ 10. Small-angle X-ray scattering (further information in Section
6.2.1) was used to confirm that the PS domains were oriented anisotropically.
2.2.1 Biostability
Compression moulded polymer samples, six of each type, were cut into strips which were
first dried and weighed 3 times before being placed in polypropylene Fisher centrifuge
70 Chapter 2. Polymer selection
Figure 2.9: Schematic of the moulding process for fabricating oriented samples. A long
block of polymer is placed in the centre of the mould. The mould is compressed, squeezing
the polymer into a flat sheet. The sides of the mould constrain the growth of the sheet
to one direction, leading to uniaxial orientation of the PS cylinders in the sample.
tubes. Distilled water or hydrogen peroxide solution (30%) was added to each tube so
each sample was immersed according to ISO 10993:13. The tubes were stored for 3
months at 70 ◦C. The samples were dried and weighed after the degradation test.
Six samples of each material were also stored at room temperature in oxidising and acidic
conditions. These samples were analysed using FTIR with Attenuated Total Reflectance
(ATR) using a KBr crystal.
2.2.2 Tensile properties
The uniaxial tensile properties of the polymer samples were measured using tensile dog-
bone samples which were cut using a sharp edged stamp. The planar tensile properties
of the polymer were measured using a sample with a high aspect ratio. These samples
were cut using a sharp scalpel or scissors. The dimensions of both sample types are
illustrated in Figure 2.10. Samples were cut such that they are stretched either parallel
(L) or perpendicular (R) to the PS domain orientation. Testing was performed using a
(a) (b) (c)
Figure 2.10: Dimensions of samples used for mechanical testing, all with a thickness
of 0.3 mm). 2.10a is the tensile bar for uniaxial testing, 2.10b is planar tension and
crack growth sample (shown with initial crack), and 2.10c is the flexural sample with the
deflecting and support beams indicated by — and -·- respectively.
Chapter 2. Polymer selection 71
Texture Analyser (TA) (Stable Microsystems, Surrey, UK) using a 30 kg load cell. To
test the tensile bar samples, commercially available clamps with rubber coated jaws were
used. For the planar tensile samples a set of aluminium clamps with a width of 180 mm
were fabricated. Samples were secured using 10 cap head bolts which provided an even
gripping force along the sample. The faces of the jaws were coated with neoprene rub-
ber. The TA collected force and displacement data at 200 Hz which were subsequently
analysed using MATLAB.
Although it would be preferable to perform this mechanical testing in combination with
the chemical environment, this would require a significantly more complex sample mount
and could lead to a reduction in sensitivity (Puskas et al., 2009a). Instead, the effects of
mechanical and chemical forces were tested separately during selection.
Samples underwent a sinusoidal displacement input with a strain ratio of R = 0 (fully
relaxed minimum strain). During tensile testing the maximum displacement was in-
crementally increased, with 120 cycles performed at each strain level. The TA system is
displacement controlled, and so strain was determined during a MATLAB post-processing
routine.
Strain energy density, W , at strain , for a fully relaxing stable cycle was calculated using:
W =
∫ 
0
σd (2.17)
2.2.3 Flexural properties
Unidirectionally oriented, compression moulded samples (dimensions shown in Figure
2.10c) were tested according to ASTM D790, with bending both perpendicular (R) and
parallel (L) to the PS cylindrical domain direction (ASTM International, 2002). Support
bars were spaced by 20 mm, the crosshead speed was 5 mm s−1 and reached a maximum
deflection of 5 mm. 20 cycles were performed on each sample to reduce transient stress
relaxation. Flexural strain in the 3 point bend scenario is calculated as follows:
flex =
6Dt
L2
(2.18)
where flex is the strain at the outer surface, D is the central deflection, t is the sample
thickness, and L is the support span.
Flexural stress in the outermost fibres for samples with a rectangular cross-section is
72 Chapter 2. Polymer selection
calculated using:
σflex =
3FxHL
2bf t2
[
1 + 6
(
D
L
)2
− 4tD
L2
]
(2.19)
where σflex is the stress at the outer surface, FxH is the crosshead deflector load, bf is the
sample width, L, t and D are the sample dimensions as before. The flexural modulus is
calculated as a ratio of stress to strain:
Eflex =
σflex
flex
(2.20)
For an anisotropic material, the flexural modulus is dependent on the direction of flexure
relative to the fibre direction. Parallel modulus, Eflex‖, refers to the stiffness when the
contours of constant displacement are parallel to the stiff cylinders, and Eflex⊥ refers to
the perpendicular modulus.
2.2.4 Crack growth
Planar samples were cut from the oriented sheets and a 25 mm crack was created in the
sample (Figure 2.10b) using a surgical scalpel. The samples were sinusoidally strained
with R = 0. The samples were conditioned using 300 to 500 cycles at 110% of the
test strain to establish a natural fracture surface and crack tip, and reduce transient
stress relaxation. At high crack growth rates, 2000 cycles were performed, at low rate
10,000. Crack length was measured by recording images with a camera every 100 s with
subsequent manual analysis in MATLAB. The resolution of the camera was sufficient to
allow the crack length to be measured to a precision of 0.05 mm.
2.3 Results
2.3.1 Manufacturing
All the tested polymers were thermoplastic and flowed during compression moulding.
Several important observations were made during polymer processing, I observed that
many of the materials displayed significant shrinkage when cooled. This occurred regard-
less of the rate at which the materials were cooled, though was greatest when cooling
was fastest. The shrinkage was greatest for SIBS and led to the shrink-induced cracking
as shown in Figure 2.11. This could be mitigated by lining the moulds with non-stick
Chapter 2. Polymer selection 73
parchment, although this left the surface with a mottled texture and would be infeasible
in a complex injection moulding tool.
Figure 2.11: Sample of SIBS30 (width=17 cm) after compression moulding. The sample
was cooled while in the aluminium mould, resulting in rapid shrinkage. The polymer
adhered to the mould, and so holes were formed as the sample relaxed.
The difficulty in moulding this polymer is supported by reports from the Stony Brook
group, who also compression moulded SIBS materials. They reported oxidation of SIBS
during moulding, which may have occurred from their use of high temperatures.
2.3.2 Biostability
I used the relative stability of the polymers in aqueous, acidic, and oxidising conditions
as a criteria for selection. At room temperature, I did not record any mass change in the
polymers, though FTIR indicated that oxidation may have occurred due to the fall in the
C−C peak at 1660 cm−1 (Figure 2.12), for the unsaturated polymers (SIS and SI/BS).
Figure 2.13 shows the proportion of mass lost from each sample when incubated at 70 ◦C
in air, hydrolysing (H2O), and oxidising (H2O2) conditions. There was a loss in mass over
60 days at temperature regardless of whether the materials were stored in air, hydrolysing,
or oxidising conditions. There is a large spread in some results as sample containment
was occasionally lost. Most notably, severe (> 2%) mass loss was observed for the
unsaturated BCPs, containing an isoprene (SIS18 and SIS30) or butadiene (SI/BS19) soft
block, causing the samples to lose their structural integrity. I eliminated SIS18, SIS30
and SI/BS19 from the selection procedure. The saturated BCPs (SE/BS20, SE/BS30,
SIBS30 and SEPS22) become the preferred materials for use in a prosthetic heart valve.
74 Chapter 2. Polymer selection
1000150020002500300035004000
0
5 · 10−2
0.1
0.15
O
* 
Wavenumber cm−1
A
b
so
rb
an
ce
SIS30
Acid
Processed
H2O2
(a)
1,0001,5002,0002,5003,0003,5004,000
0
0.2
0.4
0.6
C-O stretch
1050-1150
O* 
Wavenumber cm−1
A
b
so
rb
an
ce
SIS18
Acid
Processed
H2O2
(b)
1,0001,5002,0002,5003,0003,5004,000
0
5 · 10−2
0.1
0.15
O* 
Wavenumber cm−1
A
b
so
rb
an
ce
SIBS30
Acid
Processed
H2O2
(c)
1,0001,5002,0002,5003,0003,5004,000
0
5 · 10−2
0.1
O
* 
Wavenumber cm−1
A
b
so
rb
an
ce
SI-BS19
Acid
Processed
H2O2
(d)
Figure 2.12: (a) to (d) FTIR spectra of polymer samples before (green) and after degra-
dation in oxidising and acidic conditions. *, O, and  indicate the regions of absorbance
due to the stretching of O−H, C−C, or C−O bonds which occur at 3700 to 3200, 1680
to 1620, and 1760 to 1665 cm−1 respectively. (a) shows SIS30, (b) shows SIS18, (c) shows
SIBS30, and (d) shows SI-BS19, refer to Table 2.2 for material details.
0.2
0.4
0.6
0.8
1
M
as
s
lo
ss
p
ro
p
or
ti
on Air
Water
Oxidising
SE/BS20 SE/BS30 SI/BS19 SIBS30 SIS18 SIS30 SEPS22 PU2363
0
0.02
0.04
0.06
0.08
0.1
Figure 2.13: Proportion of mass lost during accelerated degradation test for eight samples
incubated at 70 ◦C in air, oxidising, and acidic conditions. Error bars indicate 1 standard
deviation from the mean. See Table 2.2 for details of the materials.
Chapter 2. Polymer selection 75
2.3.3 Tensile properties
I eliminated SIS and SIBS materials from the selection procedure (on the basis of chem-
ical stability and process-ability), and so the shortlist comprises SEBS29, SEBS20 and
SEPS22. I performed tensile tests on the planar and tensile bar geometries for each poly-
mer. A quadratic fit was plotted to the stress-strain data from 0 to 50% strain, as shown
in Figures 2.14a and 2.14b. All the polymers have a lower modulus at higher strain, and
are thus strain softening. The polyurethane (PU2363) is the least strain softening. This
may indicate that the PS cylinders in the styrenic BCPs can be broken when strained.
0 5 · 10−2 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5−0.5
0
0.5
1
1.5
2
2.5
3
3.5
4
Strain
S
tr
es
s
(M
P
a)
SEBS29 L
SIBS30 L
SEBS20 L
SIS30 L
PU2363
SEPS22 L
(a)
0 5 · 10−2 0.1 0.15 0.2 0.25 0.3 0.35−0.5
0
0.5
1
1.5
2
2.5
3
3.5
4
Strain
S
tr
es
s
(M
P
a)
SEBS29 L
SIBS30 L
SEBS20 L
SIS30 L
PU2363
SEPS22 L
(b)
Figure 2.14: Tensile stress strain properties of the shortlisted polymers, from an extending
stroke during a sinusoidal oscillation, plotted as a mean quadratic fit from all samples
(n≥ 3 for each polymer). Errors are indicated by the modulus measurements shown in
Table 2.3. (a) are the planar samples and (b) are the uniaxial samples.
For crack propagation and lifetime analyses, the strain energy density relation is more
appropriate (Section 2.1.6). These data were collected by performing sinusoidal oscilla-
tions at increasing strains, with 30 oscillations at increasing maximum strains. Strain
energy density up to the maximum strain was calculated for 5 cycles at each strain level.
Up to 10 strain levels were tested, between 0 and 50% strain. A quadratic curve was a
good fit to these data, which are shown in Figure 2.15.
The use of a stepwise increasing strain - as opposed to the integration of a single stress-
strain curve up to 50% strain - is appropriate for this scenario. The candidate polymers
are viscoelastic, and so are sensitive to strain rate and the maximum strain that is reached
during an oscillation. The stepwise method avoids the strain response at low strains being
masked by artefacts from high strains.
For ease of comparison, the moduli of the polymers is summarised in Table 2.3. The
76 Chapter 2. Polymer selection
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
0.5
1
1.5
2
2.5
3
3.5
Strain
S
tr
ai
n
E
n
er
gy
D
en
si
ty
(M
P
a)
SEBS29 L
SIBS30 L
SEBS20 L
SIS30 L
PU2363
SEPS22 L
(a)
0 5 · 10−2 0.1 0.15 0.2 0.25 0.3 0.35
0
0.2
0.4
0.6
0.8
1
Strain
S
tr
ai
n
E
n
er
gy
D
en
si
ty
(M
P
a)
SEBS29 L
SIBS30 L
SEBS20 L
SIS30 L
PU2363
SEPS22 L
(b)
Figure 2.15: Tensile strain energy density properties of the candidate polymers for planar
(a) and uniaxial (b) geometries. At least 3 samples were used for each geometry Each
curve is a quadratic fit to the maximum strain energy density from 30 sinusoidal oscilla-
tions at stepwise increasing maximum strains. Errors are indicated by the results shown
in Table 2.3.
SEBS material was tested with a 20 and 29% PS fraction and, as expected, the higher
PS fraction results in a higher modulus.
Table 2.3: Young’s moduli (E), secant moduli at 20% strain (Esec20), and strain energy
density at 20% strain (W20) (all MPa) for planar (n≥ 3) and dumbbell (n≥ 4) samples.
Candidate BCP materials were measured parallel (L) to the PS cylinder orientation and
the reference polyurethane is isotropic. Listed errors are the maximum recorded deviation
from the mean value.
Planar Dumbbell
Material E ± Esec20 ± W20 ± E ± Esec20 ± W20 ±
SEBS20 L 5.2 1.3 4.8 0.9 0.23 0.10 4.5 0.7 4.2 0.5 0.15 0.03
SEBS29 L 10.1 3.0 9.5 1.9 0.44 0.14 12.6 2.9 10.3 1.9 0.34 0.05
SIBS30 L 4.0 0.5 4.3 0.5 0.18 0.08 4.7 0.3 5.2 0.3 0.25 0.07
SIS30 L 6.4 1.2 5.9 1.1 0.25 0.18 4.0 1.5 4.4 1.3 0.20 0.10
PU2363 8.3 1.1 8.8 1.4 0.42 0.29 10.4 1.1 10.9 0.6 0.45 0.09
SEPS22 L 5.0 0.6 4.1 0.7 0.23 0.05 8.3 0.4 7.6 0.2 0.31 0.04
2.3.4 Flexural properties
I collected data on the flexural properties of the materials when PS domains are oriented
both parallel and perpendicular to the bending direction (Table 2.4). The flexural modu-
Chapter 2. Polymer selection 77
lus gives an indication of the maximum force and energy required to reverse the curvature
of the leaflet.
The flexural properties of the material are a factor (along with the leaflet thickness) in
the energy required to open the heart valve leaflets. During bending, a variety of strains
are experienced throughout the thickness of the material. As the samples are strained
beyond the linear region of stress-strain, it is more appropriate to compare flexural secant
moduli which amalgamate the stress response for a material in a variety of strain states.
During systole the native valve experiences flexural strains up to 15% (Balachandran
et al., 2011).
Using the flexural stiffness data in Table 2.4, the anisotropy ratios, Eflex⊥/Eflex‖, were cal-
culated. The anisotropy ratios of the materials vary between 3.15 (SIBS30 and SEBS20)
and 9 (SEBS29). In the case of SEBS the ratio increased from 3.15 to 9 as PS content
changed from 20% to 29%. For the SIBS and SEPS materials, the manufacturer supplies
the materials with differing PS content, however they do not all exhibit a cylindrical
(anisotropic) morphology.
2.3.5 Fatigue and performance indices
I recorded crack length against number of cycles. In general, samples showed a consistent
crack growth rate with respect to the number of cycles. I halted experiments when at
least eight crack lengths could be recorded, or the width to length ratio of the sample
became too low. As an example of the data, the crack tip position for SEBS20 under
50% sinusoidal strain is plotted in Figure 2.16.
I frequently observed nucleation of cracks away from the artificial crack tip. These spon-
taneous cracks were generally adjacent to the clamps. In many cases the cracks that
formed were relatively stable and the test crack continued to grow at an unchanged rate.
Interestingly, in the case of SIS30 I observed the bridging of the crack tip, a phenomenon
that may arise from the strain crystallisation that is observed in natural rubber (polyiso-
prene) (Zarrin-Ghalami and Fatemi, 2012). Sample to sample variation may have arisen
due to the poor control of cooling rate. The plates of the mould were cooled using a
slow stream of running water from a tap. An improvement for future work should utilise
embedded oil cooling channels, giving better control of cooling rate.
Figure 2.17 shows the crack growth data and tearing energy, with a power-law fit. The
coefficient and exponent of the power law models can then be used to determine the
fatigue life for the polymer as a function of strain energy density using Equation (2.14).
78 Chapter 2. Polymer selection
0 2000 4000 6000 8000 10000
0
5
10
15
20
25
30
35
40
45
Number of cycles
Cr
ac
k 
le
ng
th
 / 
m
m
Figure 2.16: Crack tip position for SEBS20 with PS cylinders oriented parallel to the
stretching direction with εmax=0.21. Linear fit is c = 0.00165N + 23.7.
I determined the maximum allowable strain energy density required to reach 107 cycles.
100 101 102
10−5
10−4
10−3
10−2
Strain energy release rate (log kJ)
C
ra
ck
gr
ow
th
ra
te
(l
og
m
m
/c
y
cl
e)
SEPS22 L
SEBS29 L
SEBS20 L
SIS30 L
SIBS30
PU2363
Figure 2.17: Crack growth rate against strain energy density for the shortlisted BCPs
with PS cylinders oriented in the direction of strain. A power law model is fitted to each
set of points, and the coefficients are reported in Table 2.4.
The macroscopic crack growth experiments allow prediction of lifetime based upon Equa-
Chapter 2. Polymer selection 79
tion (2.15). A maximum strain energy density, Wmax; and stress, σmax, can be determined
based upon a minimum fatigue life of 1× 107 cycles, corresponding to a valve lifetime of
about 100 days at 1 Hz. This lifespan lies within the power-law region (region 3 in Figure
2.8) and could also be compared to industry fatigue data, for example the data which
can be obtained from Granta Design’s CES Material Selection software, and serves as an
absolute minimum for a material to be used in the heart valve.
I combined the leaflet opening (Equation (2.2)) and fatigue (Equation (2.4)) constraints
to form the performance index. Eliminating the constants yielded a material performance
index, IP , which is minimised for optimal materials:
min (IP ) ∝
(
Eflex
σf 3
)
(2.21)
A crucial tenet of this thesis is that leaflet anisotropy should lead to improved prosthetic
heart valves. There are several ways in which anisotropy may improve the valve; first,
in Section 1.10.1 the results of Zaffora (2011) were cited. He optimised anisotropy using
FE modelling, which led to a reduction in the maximum leaflet stress during diastole. It
may be the case that the material’s durability is also anisotropic, and so a greater strain
energy density and stress can be accommodated in the fibre direction. These advantages
may allow the denominator of Equation (2.21) to be increased, reducing IP , to I‖.
Anisotropy may bring further reductions to IP by reducing the numerator of Equation
(2.21). The stresses generated in a leaflet during opening are dominated by flexure. If
the stiff fibres are oriented parallel to the contours/flexure direction/fold line, a lower
force is required to bend the material than if they are oriented perpendicular. If leaflet
opening is limited by bending in the belly of the leaflet (as will be discussed in Chapter
5), then horizontally oriented cylinders in the belly allow the leaflet to open more easily,
faster and further.
Prosthetic heart valve leaflets have a Gaussian curvature greater than zero (i.e. they do
not have a perfectly cylindrical form). They can be defined by two radii at each point with
the tighter of these radii requiring more energy to reverse. By orienting stiff cylinders
perpendicular to the tighter radius, the energy require for reversal can be reduced. For
a hemispherical surface, where both radii of curvature are equal, anisotropy causes one
direction to reverse before the other, along a lower energy path (Lu and Xu, 2003).
Practically, this means that anisotropy in the leaflet may allow the curvature to reverse
in one direction before the other, requiring less force.
Optimal anisotropy can be included in the material performance index (previously Equa-
80 Chapter 2. Polymer selection
tion 2.21), so for the best case material performance index, I‖, as:
min
(
I‖
) ∝ (Eflex‖
σ3f‖
)
(2.22)
where Eflex‖ is the flexural modulus when bending contours are parallel to the fibre
orientation, and σf‖ is the fatigue stress when the material is stretched parallel to the
fibre direction. If the above assumptions regarding leaflet opening are not true, then in
the worst case scenario, where bends are perpendicular to the oriented cylinders. An
alternative performance index, I⊥ can be expressed using Eflex⊥ in place of Eflex‖ in
Equation (2.22).
I analysed the materials using both performance indices, and these results, together
with the constituent predicted fatigue stress and flexural moduli are shown in Table
2.4. The strain energy density and crack growth curves (as shown in Figure 2.17) are
characterised by power-law model exponents, F , and coefficients, B. As Wmax is obtained
by extrapolating the curves, it is dominated by the value of F .
Table 2.4: Summary of durability and flexural stiffness properties which yield the per-
formance indices for the candidate materials. F and B correspond to the exponent and
coefficient in the crack growth model (Equation 2.10). Fatigue lives are estimated using
the following parameters (from Equation (2.14)) Nf = 10
7, c0 = 0.02 mm, cf = 0.3 mm
and k = 1.2. Wmax and σmax are recorded in MPa. I‖ assumes that the critical bending
direction during opening is parallel to the cylinder orientation - and so opening is easier.
I⊥ assumes perpendicular.
Crack growth 107 cycle fatigue limits Eflex (GPa) Flexural PI
Material B F Wmax σmax ⊥ ‖ I‖ I⊥
SEBS20 L 5.7× 10−4 1.29 0.002 0.09 1.3 4.1 1559.7 4873.8
SEBS29 L 5.5× 10−4 1.05 0.0003 0.18 1.7 15.3 312.4 2792.1
SIBS30 L 1.9× 10−7 4.63 5.89 1.13 1.1 3.5 0.8 2.4
SIS30 L 2.5× 10−5 1.43 0.041 0.22 0.4 2.3 36.6 230.0
PU2363 1.3× 10−6 2.52 1.35 0.84 36.1 36.1 61.1 61.1
SEPS22 L 8.3× 10−6 3.83 1.81 0.45 0.8 4.1 8.9 43.9
I compared materials using the performance indices I⊥ and I‖. Having eliminated SIS
and SIBS materials from the selection procedure (on the basis of chemical stability and
process-ability) the shortlist comprises SEBS29, SEBS20 and SEPS22. SEPS22 has the
lowest I⊥ and I‖, suggesting that it is the best material to select. SEPS22 is also predicted
to perform better than PU2363, the reference polyurethane (Thomas and Jayabalan,
Chapter 2. Polymer selection 81
2009). To put the differences in performance index in a physical perspective, if SEPS22
and PU2363 valves were designed with optimal thicknesses such that they can reach an
equal predicted lifetime, the PU2363 valve would require an opening force at least 50%
greater than the SEPS22 valve (in the most “poorly” oriented scenario).
2.4 Discussion
BCPs can be engineered to have anisotropic properties amenable to producing a bioin-
spired polymeric heart valve. It is relatively easy to envisage how stiff PS cylinders may
appear like a short-fibre reinforced polymer. The proposal to embed fibres in a prosthetic
valve is not novel. Cacciola et al. (2000) embedded high performance polyethylene fibres
in an EPDM matrix (using a dip coating process). The valve was tested in vitro but
suffered from excess regurgitation. Durability testing of these prototype valves was not
reported in the literature. An alternative fibrous polymeric prosthetic heart valves was
proposed and tested by Gallocher et al. (2006) who observed catastrophic in vivo fatigue
failure of SIBS coated polyester leaflets, which was attributed to fibre-bulk dehiscence.
The need to consider all means of degradation and durability from the outset of design is
imperative. The novelty of using BCPs to provide fibre reinforcement comes from their
self-assembly during processing and the intimate chemical linkage between the matrix
and fibres.
There is reason to be optimistic that BCP fibres are less susceptible to delamination
than conventional fibres. BCP fibres are fundamentally different from short-fibre filled
composite polymer as continuous polymeric chains pass between the fibres and the bulk,
covalently bonding the fibres to the bulk and reducing the chance of delamination.
In this chapter my aim was to select a suitable polymer. The first criterion that materials
must pass was adequate biostability. No absolute biostability test exists, and so a stan-
dard, relative, accelerated testing method was employed. Stasiak et al. (2011) proposed
that unsaturated polymers - which are generally disregarded due to their vulnerable C−C
bonds - might be suitable if they are used at body temperatures in the absence of UV
radiation. Although they may be stable under physiological conditions, the accelerated
testing definitively demonstrated that the unsaturated materials are less stable than sat-
urated ones. Thus, unless the fatigue or biocompatibility performance of the unsaturated
polymers is significantly superior, they should not be considered as potential materials
for a prosthetic heart valve.
During compression moulding, I was surprised to observe serious crack-inducing shrinkage
82 Chapter 2. Polymer selection
of the polymers, particularly of SIBS30. Temperature dependent shrinkage generally
results in ≤1% polymer shrinkage (Brydson, 1995). Shrinkage is of most relevance in the
polymer rolling industry, where mill rollers are used to produce sheets of polymer. Upon
exiting the rollers the sheets shrink in width and length, while growing in thickness,
giving rise to the term “mill shrinkage”. Four processing regions have been identified
in the rolling industry; in the first three regions shrinkage leads to failure (splitting,
crumbling, roughening), and only in the fourth region can a part be properly formed
(Tokita and White, 1966). Shrinkage after moulding can be attributed to the relaxation
of the strained polymer chains over various time scales. The viscoelastic features of
polymers are ever-present. At higher temperatures they continue to exhibit solid-like
yielding and tearing, which can occur as a polymer shrinks (White and Tokita, 1968).
To remedy shrinkage one can reduce the strain and strain rate, or increase the tempera-
ture. Variation in the polymer molecular weight distribution alters shrinkage, but does
not eliminate it (White and Tokita, 1968). In the plate compression moulding system,
I was not able to eliminate shrink-induced cracking despite reducing the strain rate to
0.02 s−1 and raising the temperature to 200 ◦C - above which oxidation occurred. How-
ever I did find that it was possible to alleviate the cracking associated with shrinkage
by lining the plates with baking parchment, relieving the pressure from the plates af-
ter moulding, and cooling slowly. The baking parchment reduced the “pinning” of the
polymer which leads to tears during shrinkage. Pressure relief allowed the polymer to
contract and relieve internal stress by increasing its thickness. The use of silicone grease
to reduce pinning is plausible, but leaves residues which are undesirable for fatigue and
biocompatibility considerations. As the above techniques would not be suitable for use
in a prosthetic heart valve tool, I eliminated SIBS30 from the selection process.
The polymeric materials I tested are comparable in stiffness to loaded, native aortic
valve tissue. Thubrikar et al. (1980) measured the native leaflet modulus in vivo, in dogs
using fluoroscopic markers in conjunction with pressure measurements. They reported
the leaflets to have a modulus of 5.4 MPa during diastole, which is similar to the BCPs
with a PS content of ∼ 20%. The biological tissues are of course non-linear in their
stress-strain relationship, meaning that the first few % of strain occur when the tissue’s
stiffness is 0.5 to 0.003 times lower than the stiffness when loaded (and strain reaches
10%) (Thubrikar et al., 1980). The polymeric materials that I tested do not possess such
a useful property and thus this feature of the native tissues cannot be mimicked.
In the prediction of fatigue, I observed significant variations in the maximum strain energy
density and stress that each material could withstand. SIBS30 was able to withstand the
greatest stress, though it did suffer from spontaneous crack nucleation which might have
Chapter 2. Polymer selection 83
arisen from moulded-in stresses, poor orientation, large flaws, or contaminants during
moulding. Pinchuk et al. (2008) also claimed that SIBS had a high fatigue life but
suffered from creep. The durability of the elastomers containing polyisobutylene may be
due to its stereoregularity which increases strain crystallisation (Gonzalez-Alvarez et al.,
2004). SEPS22 was the next best material. The SEBS materials displayed the worst
fatigue parameters. It is not advisable to directly compare these fatigue predictions with
those values issued by ASM via the CES EduPack (Granta Design, Cambridge, UK) as
they use R = −1, indicating compression and their values are estimates.
El Fray et al. (2006) also performed a fatigue study on SIBS30 using their stepwise increas-
ing strain/stress test (SIST). The material was isotropic (due to uncontrolled compression
moulding), and was compared to medical polyester, polyurethane and silicone materials.
The SIST method was not able to provide a stress or strain limit for samples, making
it hard to make quantitative comparisons. However, the authors did note that SIBS30
crept significantly, which could be attributed to the rearrangement of PS domains in
their isotropic samples. The method I have utilised in this study required all materials
to be pre-conditioned to remove the significant primary creep which occurs in the first
100 cycles.
Prediction of fatigue requires the extrapolation of crack growth parameters to low strain
energy densities. The results of Lake and Thomas (1967) covered a vast range of crack
growth rates, as is shown in Figure 2.8. I compared the crack growth rates recorded for
BCPs in these experiments (Figure 2.17) with the results of Lake and Thomas (1967)
(note the differing units for crack growth rate). The critical strain energy density for
survival, Wmax, can be converted to a tearing energy, Te. The results I report were within
the bounds of regime 3 which fit to the power-law model observed by Lake and Thomas
(1967). Regime 2 may also be approximated by a continuous power law, though their
experimental results start to deviate, suggesting that a linear model may become more
appropriate. At T0 there is a step decrease in crack growth rate, leading to a regime of
no crack growth where the power law model would be inappropriate, in the absence of
chemical factors (regime 1).
With reference to the results in Figure 2.17, the minimum permissible tearing energy to
reach 107 cycles is ∼1× 10−1 kJ m−2, which is greater than the lower limit of regime 2
reported by Lake and Thomas (1967), indicating that these results can be extrapolated to
this number of cycles. Below 1× 10−1 kJ m−2, crack growth rate will continue to decrease
up to the threshold tearing energy T0 at which point the crack growth rate will drop to
zero.
While extrapolation may be safe in this range, the role that environmental factors in
84 Chapter 2. Polymer selection
fatigue, such as swelling, has not been accounted for (Cho et al., 2000). Furthermore,
this method is unlikely to hold up during multiaxial loading scenarios as not all the strain
energy is available for the growth of the crack (Mars and Fatemi, 2002).
The differing performance of materials may be attributed to differences in molecular
architecture, processing and conditioning. Roche and Perier (2013) investigated the in-
fluence of elastomer formulation on cyclic crack growth. Their study was focussed on
filled styrene-butadiene rubbers. They observed few differences between growth rates re-
gardless of whether filler particles were evenly or poorly dispersed. The most significant
differences that were observed in crack growth rates was between polymers which were
aged for different times, suggesting that specific aspects of processing may in fact be
dominant in the determination of elastomer lifetime. The growth of cracks in BCP may
be significantly influenced by the molecular architecture (Lach et al., 2005; Staudinger
et al., 2008). Using the Essential Work of Fracture (EWF) concept to compare various
blends of symmetric and asymmetric styrene-b-(styrene-random-butadiene)-b-styrene tri-
block copolymers, Staudinger et al. (2008) reported that crack blunting was controlled
by the dimensions of the soft block phase, whereas the rate of stable crack propagation
was controlled by the styrene phases.
The rate of crack growth can be influenced by polymer hysteresis (Mars and Fatemi,
2002). Polymers dissipate energy when stretched, which is reflected by hysteresis in a
stress-strain curve. Hysteresis reduces the rate at which cracks grow in elastomers (see
Section 2.1.6). Polymers with more hysteresis have a lower power law exponent, F , in
crack growth models of dc/dN vs. Te (Equation (2.10)) (Lake and Thomas, 1967). Further
work, outside the scope of this thesis, is needed to determine whether the observed
differences in crack growth rate can be attributed to one of the above factors.
2.5 Conclusion
For the design of a prosthetic heart valve, the conflicting goals of improving durability
while minimising resistance to blood flow must be reconciled. Myself, working in the
Cambridge Group, have been inspired by the anisotropic structure of the native heart
valve leaflets, and so have proposed a set of BCPs which can be processed to display
anisotropic mechanical properties. I tested the stability of the candidate materials under
accelerated degradation conditions (ISO 10993:13), which indicated that the unsaturated
BCPs were more susceptible to degradation. The saturated BCPs met the in vitro,
qualitative requirements of ISO 10993:13 for long term implantable materials.
Chapter 2. Polymer selection 85
I have proposed a novel method for selecting materials for polymer prosthetic heart valves
based upon their mechanical properties. Principally, this relies upon making predictions
of the materials’ cyclic mechanical fatigue life. I combined the experience from the test-
ing of rubbers for the automotive industry, with techniques from design engineering to
produce a method allowing polymeric candidate materials to be compared. The method
which I have developed is a general, quantitative, and robust means of rapidly assess-
ing materials for use in a polymeric prosthetic heart valve, and could also be used for
other applications requiring both flexibility and durability. Valve design is driven by the
choice of material, but this method allows design and material to be de-coupled, making
screening quicker and more efficient.
I collected the tensile and fatigue properties of the short-listed materials to generate a
material performance index. SIBS30 had the most favourable index, although it was not
possible to satisfactorily compression mould the material without lining the mould, which
would not be possible in a complex geometry. SEPS22 had the second most favourable
index among the saturated BCP, and could be moulded. Furthermore, well oriented,
saturated BCP have a superior performance index to a reference polyurethane material.
SEPS22 was selected as the most suitable material for use in a flexible leaflet poly-
meric prosthetic heart valve, based upon chemical stability, process-ability, flexibility,
and durability, and will thus be taken forward for a more detailed analysis in the rest of
this thesis. Another criterion for suitability is biocompatibility, which I will consider in
the next chapter.
86 Chapter 2. Polymer selection
Chapter 3
Biocompatibility
3.1 Introduction
This chapter discusses the issues concerned with the biocompatibility of materials that
could be used in a polymeric prosthetic heart valve. Biocompatibility testing is performed
on the BCP, SEPS22, which was selected in Chapter 2, and the other shortlisted BCPs.
Before the biocompatibility of a material is assessed, biocompatibility must be defined.
Biocompatibility is the biological response when a foreign material, as part of an in
or ex vivo device, comes into contact with a host. The response will develop over time,
becoming either clinically acceptable and tolerable to the patient, making it biocompatible,
or having some clinical effect making it intolerable (Williams, 2014).
Upon closer examination of this definition, several issues require clarification. First, the
property of biocompatibility is only applicable to a material-host system. For example,
PTFE is in widespread clinical use in a multitude of devices and has excellent biocom-
patibility under international (ISO and FDA) standards. However, when used in metal-
on-plastic hip implants, small fragments of PTFE can be abraded and lead to a massive
local inflammatory reaction (Charnley, 1963). The hydrophobic surface of PTFE makes
cell adhesion particularly difficult, rendering it useless in tissue engineering scaffolds and
bioreactors. As such, it is only appropriate to say that “PTFE within a cardiac delivery
catheter, for example, is a biocompatible system”.
Second, there is no linear spectrum of biocompatibility. The response of a biomaterial-
host system is a complex combination of many entangled pathways with positive and
negative feedback loops; there is no single measure which confers biocompatibility and
each response is transient (Williams, 2014). For example, PTFE grafts are used to
87
88 Chapter 3. Biocompatibility
bypass stenosed blood vessels. Upon implantation they display a minimal host response,
but over longer durations the grafts consistently become stenotic as a result of neointimal
hyperplasia (Kelly et al., 2002). As I seek to minimise clinically observable effects, I may
have to resort to the “least worst option”, for there is no known perfectly biocompatible
material (Williams, 2014).
In this chapter I will cover whether the “styrenic BCP-major artery system” may result
in clinically acceptable effects.
3.1.1 Host response pathways
The body’s response to any implanted material is rapid. Upon contact with blood,
proteins are immediately (≤1 s) adsorbed to the implanted material’s surface. The com-
position of this adsorbed layer will change over time as proteins may rearrange or de-
tach, whilst other proteins may competitively bind. This protein pseudo-surface is of
paramount importance as it interacts with cells and thus determines the long term re-
sponse to the implanted material (Anderson et al., 2004).
The host response to implanted materials is spatially and temporally specific. For ex-
ample, the surgical implantation of a prosthetic heart valve leaves a wound in the aortic
root. In order for this to heal there is a carefully mediated cascade of biological re-
sponses. To stem the flow of blood from a wound and provide a scaffold on which healing
can take place, a thrombus, or clot, must be formed rapidly. This is initiated either by
the activation of factor VII which is triggered by tissue damage (producing tissue factor),
or contact with foreign (non-endothelial) surfaces, activating factor VIII. Both of these
processes result in an amplifying cascade of protein changes, converging at the activation
of factor X. Factor Xa (activated factor X) converts prothrombin to thrombin, triggering
fibrinogen to polymerise into fibrin strands, this becoming associated to form a fibrin gel,
creating a surface upon which healing can take place. Although this process is entirely
physiological in the wound healing response, a foreign material can lead to catastrophic
perturbations, such as a stroke-causing embolus of a clot.
While the coagulation cascade is protein-based, whole cells are also rapidly activated.
Platelets follow an adsorption, activation, and aggregation cascade. The most abundant
platelet surface receptor is glycoprotein IIb/IIIa, which binds to von Willebrand Factor
(vWF), fibrinogen, fibronectin, vitronectin or thrombospondin, allowing adsorption to a
protein coated surface and platelet aggregation.
When platelets encounter collagen or vWF, or an absence of nitric oxide, endothelial-
Chapter 3. Biocompatibility 89
ADPase or prostacyclin, they are activated. This results in a shape change and the
formation of pseudopodia from their surface. They secrete granules and initiate the
arachidionic cycle which leads to the production of thromboxane A2, initiating a positive
feedback loop, causing a platelet plug to form (Napier, 2012).
Platelets’ response to shear stresses is dependent on both the magnitude of shear and
duration (Ramstack et al., 1979; Sheriff et al., 2010). While red blood cells are lysed at
about 600 Pa, platelets are activated by shear stresses as low as 10 Pa, which may be found
in pathological regions of the vasculature (Anderson et al., 1978). Short, intense bursts
of shear stress, of 255 Pa for as little as 7 ms, are sufficient to liberate β-thromboglobulin
(β-TG) as part of the clotting cascade (Wurzinger et al., 1985). The shear activation of
platelets occurs through the binding of glycoprotein Ib-IX-V complex on the surface of
the platelets to vWF, activating fibrinogen receptors and leading to platelet aggregation
(Andrews et al., 1997).
The formation of a platelet-fibrin mesh at a lesion or foreign body becomes a scaffold
upon which the acute inflammation process takes place. Damage to the vasculature
results in the release of cytokines causing the emigration of leukocytes to the implant and
perivascular. In the minutes and days following injury, neutrophils (polymorphonuclear
leukocytes) are the most abundant cell type present at the wound. Neutrophils are
shortlived at the implant site, apoptosing after 24 to 48 h, to be replaced by monocytes.
These inflammatory cells are able to differentiate into macrophages, which aim to engulf
the material so that it can be excreted. However, when foreign objects are greater than
1 µm in size this becomes “frustrated phagocytosis”. Initially the cell will attempt to
degrade the material surface, producing a hydrolytic and oxidative environment. If this
does not successfully break down the material, a continuous fibrotic capsule will be formed
which shields the host from the implant. To form this capsule, phagocytic cells fuse into
multinucleated giant cells which combine with extracellular matrix (ECM) components
and fibroblasts to surround the foreign body. This capsule may range in thickness from
tens to hundreds of microns, and generally reaches a quiescent phase within a month
(Ainslie et al., 2009; Anderson et al., 2004).
The innate immune system also responds by recruiting phagocytic cells to the wound
and activating the complement system. The complement cascade consists of three ac-
tivation routes: classical, lectin, and alternative, which converge with the formation of
complement protein C3 convertase. The alternative route commences with C3 and C3b
adsorbing to a material, which lead to the production of C3 convertase. C3 conver-
tase triggers the terminal pathway, leading to the production of serum complement 5b-9
(SC5b-9), the membrane attack complex, which forms pores in target cells, causing them
90 Chapter 3. Biocompatibility
to lyse (Andersson and Andersson, 2003). A biocompatible material-host system will
result in minimal activation of the complement cascade.
3.1.2 Host response to prosthetic heart valves
Prosthetic heart valve implants in current clinical use are passive, and the prostheses
are recognised as foreign and can disrupt the normal physiology. Upon implantation,
the leaflets of the valve are exposed to blood, causing proteins to be adsorbed, and
contributing to the activation of the coagulation cascade. Combined with the effects of
fluid shear stresses, the coagulation cascade may be activated and thrombi generated.
The only surface which is truly compatible with blood is a layer of endothelial cells. In
a classical implant which is implanted without endothelial cells, spontaneous, natural
endothelialisation of the implant must be relied upon to achieve this level compatibil-
ity. Unfortunately, trananastomotic growth is self limiting, and it is rare for endothe-
lial cell layers to extend more than 1 to 2 cm from an anastamosis on a static graft.
Glutaraldehyde-fixed biological tissues can have some residual cytotoxicity, reducing the
likelihood of endothelial cell proliferation. New methods of detoxification are able to
reduce inflammation and calcification, though have a minimal effect on thrombogenicity
(Remi et al., 2011). Micro thrombi are rapidly and persistently formed on the leaflet
before endothelialisation can take place. In vitro endothelialisation prior to implantation
can allow a confluent layer of endothelial cells to form, which can lead to a reduction in
thrombogenicity. However, the method is complex, and results in a minimal change in
calcification and inflammation relative to detoxification (Trantina-Yates et al., 2001).
Although there is an ongoing search for materials which promote complete endothelial-
isation, the prospects for successfully forming an endothelial shell on a prosthetic heart
valve are slim. The most likely candidates for endothelialisation should be autographic
transplants, which are performed in the Ross procedure (by grafting the pulmonary valve
to the aortic position). Unfortunately, explanted transplants had a thin fibrous sheath
formed around the transplanted valve, indicative of minor inflammation without endothe-
lialisation. The very act of implantation or transplantation seems to be enough to trigger
non-immune-related cell death (Hopkins, 2006)
Acute inflammation is inevitable in the minutes and days after implantation, and this may
be followed by a chronic inflammatory response. For a classical implant, it seems that
the best possible trajectory is the formation of a stable fibrous capsule which signifies the
end-stage of healing. Wound healing then requires the production of granulation tissue in
which fibroblasts proliferate, allowing the formation of a fibrous capsule. Alternatively,
Chapter 3. Biocompatibility 91
chronic inflammation can lead to the fusion of macrophages to form foreign body giant
cells which may prevail for the lifetime of the implant, though it is unknown whether
they remain activated (Anderson, 2001).
The growth of fibrotic tissue onto the leaflets of a bileaflet mechanical heart valve can
severely interfere with their function, and provides material for embolisation. Pyrolitic
carbon is used to coat the surface of mechanical heart valve leaflets due to its lack of
thrombogenicity and durability, but the material’s surface does not resist fibrous encap-
sulation. Mechanical heart valve designers have instead placed a small ridge on the leaflet
housing which halts tissue growth1.
On the one hand, fibrotic growth on the leaflets of a prosthesis may be beneficial as it
shields the cells from the foreign material. On the other hand, ingrowth of fibrous tissue
into the sewing ring is absolutely necessary. The prosthetic valve must form a tight seal
with the aortic root to avoid paravalvular leak and to secure the valve. Polyester knits
are the materials most commonly used for the sewing ring. Polyester is commonly used
for long term implantable materials; it can be found in septal occluders, hernia repair
meshes, wound dressings, breast supports and left ventricular assist devices (Weinberg
and King, 2004). The high surface area and porosity of fabrics generally promotes a
high ratio of macrophages and foreign body giant cells to fibrotic cells, and polyester
does invoke a mild, long-term inflammatory reaction (Anderson, 2001; Sigler and Jux,
2007). Nevertheless its widespread use can be explained by its integration with scar
tissue, relatively low inflammatory response and lack of biodegradation.
The chronic inflammation of permanent foreign bodies is highly dependent on anatomical
location, animal host for testing, and the time period of observation. So the results of
certain materials and designs cannot easily be translated. As such, in vitro and generic
in vivo testing systems for chronic inflammation are not often suitable, and chronic
inflammation will not be tested in this chapter.
3.1.3 Haemocompatibility testing systems
The results of haemocompatibility testing can be strongly influenced by the choice of
testing procedure. The test should simulate the conditions experienced by blood and
cells around prosthetic heart valves, with particular attention to shear stresses imparted
on the blood.
Static tests are generally unacceptable for the assessment of haemocompatibility. To
1Personal communication with Professor Richard Bianco, 2015.
92 Chapter 3. Biocompatibility
simulate realistic blood flow situations, systems use either centrifugation, planar flow
chambers, loop systems, circulation models or agitators to generate realistic shear stresses
(Braune et al., 2013). The system should be selected for reproducibility, sensitivity, prac-
ticality, and flexibility, whilst demonstrating accurate representation of the environment
in which blood flows.
I used the Chandler loop system which consists of loops that are 2/3 filled with mildly
anticoagulated blood, with the remaining 1/3 filled with air, as show in Figure 3.1. A
strip of test material is tightly placed in the loop such that it rotates with the tubing
without being dislodged by the blood. The loops are then rotated in a water bath,
causing physiological shear rates of blood over the material. The Chandler Loop method
has a high ratio of sample surface area to blood volume to maximise sensitivity, with the
sample surface area similar to that of a prosthetic heart valve, with a minimum amount
of blood from a single donor, so enabling repetition and multiple materials to be tested
(Seyfert et al., 2002).
The standard Chandler Loop setup has an air-blood interface in the loop which is nec-
essary for inducing flow, but also causes significant blood trauma and thus leads to an
increase in leukocyte activation, thrombi and protein denaturation. The Modified Chan-
dler Loop uses polyvinylchorlide (PVC) loops with a covalently bonded heparin coating
to reduce background activation (Sinn et al., 2011; Streller et al., 2003).
Figure 3.1: Chandler loop set up with six loops rotating in a water bath.
Chapter 3. Biocompatibility 93
3.1.4 Reference materials
The parameters associated with biocompatibility vary from donor to donor, and from day
to day for a single donor (Vienken, 2008). As such, it is impossible to predict the extent
of an in vivo response, from ex vivo data, without the use of reliable reference materials.
Low-Density Polyethylene and Polydimethylsiloxane (PDMS) (both gamma irradiation
sterilised) were proposed as Primary Reference Materials for biocompatibility testing in
1984 by the National Heart, Lung and Blood Institute (NHLBI). The materials initially
served as adequate reference materials for in vitro, ex vivo, and in vivo testing but the
gamma sterilisation resulted in some cytotoxicity. However, there was no full evaluation
of the materials at the time of implementation, and they have not received widespread
use in medical devices, nor as reference materials in testing (Belanger and Marois, 2001).
Without a universal standard, reference materials must have a well characterised in vivo
response for the appropriate host system. Samples must also be manipulated inside the
flexible PVC tube in the Chandler Loop system, so they should be thin and flexible.
Expanded polytetrafluoroethylene (ePTFE) and polyester (PES) are commonly used as
reference materials. ePTFE has good haemocompatibility so serves as a positive control,
with the proviso that it can promote stenosis narrow diameter vascular grafts. PES is
mildly thrombogenic and inflammatory, so serves as a negative control.
3.2 Materials and methods
3.2.1 Materials and sample preparation
I tested the proposed BCP, from Chapter 2, SEPS22, for biocompatibility using a Modi-
fied Chandler Loop. ePTFE and PES vascular grafts, manufactured by Jostent, were used
as reference materials. I also compared the selected polymer to a Bovine Pericardium
patch, fixed with glutaraldehyde (trade name Peri-Guard, with Apex Processing, Synovis
Life Technologies, Inc. MN, USA). Three other shortlisted BCPs: SIS30 and SI-BS19
(both Kraton), and SIBS30 (Innovia LLC), were previously tested by Prof. Hans Wen-
del’s group (Klinisches Forschungslabor der Klinik fu¨r Thorax-, Herz- und Gefa¨ßchirurgie,
Universita¨t Tu¨bingen) using the same system.
I compression moulded the BCP samples at 160 ◦C in an electrically heated hydraulic
press to a thickness of 0.3 mm. Melt flow during moulding was minimal, so the samples
were not mechanically anisotropic. Strips of 9× 150 mm were cut from the polymer sheets
94 Chapter 3. Biocompatibility
and reference materials for tests in the Modified Chandler Loop. Samples were sterilized
using 70% ethanol before testing.
3.2.2 Coating of materials
In the event that a material does not appear to be sufficiently biocompatible it may
be possible to coat the material, or modify the surface, to mediate the host response.
Certain synthetic coatings seem to be able to reduce fibrinogen adsorption and thus
cellular adsorption significantly, such as polyethylene glycol, zwitterionic groups includ-
ing phosphorylcholine, sulfobetaine, carboxybetaine, and polycarboxybetaine (Vladkova,
2013). Alternatively, biological molecules may be grafted to the surface to have an active
influence. Surface heparinisation is one of the most promising methods for improving
the blood compatibility of polymeric materials. Heparin is a highly sulfated, anionic
blood polysaccharide (5–25kDa) that naturally occurs on the intravascular endothelium.
Heparin binds to anti-thrombin anionically, accelerating the rate at which anti-thrombin
inactivates thrombin.
I coated one set of SEPS22 samples with the Corline Heparin (CH) Surface. The CH
coating consists of the covalent binding of macromolecular heparin which conjugates to
an inert polyamine chain. End-on attachment of heparin, with a spacer arm, is the
most effective way to retain the activity of heparin when surface bound (Golander and
Larsson, 1986). The substrate polymer was primed for coating by forming a cationic
surface on the uncharged synthetic material. This was then heparin coated by exposing
to a dilute water solution containing CH conjugate, manufactured by Corline® (Uppsala,
Sweden). CH conjugate binds to any cationic surface with very a strong affinity due to
the high concentration of anionic groups (Kidane et al., 2004; Wendel and Ziemer, 1999).
70 moles heparin are irreversibly bound to each mole of the surface-bound carrier chain.
The coated surface has a concentration of heparin of 0.5 to 1 µg cm−2, which has the
capacity to bind 2 to 4 pmol cm−2 of antithrombin (Christensen et al., 2001).
3.2.3 Donors
Blood was drawn from healthy volunteers aged between 20 to 50 years. Donors were
chosen with exclusion criteria eliminating smokers, drug-taking in the last 2 weeks, es-
pecially aspirin, oral contraceptives, and non-steroidal antiinflammatories. Blood was
carefully drawn without stasis by venipuncture with butterflies (1.4 mm) directly into
sterile pre-anticoagulated containers (1.5 IU/ml Heparin-Natrium 25000, Rathiopharm
Chapter 3. Biocompatibility 95
GmbH, Ulm, Germany). Unless otherwise stated, six, screened, consenting donors were
used for each of these tests.
3.2.4 Chandler Loop method
Samples of material measuring 9× 150 mm were inserted into heparin coated PVC tubes
(50 cm length) (Carmeda coating, Medtronic), and the ends of the tubes were joined with
silicone tubing to form continuous loops. 20 ml of blood was added to each tube. The
loops were rotated vertically at 30 rpm in the water bath maintained at 37 ◦C. After
90 min of circulation the blood was collected in appropriate syringes for analysis as fol-
lows: 2.7 ml with 1.6 mg ml−1 potassium-EDTA added for complement and blood cells
analysis; 10 ml citrated blood (0.14 ml of 0.106M C6H5Na3O7 · 2 H2O citrate) for acute
inflammation, coagulation cascade activation and haemolysis analysis; 4.5 ml anticoag-
ulated blood (450 µl of 0.109M; citrate-theophylline-adenosine-dipyridamole (CTAD),
Becton Dickinson GmbH, Heidelberg, Germany) for β-TG analysis. The samples were
centrifuged immediately at 2000 G for 20 min with a cryofuge (Model 8000, Heraeus, Os-
terode, Germany) to separate plasma samples. These were then aliquoted into 400 µl
samples, shock-frozen in liquid nitrogen and stored at −70 ◦C.
3.2.5 Parameters for analysis
According to ISO 10993:4, blood-contacting materials should be evaluated for thrombo-
genicity, platelet activation, coagulation, protein adsorption and inflammatory markers.
Thrombogenicity
Scanning electron microscopy (SEM) was used to observe platelet adhesion and aggre-
gation, leukocyte adhesion, and fibrinogen adsorption on the surface of the material.
After circulation in the Chandler loop, the polymer samples were incubated overnight
in 2% glutaraldehyde (Serva, Heidelberg, Germany) containing PBS (phosphate buffered
saline, Invitrogen Gibco, Karlsruhe, Germany) solution and subsequently rinsed in pure
PBS. The remaining water was removed from the samples using 40 to 100% of ethanol
(Merck, Darmstadt, Germany) in ascending concentrations. Finally all samples were crit-
ical point dried and sputtered with gold palladium before analysis with SEM (Cambridge
Instruments, Cambridge, UK, type 250 MK2).
96 Chapter 3. Biocompatibility
Platelet activation
The activation of platelets leads to the release of β-TG. β-TG concentration was measured
using an ELISA (Diagnostica Stago Asnieres, France).
Coagulation cascade
The formation of thrombin causes fibrinogen to cleave into fibrin monomers. The concen-
tration of thrombin-antithrombin complex was measured by ELISA (Siemens Healthcare
Diagnostics Products GmbH, Marburg, Germany).
The adsorption of factor XII and fibrinogen to a surface leads to their conformational
change and the activation of platelets. Platelets adsorb via their GPIIb/IIIa receptor,
leading to their activation. The concentration of GPIIb/IIIa (also known as CD41 anti-
gen) was measured using an in house ELISA method, as devised by Paul et al. (2009).
Inflammation
Local inflammation is characterised by the activation of neutrophils and phagocytes seek-
ing to expel a pathogen from the body. Neutrophils release the proteinase PMN-elastase
(Erlenko¨tter et al., 2008). The concentration of PMN-Elastase was measured by ELISA
(Milenia Biotec GmbH, Giessen, Germany).
Protein adsorption
The adsorption of proteins to the material surface determines the host response. The
blood contacted materials are placed in EDTA to preserve the state of the adsorbed
plasma proteins (Engberg et al., 2011). The concentration of fibrinogen and CD41 were
measured using in-house methods
Blood cell count
Red and white blood cell and platelet counts were performed on EDTA-blood (potassium-
EDTA, 1.6 mg ml−1) immediately after sampling using a fully automated cell counter
system (micros 60 ABX Hematology, Montpellier, France). Haemolysis was detected
using a colorimetric assay for free plasma haemoglobin (Cyan haemoglobin test, UKT,
Germany).
Chapter 3. Biocompatibility 97
Complement system
The activation of the complement cascade leads to the release of serum complement
protein 5b-9 (SC5b-9), whose concentration was measured with an ELISA (Quidel, San
Diego, CA, USA).
3.2.6 Coating durability
To assess the durability of the heparin coating I aged heparin coated samples in vitro
in three ways: static in PBS solution, dynamically strained in PBS solution, and in a
3% hydrogen peroxide solution which simulates the oxidising conditions which may be
experienced during an inflammatory reaction. Coated and uncoated control samples were
also used. All samples were kept at 37 ◦C for 3 weeks.
Heparin is the most negatively charged molecule found in nature, so a heparin coated
sample should have a lower surface energy. This change in surface energy might be
measured by investigating the water contact angle on the polymer surface. The contact
angle which a drop of water makes with the sample is a measure of the hydrophobicity of
the material. I performed a preliminary experiment to investigate whether there was a
difference in contact angle between coated and uncoated samples. On uncoated samples
of SEPS22 I found the contact angle to be 92.0° compared to 78.6° on heparin coated
specimens. This difference was significant (p≤ 0.001). Water contact angle on the samples
was recorded at seven spots.
Although the presence of heparin may be inferred from the water contact angle on the
surface, this gives no indication of the presence of a functional group. Kristensen et al.
(2006) used X-ray photoelectron spectroscopy (XPS), and Harada et al. (2005) used
Fourier Transform Infra Red spectroscopy with an attenuated total reflectance sample
mount (FTIR-ATR) to evaluate surface concentrations of heparin.
In XPS, an X-ray is fired at the surface leading to the emission of outer shell photoelec-
trons from surface molecules. A specific fingerprint of photoelectron energies can be used
to identify and measure the concentration of specific groups which are found in the outer
10 nm of the surface. I sent samples for XPS analysis at the Nexus facility at Newcastle
University. XPS was performed using Kratos Axis Nova XPS system with an Al Kα
(1486.6 eV) source. Five spots upon three samples of each material were analysed.
In FTIR, infra red light is shone on to a material at a range of frequencies. Each chemical
bond in the material absorbs radiation at a characteristic frequency. The reflected or
98 Chapter 3. Biocompatibility
transmitted light is detected which is analysed by a fourier transform, to derive the
absorption at a range of frequencies. I performed FTIR using an ATR sample mount,
unlike Harada et al. (2005), I was not able to consistently detect changes in the relevant
absorption peaks on the polymeric substrates as they appeared to be masked by a peak
from COH groups, so this method was not used.
The concentration of heparin which leached from the coating into the PBS solution was
determined using the toluidine blue method (van Bilsen et al., 2008). Polymer samples
(65 cm2) were stored in glass vials containing 20 ml of PBS, pH 7.4 at room temperature.
Periodically, 2 ml of the PBS solution was taken and mixed with a 0.005% toluidine blue
solution. Following heparin-dye-precipitation, the absorbance of the dye-depleted solution
was measured at 631 nm and the concentration of the released heparin was calculated from
the heparin calibration curve, as constructed using sodium heparin salt from intestinal
mucosal source (Sigma, UK).
3.2.7 Effect of anticoagulant drugs
The preferable scenario would be for the in vitro haemocompatibility study of a material
to mimic the expected in vivo scenario as closely as possible. This suggests that antico-
agulant therapies should be implemented into the testing. However, Braune et al. (2013)
noted that no internationally accepted guidelines are defined for the concentrations of
anticoagulant to be used in vitro. Salicylic acid (Bayer, Germany) was added in vitro to
yield a concentration of 20 mg l−1.
3.2.8 Analysis
For the post-processing of data I used MATLAB (Mathworks, MA, USA). Results were
plotted as histograms with error bars representing the mean ± standard error. A two
tail ANOVA with the Bonferroni correction was used to compare normally distributed
samples of equal size., significant comparisons with p≤ 0.05, 0.01, 0.001 are indicated by
marking with *, **, and ***, respectively.
3.3 Results
In the polymer selection of Chapter 2, a set of BCPs were tested by our collaborators
for their haemocompatibility; these materials constituted “run 1”. Having selected a
Chapter 3. Biocompatibility 99
material, and collected preliminary observations on the relative biocompatibility of the
polymers, I then assessed the haemocompatibility (at Prof. Wendel’s lab) of the selected
polymer, SEPS22. This was compared to fixed bovine pericardium, as an additional
reference material due to its extensive use in bioprostheses. I also tested SEPS22 with a
heparin coating. These materials are “run 2”.
3.3.1 Thrombogenicity
SEM images of the polymeric materials are shown in Figure 3.2. No large thrombi were
found on the surface of the polymeric materials. Morphological differences were observed
between the proteins adsorbed on the surface of each polymer. Addition of the heparin
coating resulted in a very different morphology of the protein layer. A subconfluent
protein layer is adsorbed onto the uncoated materials. The protein layer is fibrous with
some cellular entrapment and joining of fibres. In contrast, the heparin coated samples
resulted in the formation of a weaker fibrous mesh, suggesting a lower level of protein
adsorption. Holes in the protein layer were observed on some uncoated samples, which
may have arisen due to the detachment of adsorbed cells during the drying process.
SEM images of the ePTFE (not shown) also had a subconfluent protein layer, which
did not have the fibrous morphology as seen on the BCP materials. Although ePTFE is
porous, no ingress of proteins or cells could be detected by SEM. On the polyester sample
(not shown) there was significant adsorption of proteins and red blood cells, though no
large thrombus formation. While no blood was adsorbed to the surface of the polymeric
samples after rinsing in PBS, the bovine pericardium samples retained significantly more
blood, as shown in Figure 3.3. Under SEM, the sample displayed significant protein
adsorption and thrombus formation.
3.3.2 Hemolysis and coagulation
I did not find any significant differences between materials for red blood cell counts of
blood discharged from the loops after testing. Free haemoglobin concentration is a marker
of red blood cell destruction, which may occur if a material is highly incompatible with
human blood; all BCPs performed at least as well as the reference materials, with SI-
BS19, SIBS30, and uncoated SEPS22 resulting in less haemolysis than the reference
materials (Figure 3.4a).
The generation of thrombin occurs in the common pathway of the coagulation cascade.
The concentration of thrombin-antithrombin complex (TAT) in blood was measured be-
100 Chapter 3. Biocompatibility
No blood contact After 90 minutes blood contact
(a)
SEPS22,
un-
coated
(b)
SEPS22,
heparin
coated
Figure 3.2: SEM images of polymer surfaces before and after contact with blood (scale bar
is 10µm on all images). The coated and uncoated surfaces are macro- and microscopically
smooth, but have a protein coating after blood contact. The uncoated sample (a) appears
to have activated platelets adsorbed to its surface. The coated surface (b) has a confluent
layer of fibrin, but there are no activated platelets on the surface.
fore and after contact with the materials (Figure 3.4b). In both runs the fresh blood
and control tube were found to contain an identical level of thrombin, indicating that the
heparin coating of the PVC tubing was effective at reducing background activation. All
reference materials resulted in activation of the coagulation cascade. The uncoated BCP
were more activating than ePTFE, itself more activating than bovine pericardium, but
less activating than polyester. In the second run, SEPS22 was also more activating that
bovine pericardium. The addition of a heparin coating to SEPS22 resulted in the least
activation of the coagulation cascade. However, differences between materials were not
statistically significant (p > 0.05).
Chapter 3. Biocompatibility 101
(a) Photograph
(b) SEM
Figure 3.3: (a) Bovine pericardium sample after blood contact showing adherence of
blood to the surface of the material. Under SEM (b), where the scale bar is 1 mm and
100µm in the left and right images respectively, there is a mild thrombus with a fibrin
mesh with entrapped erythrocytes.
0
5
10
15
20
25
30
35
Fre
sh
 bl
oo
d (i
)
SIB
S 3
0
SIS
 30
SI−
BS
 19
eP
TF
E (i
)
Po
lye
ste
r
Co
ntr
ol t
ub
e (i
)
Fre
sh
 Bl
oo
d (i
i)
SE
PS
 22
SE
PS
 22
 + 
He
pa
rin
Bo
vin
e P
eri
ca
rdi
um
eP
TF
E (i
i)
Co
ntr
ol t
ub
e (i
i)
H
em
og
lo
bi
n 
co
nc
en
tra
tio
n 
(m
g/d
L) Run 1 Run 2
(a) (b)
Figure 3.4: Results of measurements on 13 materials as tested in two separate runs. (a)
Free haemoglobin is released during haemolysis and was measured using a photometrical
test. (b) thrombin is generated in the common pathway of the coagulation cascade as
measured by the concentration of thrombin-antithrombin complex.
102 Chapter 3. Biocompatibility
3.3.3 Platelets
I found that the BCP and reference materials led to activation of platelets. Platelets
may be adsorbed to the material surface, aggregate to each other, or become activated,
releasing α-granules. The reduction in platelet count (Figure 3.5a) indicates that platelets
were adsorbed or activated.
When platelets come into contact with a foreign surface they can be activated. Activation
proceeds through the formation of pseudopodia, adhesion, aggregation and release of
platelet factors from the granules. All uncoated materials caused a drop in platelet
count in the discharged blood (Figure 3.5a) either through activation or adsorption of
the platelets to the material. BCPs, ePTFE and bovine perdicardium showed similar
platelet counts, whereas polyester resulted in a significantly greater reduction in platelet
numbers. Activation of platelets leads to the release of β-TG (Figure 3.5b), which was
analysed by ELISA. In the first run, β-TG was significantly greater on the BCP than
on the haemocompatible control, ePTFE. In run 2, there were no significant differences
between the BCPs and reference materials. The addition of a Heparin coating to SEPS22
led to lower β-TG levels (Figure 3.5b).
Adsorption of platelets to the material surface can result in the deposition of CD41
receptor protein on the material surface (Figure 3.5c). I found that the BCP materials had
significantly less adsorbed CD41 than the negative control, polyester. Heparin coating of
SEPS22 also reduced CD41 adsorption. Platelets and leukocytes can bind to fibrinogen
which is adsorbed to the material’s surface. High levels of adsorbed fibrinogen lead
to thrombus generation and localised inflammation (Hong et al., 2001). I found that
fibrinogen adsorption was significantly lower on the BCPs than on the control materials
(Figure 3.5d).
3.3.4 Innate immune response
Foreign materials can active the innate immune system. I measured the white blood cell
count, concentration of PMN-Elastase and SC5b-9 before and after circulation in the
Chandler loop.
The BCP materials did not have a significant effect on white blood cell count. There was,
however, a significant reduction (p≤0.05) in white blood cell count for samples contacted
with polyester and bovine pericardium (Figure 3.6a). The activation of the immune
system may be inferred from the release of PMN-Elastase as neutrophils degranulate;
bovine pericardium and polyester resulted in greater releases of PMN-Elastase than the
Chapter 3. Biocompatibility 103
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
Fre
sh
 bl
oo
d (i
)
SIB
S 3
0
SIS
 30
SI−
BS
 19
eP
TF
E (i
)
Po
lye
ste
r
Co
ntr
ol t
ub
e (i
)
Fre
sh
 Bl
oo
d (i
i)
SE
PS
 22
SE
PS
 22
 + 
He
pa
rin
Bo
vin
e P
eri
ca
rdi
um
eP
TF
E (i
i)
Co
ntr
ol t
ub
e (i
i)
Fr
ac
tio
n 
of
 p
la
te
le
ts
 re
la
tiv
e 
to
 fr
es
h 
bl
oo
d
**
***
*****
***
*
***
**
***
***
***
**
***
***
***
***
***
***
***
***
***
(a)
100
101
102
103
104
105
Fre
sh
 bl
oo
d (i
)
SIB
S 3
0
SIS
 30
SI−
BS
 19
eP
TF
E (i
)
Po
lye
ste
r
Co
ntr
ol t
ub
e (i
)
Fre
sh
 Bl
oo
d (i
i)
SE
PS
 22
SE
PS
 22
 + 
He
pa
rin
Bo
vin
e P
eri
ca
rdi
um
eP
TF
E (i
i)
Co
ntr
ol t
ub
e (i
i)
β−
Th
ro
m
bo
gl
ob
ul
in
 c
on
ce
nt
ra
tio
n 
(IU
/m
L)
Run 1
Run 2
** ***
*****
**
*
**
*** **
*
(b)
0
0.5
1
1.5
2
2.5
SIB
S 3
0
SIS
 30
SI−
BS
 19
eP
TF
E (i
)
Po
lye
ste
r
SE
PS
 22
SE
PS
 22
 + 
He
pa
rin
Bo
vin
e P
eri
ca
rdi
um
eP
TF
E (i
i)
Ad
so
rb
ed
 C
D4
1 
co
nc
en
tra
tio
n 
O
.D
.
Run 1
Run 2
**
***
**
**
***
***
***
***
***
(c)
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
SIB
S 3
0
SIS
 30
SI−
BS
 19
eP
TF
E (i
)
Po
lye
ste
r
SE
PS
 22
SE
PS
 22
 + 
He
pa
rin
Bo
vin
e P
eri
ca
rdi
um
eP
TF
E (i
i)
Ad
so
rb
ed
 F
ib
rin
og
en
 c
on
ce
nt
ra
tio
n 
O
.D
.
Run 1 Run 2
*
**
***
***
***
***
***
(d)
Figure 3.5: Results of measurements on 13 materials as tested in two separate runs. (a)
Platelet count before and after contact with each of the materials. (b) shows the concen-
tration of β-TG. The presence of β-TG indicates that platelets have been activated. (c)
shows the surface concentration of CD41. The adsorption of platelets to the materials
can quantified by measuring the surface concentration of CD41 receptor. (d) The sur-
face concentration of fibrinogen is shown. Adsorbed fibrinogen is used by platelets and
leukocytes to bind to a material surface.
tested BCPs (Figure 3.6b). The coating of SEPS22 with heparin had no effect on either
of these inflammatory measures. Plasma protein SC5b-9 is used to lyse pathogenic cells
in the final stage of the complement cascade, where the alternative and classical pathways
104 Chapter 3. Biocompatibility
converge (Hastings, 1992), and is a marker of the degree of inflammatory response which
may be caused by a material. All materials led to some activation of the complement
cascade, though none as much as bovine pericardium (p<0.001) (Figure 3.6c).
3.3.5 Cell viability
Indirect contact cytoxicity tests were previously performed on shortlisted materials by
our collaborators (Prof. M. Costantino, Politecnico di Milano). They found no significant
cytoxicity associated with leaching from the materials. The proposed use of these BCPs
is in a permanent implant with direct contact with fibroblast and smooth muscle cells. In
accordance with ISO 10993-1, I tested cell viability in direct contact with these materials.
I used the standardised L929 fibroblast strain and measured cell count after 48 hours with
an MTS assay. The final cell numbers are shown in Figure 3.7.
The rapid profileration of murine fibroblasts on the surface of the materials implies that
cells are viable on the material. I did not find any significant differences between prolif-
eration on compression moulded (SEPS22), solvent cast (SEPS22 SC), or heparin coated
(SEPS22 H) surfaces of SEPS22 (P ≥ 0.15). Only the uncoated SEPS22 led to signif-
icantly more proliferation than the PS control well (P = 0.034). Differences in man-
ufacturing which may lead to variations in surface texture did not affect the extent of
proliferation of cells upon the surface; furthermore the solvents used during casting were
not cytotoxic. All BCP processing resulted in relatively smooth surfaces. In contrast,
the results for ePTFE and PES were for venous graft material which had ridges of ap-
proximately 1 mm which was likely to have reduced the rate and extent of proliferation.
Proliferation was significantly lower on PES than in the PS well (P = 0.000025). I did
not find cells to be viable on the bovine pericardium surfaces, which is possibly due to
residual leaching of the glutaraldehyde used for fixation.
Proliferation of fibroblasts should ensure good colonisation around the sewing ring of a
prosthetic heart valve, and possibly aid the fibrous encapsulation of the leaflets which
may reduce the rate of calcification. The cell culture used in these experiments was static,
which means that predictions for the growth of fibroblasts on a material implanted in the
structurally and fluidly dynamic aortic root are uncertain. Furthermore, predictions of
chronic inflammation cannot be made without assessing the response of directly contacted
macrophages.
Chapter 3. Biocompatibility 105
0
0.5
1
1.5
Fre
sh
 bl
oo
d (i
)
SIB
S 3
0
SIS
 30
SI−
BS
 19
eP
TF
E (i
)
Po
lye
ste
r
Co
ntr
ol t
ub
e (i
)
Fre
sh
 Bl
oo
d (i
i)
SE
PS
 22
SE
PS
 22
 + 
He
pa
rin
Bo
vin
e P
eri
ca
rdi
um
eP
TF
E (i
i)
Co
ntr
ol t
ub
e (i
i)F
ra
ct
io
n 
of
 w
hi
te
 b
lo
od
 c
el
ls 
re
la
tiv
e 
to
 fr
es
h 
bl
oo
d
Run 1 Run 2
*
**
*
*
*
**
(a)
100
101
102
103
104
105
Fre
sh
 bl
oo
d (i
)
SIB
S 3
0
SIS
 30
SI−
BS
 19
eP
TF
E (i
)
Po
lye
ste
r
Co
ntr
ol t
ub
e (i
)
Fre
sh
 Bl
oo
d (i
i)
SE
PS
 22
SE
PS
 22
 + 
He
pa
rin
Bo
vin
e P
eri
ca
rdi
um
eP
TF
E (i
i)
Co
ntr
ol t
ub
e (i
i)
PM
N
−E
la
st
as
e 
co
nc
en
tra
tio
n 
ng
/m
L
Run 1 Run 2
*
**
***
***
***
*** ** *
***
***
*
**
***
***
(b)
100
101
102
103
104
105
Fre
sh
 bl
oo
d (i
)
SIB
S 3
0
SIS
 30
SI−
BS
 19
eP
TF
E (i
)
Po
lye
ste
r
Co
ntr
ol t
ub
e (i
)
Fre
sh
 Bl
oo
d (i
i)
SE
PS
 22
SE
PS
 22
 + 
He
pa
rin
Bo
vin
e P
eri
ca
rdi
um
eP
TF
E (i
i)
Co
ntr
ol t
ub
e (i
i)
SC
5−
b9
 c
on
ce
nt
ra
tio
n 
ng
/m
L
Run 1
Run 2
*
*
***
***
*** ***
***
(c)
Figure 3.6: Activation of leukocytes occurs during inflammation and results in the release
of PMN elastase and loss of white blood cells. Polyester and bovine pericardium both
led to a significant reduction in leukocyte counts. (a) None of the BCPs led to a fall in
leukocyte numbers. (b) The concentration of PMN elastase was analysed by ELISA, all
materials led to release, though all BCP were significantly less inflammatory than both
polyester and bovine pericardium. (c) SC5b-9 is associated with the final stage of the
complement cascade and leads to the formation of membrane attack complexes that are
used to perforate the membrane of a pathogen or cell. All materials led to the activation
of the complement cascade, though all polymers were significantly less inflammatory than
bovine pericardium.
106 Chapter 3. Biocompatibility
SE
PS
22
SE
PS
22
SC
SE
PS
22
H
P2
36
3
BP
Et
O
H
PS
we
ll
PE
PT
FE
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
·104
N
u
m
b
er
of
ce
ll
s
Figure 3.7: Cell numbers on material that were seeded with 5× 103 cells, and cultured
for 48 h. Cell density was measured using an MTS assay. Error bars indicated standard
deviation from the mean (n=7). PS refers to polystyrene.
3.3.6 Coating durability
I aged the heparin coated samples for 3 weeks in air, PBS, and oxidising solutions. The
amount of heparin on the surface can be detected by XPS which quantitatively detects
the presence of sulphur in −OSO3 and −NSO3 groups of active heparin (Figure 3.8b).
The heparin coating also influences the hydrophobicity of the material, which I measured
by contact angle. The XPS and water contact angle results are shown in Figure 3.8.
The heparin coated samples which were strained dynamically have a significantly greater
contact angle than those that were held statically (P = 0.0029) and a lower sulphate
group concentration, which suggests that the coating may have been affected by cyclic
strain. Alternatively, strain may alter the surface morphology, which then might affect
the interaction with the water droplet. The contact angle of samples which were held in
oxidising solution was equal to that of the control samples (no coating), and significantly
greater than the coated samples which were stored in buffer solution. This indicates that
the coating was degraded by oxidising species.
I found a reduction in surface sulphur concentration (Figure 3.8c) for both the dynami-
cally strained and the static samples, though the difference between them was not statis-
tically significant. There was a large reduction in surface concentration of sulphur for the
samples which were statically stored in oxidising solution. The least amount of heparin
Chapter 3. Biocompatibility 107
was present on samples exposed to oxidising solution. Contrary to the difference in con-
tact angle between the statically and dynamically aged samples, the average area of their
S 2p spectra were very similar. Concerning the samples from PBS solution, phosphate
spectra were also detected, suggesting that the heparin coating became partially covered
by a layer of phosphate salts. This could also have contributed to a slight decrease in the
sulphur peak from heparin.
Figure 3.8d shows that the percentage of released heparin increased sharply during the
first day of the immersion and then the release curve plateaued at around 8% of the
adsorbed heparin. The observed heparin release was probably due to physically adsorbed
heparin on the polymer surface, which diffused into the PBS solution during the first
day of the experiment. As the heparin coated samples were thoroughly washed during
sterilisation, and the adsorbed protein profiles are significantly different, it is unlikely
that the biocompatibility improvements are a result of heparin leaching.
3.4 Discussion
In this chapter I used the Chandler Loop model to characterise the ex vivo response of
human blood to implantable materials. The Chandler Loop model is an established means
of simulating thrombus generation, as observed in vivo (Gardner, 1974). Furthermore,
the average shear rates generated in the Chandler Loop can be estimated for comparison
to the physiological environment (Touma et al., 2014). The selected loop dimensions
and rotation rate create shear rates which are comparable to those found in a major
artery, between 50 and 300 s−1(Hathcock, 2006). However, the continuous rotation of the
loop does not produce the transient flow seen at the aortic root. The Modified Chandler
Loop was successful at eliminating background activation of the coagulation cascade;
the control loop (containing no sample) was not significantly different to fresh blood for
coagulation parameters.
Day-to-day and donor-to-donor variation is a significant consideration in biocompatibil-
ity testing (Seyfert et al., 2002). As such, absolute analyte levels for the test materials
should be compared to the reference material results. PTFE, polyester, and bovine peri-
cardium are used extensively in in vivo and ex vivo vascular implants and are appropriate
choices of reference materials as the host response is well characterised. ePTFE is an in-
ert biomaterial with good haemocompatibility (ASTM International 2009). ePTFE is
characterised by its highly hydrophobic surface and lack of functional groups, but its
major performance draw back as a graft material is its restenosis in small vessels. As
108 Chapter 3. Biocompatibility
Un
co
at
ed
St
at
ic
PB
S
D
yn
am
ic
PB
S
O
xi
di
sin
g
Ai
r
0
10
20
30
40
50
60
70
80
90
100
C
on
ta
ct
an
gl
e
(
℃
)
(a)
(b)
(c)
(d)
Figure 3.8: Analysis of heparin coating after ageing of samples under various conditions.
(a) shows the water contact angle on SEPS22 polymer coated with Corline Heparin
Surface after ageing. At least 8 readings were taken for each sample and the error bars
represent the standard deviation of samples. (b) shows the XPS measurements of the
surface of heparin coated polymers with a peak at 169.5 eV which can be attributed to
the −OSO3 and −NSO3 groups in heparin. (c) shows the area of the XPS sulfur peaks
for the test samples stored in air, PBS, and H2O2, and dynamically strained in PBS. (d)
heparin concentration leached from coated samples incubated in PBS for 11 days. The
XPS and heparin leaching graphs were prepared by Dr J. Stasiak.
a prosthetic heart valve material it has suffered from excessive calcification and poor
durability (Bezuidenhout and Zilla, 2014). The use of polyester (Dacron) in heart valve
applications is limited to the sewing ring where it is rapidly encapsulated in fibrous tissue.
Chapter 3. Biocompatibility 109
In applications of continuous contact with blood the material has resulted in the failure
of polymeric prosthetic heart valves tested in vivo (Gallocher, 2007).
The experiments performed here on bovine pericardium in the Chandler Loop are the
first to be reported in the literature. The pericardium had minimal thrombogenicity
and activation of the coagulation cascade, and these are reflective of its in vivo perfor-
mance; pericardial leaflets on prosthetic heart valves do not require major anticoagulation
therapies, though patients are generally prescribed aspirin. Among all materials, the in-
flammatory measurements (leukocyte count, PMN elastase, SC5b-9) were greatest for
blood contacted with pericardium, which might be expected from a xenographic trans-
plant. Improved pericardium processing methods, with alternatives to glutaraldehyde
and post-fixation washing, could improve cell viability and reduce inflammation (Remi
et al., 2011).
The BCP materials did not cause significant haemolysis or activation of the coagulation
cascade relative to ePTFE. Activation of platelets can be inferred from a release of β-
TG, whilst the adsorption of CD41 to a surface is an artefact of platelet adsorption
to the surface. Platelet activation is lowest for ePTFE, but is closely followed by the
test BCPs. High levels of adsorbed CD41 might not correlate with SEM images of
adsorbed platelets because they could have been washed off due to flow during the test
(Weber et al., 2002). Polyester resulted in significantly greater destruction of platelets and
adsorption of CD41. The test BCPs were significantly less activating and adsorbing than
polyester. Contact with ePTFE resulted in significantly less β-TG release than contact
with BCP. However, adsorbed CD41 levels were not significantly different between BCP
and ePTFE, possibly due to the hydrophobic surface of both materials. The addition of
heparin coating mediated the activation and adsorption response, resulting in the lowest
platelet activation amongst test materials.
The absence of hydroxyl and amine groups in the shortlisted biomaterials is a useful prop-
erty in minimising the activation of the (alternative) complement cascade (Chenoweth,
1987). The protein adsorption profiles for the BCPs are similar to ePTFE and bovine
pericardium, so the long term biocompatibility performance of the polymers should also
be similar. As these protein adsorption levels are taken at physiologically relevant flow
rates, similar levels of cell adhesion may also be possible in vivo. The adsorption of pro-
teins is indicative of the long term performance of a biomaterial, in particular, elevated
fibrinogen adsorption aids the adhesion of platelets and macrophages which can lead to
a chronic, local, inflammatory reaction. SEPS22 and SI-BS19 resulted in significantly
lower protein adsorption than occurred on ePTFE. Heparin coated samples resulted in
a minimum of CD41 and fibrinogen adsorption suggesting that as long as the coating is
110 Chapter 3. Biocompatibility
present and active, superior biocompatibility may be achieved.
The strength of adsorption of proteins to the surface may indicate the ease with which
inflammatory cells bind to the materials. In run 1 the BCP and ePTFE had lower adsorp-
tion of CD41 and fibrinogen than polyester, supporting the claim that these materials
are less inflammatory. Interestingly, in run 2, SEPS22, Heparin coated SEPS22, and
pericardium were less adsorbent than ePTFE, suggesting that the long term binding of
inflammatory cells to ePTFE is greater than to other materials. While ePTFE certainly
has acute haemocompatibility, the long term performance of ePTFE grafts in narrow
vessels is poor due to neointimal hyperplasia (Kelly et al., 2002). The binding of inflam-
matory cells can be predictive of the long term performance of implanted materials, but
long term predictions of inflammatory behaviour are not advisable from such short term
studies.
The surfaces of the BCPs are dominated by −CH3 groups. Sperling et al. (2005) pro-
duced alkanethiols with uniform monolayers containing either −CH3, −COOH or −OH.
Platelets adhered to the −CH3 surface, but not to the −OH surface, whereas leuko-
cytes behaved inversely. The −OH was the only surface to induce significant complement
cascade activation, explaining the minimal complement activation observed on the test
BCPs (Chenoweth, 1987; Engberg et al., 2011).
Between the various BCP materials there were few differences in any of the haemocom-
patibility parameters. This may be attributed to the similarity of the material’s surface
chemistry. The soft block (polyethylene, polyisoprene, or polyisobutylene) is expected
to coat the surface and the minor differences in molecular architecture do not affect the
surface properties (surface energy) significantly.
The polymeric materials were not cytotoxic. In the heart valve scenario, the proliferation
of fibroblasts should ensure good colonisation around the sewing ring, and possibly aid
the fibrous encapsulation of the leaflets which may reduce the rate of calcification. In situ
endothelialisation from anastomoses over distances greater than 10 mm is unlikely. Fur-
ther work should evaluate the response of endothelial and monocyte-macrophage cultures
to surface contact.
The addition of a heparin coating is well known for improving haemocompatibility by
reducing the hydrophobicity of polymer materials (Goosen and Sefton, 1979), as well
as having a biological interaction with plasma proteins (Amiji and Park, 1993). The
reduction in protein adsorption, as measured by assay and observed under SEM, helps to
reduce coagulation and inflammatory activation, resulting in the most haemocompatible
material in these tests.
Chapter 3. Biocompatibility 111
These experiments were limited to 90 minutes of circulation, and so the long term perfor-
mance of the heparin coating was evaluated separately. The in vitro ageing indicated that
some heparin may be lost during the washing phase, followed by a slow release of heparin
over a period of weeks and months. This is the first reported study on the durability of
heparin coating on dynamically strained substrates. Dynamically stretching the polymer
did increase the water contact angle, but there was no statistically significant change in
S 2P surface concentration, indicating that dynamic strain may not in fact accelerate the
loss of heparin. On the other hand, ageing in hydrogen peroxide destroyed the coating,
suggesting that macrophages may affect the coating. The durability of the Corline Hep-
arin Surface on pyrolitic carbon was evaluated by Kristensen et al. (2006) using XPS and
antithrombin binding capacity. Although they also observed a loss in heparin conjugates
from the carrier chain after 3 weeks of continuous high shear flow, there was no reduction
in antithrombin binding capacity.
The inactivation of heparin via ion-exchange mechanisms with endogenous cations and
proteins has also been reported suggesting that long term performance is unlikely to
be satisfactory even if heparin is not leached from the surface (Amiji and Park, 1993).
Heparin may mediate the short and medium term response and ideally aid the formation
of a stable fibrous capsule, but long term performance requires further investigation,
particularly for this polymeric substrate.
3.4.1 To anticoagulate?
I have also to consider whether there was a bias in my selection of healthy donors, who may
not be clinically representative of patients receiving a prosthetic heart valve (Vienken,
2008). Furthermore, even when bioprostheses are implanted, patients are typically pre-
scribed a mild anticoagulant, aspirin or clopidogrel for example. To investigate whether
this resulted in any significant effects, I added 20 mg l−1 aspirin to blood from just one
donor. Following the same method as detailed in Section 3.2 the heparin coated BCP
was tested for haemocompatibility in blood anticoagulated with salicylic acid.
Anticoagulation of the blood samples led to a reduction in all measures of coagulation
activation and inflammation after rotation in the Chandler Loop. Anticoagulation of
blood did not affect the haemocompatibility parameters for blood which was contacted
with the heparin coated material. Overall, the addition of aspirin makes the testing
method less sensitive to differences between materials.
Low-dose aspirin is prescribed for post-operative patients as an antiplatelet drug due to
its ability to inhibit the production of thromboxane and prostacyclin. Both are lipids
112 Chapter 3. Biocompatibility
which can be found in platelets, and are involved in platelet activation. The activated
form, thromboxane A2, has prothrombotic properties; it stimulates the activation of new
platelets and increases platelet adhesion. Aspirin inhibits the ability of cyclooxygenase
I enzyme to synthesize thromboxane and prostacyclin (Kyrle et al., 1987). The heparin
coated material activates antithrombin, leading to an inhibition of thrombin and factor
Xa which breaks the coagulation cascade, making the anticoagulating action of aspirin
futile.
In summary, anticoagulation is unlikely to mediate impact of implanting a heparin coated
BCP, however, the thrombus-inducing shear stresses may still warrant anticoagulation
after the implantation of polymer leaflet heart valve.
3.5 Conclusion
I tested a shortlist of nano-cylinder forming BCP with favourable mechanical properties
for haemocompatibility and cell viability according to ISO 10993, with a view to their
use in polymer-based prosthetic heart valves. None of the BCP test materials elicited
an adverse host response in haemocompatibility or direct fibroblast contact results. In
conclusion, I am able to rank the materials, as shown in Table 3.1.
Table 3.1: Qualitative ranking of tested materials for activation of the coagulation cascade
and inflammatory reaction
Least activation of
the coagulation cascade
Heparin Coated SEPS22
Bovine Pericardium
ePTFE
SEPS22, SI/BS19, SIBS30, SIS30
Polyester
Greatest activation of
the coagulation cascade
Least acute inflammatory reaction
Heparin Coated SEPS22
SEPS22, SI/BS19, SIBS30, SIS30, ePTFE
Bovine Pericardium, Polyester
Greatest acute inflammatory reaction
The uncoated BCPs performed better than polyester reference material in terms of coag-
ulation, platelet activation and inflammation, indicating that their haemocompatibility
should be sufficient for use in a prosthetic heart valve. One of the BCPs, SIBS, has re-
ceived considerable attention as a haemocompatible material (Pinchuk et al., 2008): in this
Chapter 3. Biocompatibility 113
study, I show that this accolade may be extended to several other BCPs. Furthermore,
the addition of a heparin coating results in a significant improvement in haemocompati-
bility parameters. The heparin coating was found to be largely still present after 3 weeks
of testing under dynamic stretching in PBS, although it was degraded after contact with
oxidants. Based on these results, none of the tested BCP should be eliminated on the
grounds of haemocompatibility. Relative to bovine pericardium, the BCP are less inflam-
matory, and further development of the valve can continue to use SEPS22, as selected in
Chapter 2.
114 Chapter 3. Biocompatibility
Chapter 4
Calcification
4.1 Introduction
The principle cause of aortic stenosis is calcification of the valve. Calcification involves
the nucleation and growth of calcium phosphate crystals, necessary for bone, but un-
wanted elsewhere. Although extracellular calcium concentrations mean that the solubil-
ity product is reached for several calcium salts, active cell mechanisms have evolved to
limit nucleation and propagation of calcium deposits in healthy, soft tissues. Patholog-
ical valves and prostheses lack these processes making them susceptible to calcification
(Siddiqui et al., 2009).
In this chapter I first review the calcification of native valves before going on to consider
bioprosthetic and polymeric valves. Based upon this review, a novel testing device is
designed and built. This is used to determine the propensity of BCPs to calcify and the
effect that this has upon the material’s function.
4.1.1 Calcification of native valves
Researchers have long sought to determine the aetiology and mechanisms of native valve
calcification such that it may be predicted, prevented, and cured. In 1663 Rivie`re re-
ported “caruncle-like masses obstructing the left ventricular outflow to the aorta”, and
in 1846 Hasse suggested that degenerative valve disease was a result of valve ossificiation.
Mo¨enckeburg’s description of valvular calcification in 1904 was the first to identify the
deposition of calcium-phosphate on the leaflets, causing sclerosis (Leopold, 2012).
Aortic stenosis caused by calcification is the dominant pathology of the aortic valve:
115
116 Chapter 4. Calcification
it outnumbers cases of regurgitation by three to one in developed countries (Iung and
Vahanian, 2011). During stenosis, the aortic valve thickens and stiffens, with calcific
nodules forming on the aortic side of the leaflets. Brittle crystals are formed on the
leaflets by the most thermodynamically stable calcium-phosphate structure, hydroxyap-
atite, Ca10(PO4)6(OH)2. Hydroxyapatite (HAp) is the most abundant crystal form in
bone, though bone also contains carbonate and magnesium crystals, crystal defects, and
other types of crystalline structures making it more of a “bioapatite” (Chakraborty et al.,
2006; Demer and Tintut, 2008).
Calcific nodules are certainly associated with the macroscopic stiffening of aortic valves
and other tissues, but it appears as though HAp particles are present long before the
manifestation of disease. Bertazzo et al. (2013) used nano-analytical electron microscopy
to examine calcified and non-calcified human leaflets. They found spherical (500 nm to
5 µm diameter) particles on leaflets before any other symptoms of calcification such as
stiffening and fibrosis. These HAp spheres were not seen in bone, and they were in fact
more crystalline than bone, suggesting that therapeutic dissolution of these nucleation
points would be impossible without also dissolving bone.
HAp particles seem to be present in all valves, but osteogenic cells are certainly involved
in the heterotopic ossification of pathological leaflets (Egan et al., 2011). Circulating
osteogenic precursor cells can be found at the valve, along with native aortic valve stromal
interstitial cells and both morph into an osteoblast bone-like phenotype responsible for
the bone-like calcifications (Cho and Kim, 2009; Giachelli et al., 2005; Rajamannan et al.,
2011).
Although the stiffening and calcification of valves shares the characteristics of inflam-
mation and the presence of bone morphogenic proteins with atherosclerosis, valvular
pathologies are distinct (Akat et al., 2009). Most prominently, statins appear to be a
miracle drug with respect to atherosclerosis, but no benefit has been seen in treatment
of human valvular disease (Amann, 2008; Ehrenstein et al., 2005; Farzaneh-Far et al.,
2001; Helske and Kupari, 2007; Hermans et al., 2010). The finding of cells that cause
calcification certainly aids the search for new therapies to prevent or halt the progression
of valvular calcification (O’Brien, 2006), but it is impossible to know whether these cells
might cause the calcification of a polymeric prostheses.
4.1.2 Calcification of bioprostheses
Calcification is also associated with bioprosthetic valve failure. Pure stenosis as a result
of calcification leads to 10-15% of failures in porcine bioprostheses. 75% of failures are
Chapter 4. Calcification 117
caused by tears, which are accompanied by calcification, and are manifested haemody-
namically by regurgitation (Siddiqui et al., 2009). During analysis of explanted biopros-
theses from both humans and animals, antibodies and mononuclear inflammation were
detected, indicating that inflammatory and immune processes play a role in calcification.
However, calcific valve failure does not satisfy Koch’s Postulates of disease (Anderson
et al., 2004). There was no significant difference in the mineralisation of subcutaneously
implanted bioprosthetic valve leaflet material in mice with or without an immune system.
This implies that the observed inflammatory response is more likely to be a secondary
response to valve damage rather than a cause of failure (Levy et al., 1983; Schoen and
Levy, 2005).
In the native healthy valve, active inhibitors and calcium ion pumps reduce the con-
centration of calcium from approximately 10−3M extracellular to 10−7M intracellular,
limiting crystallisation of the ECM. The tissue of a fixed bioprosthetic valve lacks these
active cell processes. Furthermore, the fixed tissue abounds with nucleation sites such
as phosphate lipids in the dysfunctional cell membranes, alkaline phosphatase, and col-
lagen and elastin fibres (Levy et al., 1983; Schoen and Levy, 1992, 2005; Siddiqui et al.,
2009). Indeed, when bioprosthetic valve material is placed subdermally in rats, mineral
deposition commences within 24 to 48 h (Schoen et al., 1986).
Full calcification of bioprostheses takes ∼2 years in children and ∼13 years in young
adults suggesting that bioprosthesis calcification is limited by crystal growth rather than
nucleation. Calcification is faster in younger patients due to higher levels of extracellu-
lar phosphate (Jono et al., 2000), and their more active immune system which causes
more fibrocalcification (in which a combination of inflammatory and osteogenic cells are
implicated) (Hopkins, 2006; Weiss et al., 2013).
The dynamic valve environment also plays a role in bioprosthetic valve calcification.
Calcified nodules are known to localise to regions of high mechanical stress. Flexure and
shear lead to the the disruption of collagen fibres which then serve as focii for calcium-
phosphate nucleation (Bernacca et al., 1992a; Thubrikar et al., 1983). As such, stress
reductions can lead to significant improvements in bioprosthetic heart valve durability.
Anticalcification treatment of bioprostheses can disrupt calcification by removing dead
cell debris, phospholipids, and other lipids; modifying collagen structure or chemistry; or
restoring of glycoaminoglycans (Schoen and Levy, 2009). Specifically, tissue treatments
have used ethane hydroxybiphosphonate, α-amino-oleic acid, ferric or aluminium chloride
exposure, or sodium dodecyl sulphate (Vyavahare et al., 1997a). In my experiments I
opted to use an ethanol incubation step after glutaraldehyde fixation to alter the amide
groups of collagen and remove cholesterol and phospholipids, which has been shown to
118 Chapter 4. Calcification
significantly reduce calcification (Vyavahare et al., 1997b).
4.1.3 Polymeric valve calcification
Calcification was reported in 6 of the 27 polymeric prosthetic heart valve trials listed
in Section 1.8. Of these, failure as a result of calcification was reported in the valve
made from SIBS with embedded PET fabric and dimyristoyl phosphatidylcholine coating
(Wang et al., 2010) and a polyurethane valve (Hilbert et al., 1987). Calcification has also
been reported as a cause of failure for the polymeric membranes of total artificial hearts
(Imachi et al., 2001; Yang et al., 1999).
On the one hand, like bioprostheses, polymers lack the active ion pumps of living tissue.
On the other, polymers do not contain the phosphorus rich cell debris and disrupted
collagen that encourages nucleation. On blood contacted polymeric surfaces two types of
calcification have been identified. First, heterogeneous nucleation of calcium-phosphate
crystals can occur on the surface or at the interface between the leaflet surface and mi-
crothrombi or fibrous sheaths; this mechanism will be referred to as surface calcification.
Second, degenerated cells, which are found in thrombotic material or the fibrous sheath,
can become calcified; which will be referred to as calcification of adsorbed (biological)
matter (Hilbert et al., 1987).
Pettenazzo et al. (2001) performed a comparison of in vitro and in vivo calcification of bio-
prostheses. For the in vitro environment the valves underwent accelerated pulsatile test-
ing in a simulated body fluid with no biological material. The in vivo valves were identical
St. Jude Medical bioprostheses which had failed and been explanted. Using SEM and X-
ray powder diffraction (XRD), they showed that very similar hydroxyapatite-octacalcium
phosphate deposition occurred in both cases, indicating that biological mechanisms are
unlikely to be causal (Pettenazzo et al., 2001). In vitro studies by Bernacca et al. (1992b);
Boloori Zadeh et al. (2013); Hamon et al. (1982), and the in vivo trials of the “jellyfish”
valve (Imachi et al., 2001), all considered calcification to be a surface process (as opposed
to calcification of adsorbed matter).
Calcification of adsorbed biological matter is unlikely to be a major cause of failure, and
so I will not consider the chemical-biological synergies in this chapter. Nevertheless, as
BCP materials are less inflammatory and adsorbent than bovine pericardium in vitro,
as well as being smoother, they should accumulate less cellular debris, making them less
vulnerable to adsorbed matter calcification than bioprostheses.
It is a simpler task to isolate the factors which affect the rate of calcification of polymers,
Chapter 4. Calcification 119
than when biological matter is present. Heterogeneous crystal nucleation is a balance of
enthalpy vs. entropy, which is controlled by the local chemical potential (or supersatu-
ration) of the ions, the surface chemistry and morphology. The interfacial free energy at
the surface is always a positive term, and destabilises small crystal nuclei. When nuclei
above a critical size are formed, they may grow at the surface (DeYoreo and Vekilov,
2003).
With a view towards understanding physiological calcification of hard tissues such as
teeth and bones, Liu et al. (2012) performed a fundamental study into the effect of
surface functionalisation on calcium-phosphate crystallisation. Gold surfaces were func-
tionalised with equal surface densities of −COOH, −CH3 and −NH2, and exposed to
calcium-phosphate solutions. SEM, XPS and AFM were used to assess the morphology
and rate of calcium phosphate deposition. They found that amine and carboxylic acid
surfaces became calcified, whereas the methyl surface inhibited calcification and had a
lower nucleation and growth rate than the raw gold control surface.
Two mechanisms were proposed: on negatively charged surfaces (such as the surface
coated with carboxylic acid groups) Ca2+ were adsorbed and concentrated before PO4
3–
ions were adsorbed to form the crystal framework. On the positively charged −NH2
surface, homogeneous clusters of Ca2+ and PO4
3– with a negative charge imbalance
were adsorbed to the surface. In both cases, once the first surface layer is formed, the
modulating effect of the functional groups becomes weak and crystal growth proceeds
as a function of solution concentrations and by the rate of mass transfer to the surface
(DeYoreo and Vekilov, 2003).
Despite most polymer prosthetic heart valve surfaces being uncharged, calcification is still
observed. Various mechanisms have been proposed to affect surface calcification, these
can be divided into the following categories:
• Biological (without surface debris): native calcification inhibitors are removed, whilst
proteins with an affinity for calcium transport calcium-phosphate to the valve site (An-
derson, 1981; Fisher and Termine, 1985).
• Chemical cation complexation: calcium forms complexes with appropriate functional
groups within the biomaterial, polyether soft segments of polyurethanes for example,
which then sequester phosphate to maintain charge neutrality, resulting in the deposition
of insoluble calcium phosphate within the material (Hamon et al., 1982; Long and Urry,
1981).
• Mechanical: high cyclic stress and strain results in macromolecular breakdown, expos-
ing calcium binding sites (Harasaki et al., 1987; Thubrikar et al., 1983), and microcracks
120 Chapter 4. Calcification
(Imachi et al., 2001) in which calcium-phosphate nucleates.
• Surface morphology: rough surfaces promote nucleation (Boloori Zadeh et al., 2013).
Regardless of polymer type or valve design the proposed prosthesis will not contain active
cells, and thus no native calcification inhibitors will be present. Therefore the biological
hypothesis does not need to be tested in the following experiments. The chemical cation
complexation mechanism will be considered by testing various types of polymer and
coating. Similarly, mechanical mechanisms and surface morphology will be considered,
as they can be controlled by valve design and manufacture respectively.
Cyclic strain is necessary in in vitro models. Bernacca et al. (1992b) observed negligi-
ble calcification in static tests on polyurethane, but the same materials calcified signif-
icantly under cyclic strain in a valve, eventually failing. In the absence of mechanical
strain, artificial surface defects and roughness on Angioflex (a proprietary polyether based
polyurethane) increased the rate of calcification (Boloori Zadeh et al., 2013). Bernacca
et al. (1992b) and (Imachi et al., 2001) both observed calcification deposits localising to
areas of high strain. Imachi et al. (2001) hypothesised that repetitive strain may create
(micron-sized) cracks in the polymer surface exposing previously covered calcium binding
sites, allowing calcium or phosphate ions to congregate, precipitate, stiffen the material
and lead to further breakdown.
None of the aforementioned studies make a distinction between polymeric prosthesis
calcification which leads to stiffening and stenosis, and calcification that is associated
with tearing and regurgitation. Presumably stiffening of polymeric materials can lead
to tearing during flexure. In the following experiments the effect of calcification on
mechanical properties will be considered. Furthermore, I will aim to find out whether
polymer mechanical fatigue, or calcification cause failure (Bernacca et al., 1992b).
4.1.4 Polymer anticalcification strategies
In response to the failure of several polymeric heart valves, various anticalcification strate-
gies have been tested. The segmented polyurethane, Biospan®, received a biphosphonate-
diethylamino coating (via bromoalkylation). Biphosphonates are well known to block the
growth of hydroxyapatite, whilst the diethylamino groups reduce the water uptake of the
polyurethane, improving the elastomer’s durability and successfully improving the valve
(Alferiev et al., 2003).
The grafting of sulfonated poly(ethylene oxide) to the polyether polurethane (Pellethane
2363-80AE) reduced calcium deposition on leaflets implanted in a dog by more than 50%
Chapter 4. Calcification 121
(Han et al., 2006). The sulfonated poly(ethylene oxide) groups are highly dynamic, hy-
drophilic, and negatively charged, possibly repelling the phosphate groups. Bernacca and
Wheatley (1999) applied heparin, taurine, 3-aminopropyltriethoxysilane and polyethylene
oxide surface modifications to polyetherurethane (chain extended with either butanediol
or ethylene diamine). They reported that polyethylene oxide modification led to a lower
durability and greater calcification, whilst the other modifications improved durability.
The UCL group tested the anticalcification potential of polyhedral oligomeric silsesquioxane-
poly(carbonate-urea)urethane (POSS-PCU) relative to glutaraldehyde-fixed bovine peri-
cardium and polyurethane, using an in-house accelerated pulsatile pressure device to load
membranes of equal thickness of the materials to 120 mmHg. The polyhedral oligomeric
silsesquioxane separates into nano-sized particles in the polymer, which increases the
tensile and tear strength of the material, as well as reducing hydrophilicity and water ab-
sorption. In their study the POSS-PCU material calcified less, and had a smaller change
in mechanical properties than the polyurethane reference sample (Ghanbari et al., 2010).
As the authors did not run the experiment without the calcium and phosphate ions, it is
impossible to ascertain to what extent the acceleration of degradation could be attributed
to calcification. Furthermore, polyurethane has a lower modulus than POSS-PCU, mean-
ing that polyurethane strained more than POSS-PCU under equal pressure, indicating
that larger or more micro-cracks may be formed in polyurethane.
4.1.5 Calcification testing
It is generally accepted that in vitro models of calcification are an acceptable model
for in vivo studies if ion concentrations are maintained at physiological levels (single
batch testing being unacceptable) (Park et al., 2000), and materials are dynamically
loaded (Bernacca et al., 1992b; Krings et al., 2006; Mavrilas, 2004; Mavrilas et al., 1999;
Pettenazzo et al., 2001). There is no consensus regarding the composition and control
of the calcifying solution. Some studies have used physiologically representative buffered
solutions which produce realistic bioapatite structures containing other ion defects (such
as magnesium and carbonates) (Golomb and Wagner, 1991; Starcher and Urry, 1973),
while others have removed the buffers, and used control of pH to maintain appropriate
concentrations of calcium and phosphate ions (Kapolos et al., 1997; Krings et al., 2006).
In the pH control method, the measurement and control of pH allows the calcification
rate to be inferred for single samples.
ISO 5840 now mandates calcification testing of devices prior to accreditation, but given
the widespread occurrence of calcification, a rational design methodology for prosthetic
122 Chapter 4. Calcification
heart valves should evaluate the potential of materials to calcify. The screening mecha-
nism should recreate physiological tensile strains (10− 15% (Balachandran et al., 2011))
and bending (∼ 10% (Corden et al., 1995)), in a physiologically representative chemical
and thermal environment.
4.2 Materials and methods
4.2.1 Calcification
A novel calcification device was fabricated in the Department of Chemical Engineering
and Biotechnology, University of Cambridge. The device, shown in Figure 4.1, allows
up to 21 samples to be simultaneously flexed and strained whilst immersed in calcifying
solution. The device is made from 10 mm PMMA, and the samples are secured using
PMMA washers and polystyrene screws (to reduce the dissolution of unwanted ions).
BCP materials (SIBS30, SI/BS19, SIS18, SIS30, and SEPS22) were manufactured by
compression moulding as previously described (see Section 2.2). Dynamically strained
oriented samples were stretched in the stiff direction. For comparison, SIBS30 was also
solvent cast. In a second run, performed under my supervision by Miranda Xu and An-
drew Cheung, the calcification of heparin coated SEPS22 was evaluated (Corline Heparin
Surface, as previously described in Section 3.2.2). Porcine pericardium tissue was ob-
tained from a local slaughterhouse (Huntingdon Life Sciences). The tissues were washed
in phosphate buffered solution before fixation in glutaraldehyde solution: 0.4% for 48 h
followed by 0.2% for 1 week. A selection of the pericardium samples underwent anti-
calcification treatment using 80% ethanol (Vyavahare et al., 1997b), these being referred
to as Pericardium AntiCa. The reference polyurethane, Pellethane 2363-80AE (further
details in Section 2.2), was also formed by compression moulding.
Samples measuring 50× 8× 0.5 mm were cut using a straight blade and mounted either
on a static plate or on the deflection device. The plate and device were placed in a plastic
container holding the calcifying solution. The container was placed in a water bath which
was held at 37 ◦C. Between experiments the container and stretching device were washed
using 0.1M nitric acid.
Chapter 4. Calcification 123
D
efl
ec
to
r
Static sample
Dynamic sample
Figure 4.1: Cross section detail of the novel calcification device showing the 3-point bend
and static test positions and mounting systems immersed in calcifying solution. The
polymeric samples are indicated by the red-blue lines. The deflector for dynamic testing
is driven by a rotating cam. The static samples are situated directly below the strained
samples.
Calcifying solution
The calcifying solution contained ions at the concentrations listed in Table 4.1. The
solution was brought to pH 7.2 using HCl solution. All reagents were purchased from
Sigma Aldrich (Gillingham, UK) and solutes were dissolved in reverse osmosis water.
The solution was changed every 3 days during the experiment to maintain the electrolyte
concentrations and pH. Samples were also tested in a solution that did not contain Ca2+
or HPO4
2– ions (“non-calcifying solution”), but was otherwise identical.
4.2.2 Analysis of calcification
After calcification, samples were briefly rinsed in distilled water. The samples were cut
from the device and dried before the calcium-phosphate deposits were dissolved by placing
the samples in 0.5M nitric acid solution. Calcium concentration has been evaluated by
other authors using inductively coupled plasma (Boloori Zadeh et al., 2013), colorimetric
assay (Ghanbari et al., 2010), and atomic absorption spectrometry (AAS) (Brockbank
and Song, 2003; Pettenazzo et al., 2001). I selected AAS (using Varian AA 240, now
Agilent) for accuracy and speed. In AAS, the atomised solution absorbs at a particular
wavelength. While absorption should be linearly related to the ion concentration via the
Beer-Lambert law, this is rarely the case. A rational curve (Equation (4.1)) was fitted to
124 Chapter 4. Calcification
Table 4.1: Ion concentrations in calcifying solution and their plasma equivalents. The
solution was brought to pH 7.2 using HCl solution.
Ion Calcifying ion concentration (correspond-
ing plasma concentration (Kokubo, 1990))
(mM)
Na+ 142 (142)
K+ 5 (5)
Mg2+ 1.5 (1.5)
Ca2+ 2.5 (2.5)
Cl– 128 (103)
HCO3
– 27 (27)
HPO4
2– 1 (1)
MOPS buffer 5
the calibration readings using OriginPro 9 software.
a
c
= P +Qa+Ra2, (4.1)
where a is the measured absorption, c is the concentration of the solution, and P , Q, and
R are curve fitting parameters.
Tensile testing
The flexural stiffness of polymers was tested before and after static calcification using a
3-point bend fixture (Texture Analyser, Stable Microsystems, Surrey, UK).
Scanning electron microscopy
After calcification testing, dried polymeric samples were imaged using scanning electron
microscopy (SEM), with elemental ratios assessed using energy dispersive X-ray (EDX).
Atomic force microscopy
Sample roughness was evaluated using atomic force microscopy (AFM) (Veeco Dimension
3100, NY, USA). The root-mean-squared height at four spots on each sample was used
to characterise surface roughness. AFM measurements were performed by Miranda Xu
and Andrew Cheung, whose masters’ research project I supervised.
Chapter 4. Calcification 125
4.3 Results
4.3.1 Calcium surface concentration
The surface concentration of calcium on samples is shown in Figure 4.2. Dynamic strain
resulted in a statistically significant increase in calcification for all polymeric samples
(Student’s t test, p < 0.05). Absolute surface concentrations could not be compared
between the three runs, I, II, and III, as the mechanical, chemical, and thermal conditions
were not identical. In Figure 4.2b the calcium surface concentration after dynamic III
on the heparin coated SEPS22 is significantly lower (p < 0.05) than upon the uncoated
polymer.
4.3.2 SEM and EDX
A selection of SEM images are presented in Figure 4.3. The extent of crystallisation is
much greater on the dynamic BCP samples than on the static ones. Surprisingly, on the
static BCP sample surfaces (Figures 4.3b and 4.3d) a “mud cracking” pattern is present,
but could not be found on the BCP control samples. On the dynamic samples, the surface
cracking could only be found where the calcium phosphate layer was thin. Contrary to
expectations, such as according to the surface morphology hypothesis, crystalline deposits
did not localise to the cracks. The “mud cracking” surface feature could have been caused
by chemical or environmental attack (Hayes et al., 2015). This behaviour was also seen on
samples tested in the Chandler Loop for their haemocompatibility (Figure 3.2). Surface
cracking was observed on both the more durable (saturated and coated) polymers and
the unsaturated BCPs, which failed.
Liu et al. (2012) took images of calcification on uniformly functionalised surfaces of −NH2
and −CH3. In comparison to their images, those produced here do not resemble the
globular deposits caused by homogeneous nucleation as seen on the functionalised surface.
Furthermore there was an absense of a uniform HAp layer on the surface which developed
on their −COOH surface. The more dense and uniform distribution of crystals observed
on dynamic materials may result from fluid shear forces removing larger particles.
Analysis of the elemental concentrations of the crystals revealed Ca, P, Mg and Na
in decreasing concentrations. The atomic ratio of calcium to phosphorus indicates the
calcium-phosphate crystalline state. In HAp, Ca:P=5:3. The mean ratio over all sam-
ples was 1.96:1, with some points containing Ca-P ratios greater than 3:1, suggesting
that fully developed HAp was yet to form. EDX measurements performed by Pettenazzo
126 Chapter 4. Calcification
SI
S1
8
SI
/B
S1
9
SI
BS
30
SE
BS
20
Pe
ric
ar
di
um
Pe
ric
ar
di
um
A
nt
iC
a
P2
36
3
SI
BS
30
SC
0
2
4
6
8
10
12
14
16
2 2
2
C
al
ci
u
m
co
n
ce
n
tr
at
io
n
(m
g
cm
−2
) Static I
Dynamic I
(a)
SE
PS
22
P2
36
3
SE
PS
22
H
ep
ar
in
Pe
ric
ar
di
um
0
2
4
6
8
10
12
14
16
∗ ∗Ca
lc
iu
m
co
n
ce
n
tr
at
io
n
(m
g
cm
−2
)
Control II
Static II
Dynamic II
Dynamic III
(b)
Figure 4.2: Results from the measurements of calcium on the samples. (a) shows the
surface concentration of calcium on the test materials after 12 days in solution. The error
bars indicate the minimum and maximum recorded values (3 to 5 samples per material).
The 2 symbol is used to indicate the samples that failed during testing. (b) shows the
calcium surface concentrations for a smaller range of materials in experiments performed
by Miranda Xu and Andrew Cheung. One experiment (bars marked II) was run with all
materials (Control, n=5; Static, n=5; Dynamic, n=5). A second experiment (Dynamic
III) concentrated on SEPS22 and heparin coated SEPS22 (n=10). In both experiments
the materials were calcified for 8 days, but there were fluctuations in temperature. The
error bars indicate the standard deviation of readings. The bars marked with ∗ are
significantly different from one another (Student’s t-test, p < 0.05). Full details on the
polymers can be found in Table 2.2.
et al. (2001) of calcified bioprostheses indicated that octacalcium phosphate (Ca:P=4:3)
was formed before HAp, though their study was considerably longer (44 vs. 8 days).
The authors also suggested that amorphous calcium-phophate was a precursor to oc-
tacalcium phosphate and HAp on the valve. Amorphous calcium phosphate is formed
with 1:1<Ca:P<11:5, which could account for the high ratios measured here (Dorozhkin,
2009). The presence of magnesium (with the modal Ca:Mg=11:1) suggests that whit-
lockite (Ca:Mg=9:1), the second most abundant crystal form in bone, may be present.
This mixed elemental composition of the crystalline deposits is indicative of bioapatite.
Chapter 4. Calcification 127
(a) (b)
(c) (d)
Figure 4.3: SEM micrographs of BCP samples before and after calcification. Image (a)
is an uncalcified sample, showing a potential surface defect or contaminant upon which
crystals might seed. Image (b) is statically calcified SIBS30, (c) is dynamically calcified
SIBS30, and (d) is statically calcified SIS18. Most of the white deposits on the surface
show calcification. The scale bar indicates 50µm. The red arrows in (b) and (d) indicate
cracks on the surface of the static samples which, contrary to expectations, the calcium-
phosphate crystals did not localise to.
4.3.3 Effect on mechanical properties
Calcification of prosthetic heart valves can lead to sclerosis or tearing. During my first
calcification study, the SIS18, SIS30, SEBS20, and SI/BS19 samples failed due to tearing
or cracking within 5 days or 9× 105 cycles (Figure 4.4). Tears occurred near to the
deflector. Split samples were sutured together to continue their dynamic movement. All
dynamically strained samples crept during the test, which aided the survival of some
samples.
In the second set of calcification studies, SEPS22, heparin coated SEPS22, P2363 and
pericardium were tested. Some of the coated and uncoated SEPS22 polymer also failed
during this test. However, failure rates should not be compared between the runs as strain
varied. Table 4.2 shows that the heparin coating brings about a marked improvement
128 Chapter 4. Calcification
in the polymer’s durability. Furthermore, a greater proportion of SEPS22 samples failed
in calcifying solution than in the non-calcifying solution, indicating that calcification
accelerated failure for this uncoated polymer.
The root mean squared roughness of SEPS22 and heparin coated surfaces was measured
to be 255± 118nm and 128± 42nm, respectively. The uncoated surface was significantly
rougher (Student’s t-test, p < 0.95) than the coated surface, which may contribute to the
higher calcium concentration.
SI
S1
8
SI
/B
S1
9
SI
BS
30
SE
BS
20
Pe
ric
ar
di
um
Pe
ric
ar
di
um
A
nt
iC
a
P2
36
3
SI
BS
30
SC
0
2
4
6
8
10
12
S
u
rv
iv
al
ti
m
e
/1
2
d
ay
s
Calcifying solution
Non-calcifying solution
Figure 4.4: Days to failure for samples in calcifying and non-calcifying solution when
dynamically strained at 2 Hz. Bars which reach 12 days correspond to samples which
did not fail. There were 3 to 5 samples of each material and the bars are shown as an
average. AntiCa indicates that the pericardium underwent anticalcification treatment.
Full details on the polymers can be found in Table 2.2.
To investigate the stiffening of polymers, secant moduli (at 5, 10, 20% strain) were
measured before and after calcification of statically mounted samples. Figure 4.5 shows
that despite a large variation in surface calcium concentration, samples stiffened by a
uniform amount of 0.1 to 0.2.
Chapter 4. Calcification 129
Table 4.2: Proportion of samples surviving immersed durability tests in calcifying and
non-calcifying tests, as performed by Andrew Cheung and Miranda Xu.
Number of samples failed in test period
Material Calcifying solution Non-calcifying solution
SEPS22 12/15 4/5
SEPS22 Heparin 5/15 3/10
P2363 0/5 Not tested
5
· 10
−2 0.1 0.1
5
0.
2
0.
25 0.
3
0.
35 0.
4
−0.1
0.1
0.2
0.3
0.4
Calcium surface concentration (mg cm−2)
C
h
an
ge
in
fl
ex
u
ra
l
m
o
d
u
lu
s
Figure 4.5: Proportional change in secant moduli (5, 10, 20%) of uncoated polymeric
materials after static calcification as a function of the calcium surface concentration.
Each point represents a polymeric material, and error bars represent the range of values.
4.4 Discussion
The synergy of mechanical stress and calcification is well documented and so the calcifi-
cation of valves is affected by the stress distribution. As such, calcification studies should
isolate material effects from design. The calcification of isolated heart valve materials by
Boloori Zadeh et al. (2013), Mavrilas (2004), and Park et al. (2000) used a continuous flow
of solution, or calcium and phosphorus levels were topped-up by each of these authors,
but none dynamically strained the material. Ghanbari et al. (2010) are the only authors
to dynamically strain isolated heart valve materials. Their test system design marks a
significant improvement in calcification testing, and demonstrated that the group’s ma-
terial, POSS-PCU, was superior to polyurethane and pericardium. No anticalcification
treatment was applied to the pericardium, as is now commonplace among bioprostheses.
The samples underwent tensile strain as opposed to flexure - an important deformation
in prosthetic heart valve calcification. Also, the softer polyurethane would have been
130 Chapter 4. Calcification
strained significantly more than POSS-PCU, whereas in an optimal heart valve softer
materials allow for thicker leaflets, reducing the stress and strain during diastole (refer
to Chapter 2 for further information).
In response to these shortcomings I proposed a novel device to improve the simulation of
material calcification in vitro. In the device, a large number of immersed samples (up to
21) simultaneously undergo a 3-point bend. Deflection was tuned to achieve physiological
strains of tensile (10 − 15%) and bending (∼ 10%), in a physiologically representative
chemical and thermal environment. The device successfully strained samples and accel-
erated their calcification. Unfortunately, sample creep resulted in changes in stress and
strain over time which the fixed displacement deflector could not adjust for. An im-
provement would be to retain the 3-point flexure characteristics, but use constant cyclic
pressure or stress.
Dynamically straining materials led to a large increase in the rate of calcification on
all materials. Pericardium samples calcified significantly more than polymeric samples.
Despite the sample size being limited no large differences were observed when treated
and untreated pericardium were compared.
Heparin is the most negatively charged biological molecule, implying that it could promote
the adsorption of Ca2+ ions and accelerate calcification. In fact, the heparin coated
surface calcified significantly less than the uncoated SEPS22. It was a surprise to observe
that the heparin coating led to a significant improvement in the durability of the polymer.
When I previously imaged blood contacted heparin coated SEPS22 (Figure 3.2), cracking
was visible beneath the confluent protein layer.
Submerged polymers failed during dynamic straining. Using some polymers can lead
to environmental stress cracking (ESC) under some conditions. Solvents with a similar
solubility parameter to the polymer lead to crack tip softening, causing crazing at the
crack tip and ultimately leading to crack growth and failure (Wright, 1996). The mud-
cracked surface that I observed in my samples seems to have an environmental cause
as it occurs in both blood and water, but is unlikely to be true ESC as no solvents are
present. Further experimentation should evaluate whether solvent cast polymers (which
have a smoother surface and fewer frozen-in stresses) also exhibit surface mud-cracking
upon contact with aqueous solutions.
Bernacca and Wheatley (1999) reported an improvement in polyurethane durability after
heparin coating. Among hydrophilic polyurethanes, coatings which reduce water absorp-
tion can lead to durability improvements (Alferiev et al., 2003). This mechanism however,
is unlikely to be at work in SEPS22 as it has a very low water absorbency and is quite
Chapter 4. Calcification 131
hydrophobic (water contact angle ∼ 90°). In my experiments there was contact with
a rigid bar, and in the valve there is leaflet-leaflet contact. I suggest that the heparin
coating reduces surface wear during contact.
In Figure 4.5 I showed that among the polymeric samples which were statically calci-
fied, the change in stiffness of the sample was not related to the surface concentration
of calcium. Over a 7-fold variation in the surface concentration, the stiffness of each
material increased by a uniform 10 to 20%. This result suggests that the concentration
of calcification is not the only marker of a change in mechanical properties.
Ghanbari et al. (2010) reported that polyurethane samples were weakened after calci-
fication testing as they displayed only 25% of their original tensile strength. However
this could be attributed to mechanical rather than chemical fatigue in their experiment.
POSS-PCU retained its tensile properties (despite also calcifying). In my study, calci-
fication led to stiffening of static samples, and snapping of flexed samples. As such, I
propose that calcification fills surface defects: this makes polymers stiffer when flexed
and eventually causing their failure, but has little effect their properties in tension. This
proposition is well aligned with the observations made of porcine bioprostheses which
typically sclerose or tear at regions of flexure (Siddiqui et al., 2009).
Various hypotheses have been proposed to account for the difference in calcification rate
between polymers (see Section 4.1.3): mechanical (Harasaki et al., 1987; Imachi et al.,
2001; Thubrikar et al., 1983), chemical (Hamon et al., 1982; Long and Urry, 1981),
surface morphology (Boloori Zadeh et al., 2013), and biological (Anderson, 1981; Fisher
and Termine, 1985). Micron-sized surface cracks were observed upon my materials but
calcium-phosphate crystals did not localise to them. Assuming that the cracks were
present during calcification, this counters the micro-crack theory of Imachi et al. (2001).
Bernacca et al. (1992b) questioned whether calcification or mechanical fatigue was causal
of failure. A higher proportion of coated and uncoated SEPS22 samples failed under
calcifying conditions than non-calcifying conditions indicating that calcification can ac-
celerate failure. It is well known that strain reduction certainly improves leaflet durability
(see Chapter 2 for further details). Strain reduction also reduces calcification, though it
is unclear at what point a reduction in calcification reduces either tearing or stiffening.
Further work would need to investigate the quantitative relationship between (cyclic)
strain and calcification rate to ascertain whether there is a critical strain level at which
calcification becomes accelerated. With regard to informing the design of the prosthetic
heart valve, the importance of strain reduction is reiterated by these studies.
132 Chapter 4. Calcification
4.5 Conclusion
In my investigation of calcification I tested the effect of calcium phosphate crystals upon
BCPs. Calcification stiffens the polymers, but also tears them during flex. Material
calcification thus has a bearing on heart valve design. The polymers calcified significantly
less than the biological materials. However, BCP durability may be compromised by
immersion in solution, and further reduced by calcification. The addition of a heparin
coating improved durability under both conditions, and could be used in the final valve
if durability is poor.
The rate of crystallisation is the sum of nucleation, growth, and removal rates; and each
one is controlled by surface chemistry, morphology, and deformations, alongside solution
chemistry and dynamics. SEM images indicated that calcium-phosphate crystals did not
localise to surface cracks. The rougher surface of uncoated SEPS22 corresponded with
its higher calcification rate. The negatively charged heparin coating would be expected
to increase calcification, on the contrary the rate was lower on heparin coated SEPS22,
suggesting that more complex chemistry would need to be investigated.
Chapter 5
Valve design
5.1 Introduction
For a prosthetic valve to approach the performance of a healthy, native valve, the prosthe-
sis’ form, the material selection and the manufacture all must work in perfect harmony.
Over 50 different designs of prosthetic heart valve have been implanted, yet none have
come close to matching the healthy, native valve for durability and haemodynamics (Dasi
et al., 2009).
The effect of the valve design on haemodynamics is the focus of this chapter. First I
identify the haemodynamic measures used to assess cardiac valves and prostheses. I go
on to consider the local anatomy and design constraints before reviewing previous leaflet
designs that have been trialled. This leads into my proposals for valve form and I go on
to identify two potential designs, which I prototyped and tested in a pulsatile flow loop.
I conclude by selecting a design for full manufacture.
5.1.1 Evaluating valve performance
A pathological valve can lead to an increase in cardiac workload, and an increase in
thrombus-inducing shear stresses in blood. A stenotic aortic valve does not fully open
during systole, whilst a regurgitant valve leaks during disatole. In order to assess the
severity of valvular pathology (which may be asymptomatic), or the function of a pros-
thesis, valvular performance must be assessed quantitatively. In the clinic, (transthoracic)
echocardiography is the standard tool used to assess cardiac valves non-invasively. Car-
diac MRI is, however, preferred for measuring the geometric orifice area (GOA) of an
133
134 Chapter 5. Valve design
open valve (Garcia and Kadem, 2006). Sonic reflections are used to detect structures in
the heart, and the Doppler effect is used to measure blood flow velocities. Echocardiogra-
phy gives direct measurements of velocities in the left ventricular outflow tract, upstream
of the valve (indicated by the suffix v), in the valve orifice (indicated by the suffix o), and
on the aortic side (indicated by the suffix a) of the valve.
An estimate of the energy loss over the valve is often taken as a mean (systolic) (∆Pm),
or peak pressure drop across the valve (∆Pmax), which are inferred from velocity data
(using the Bernoulii equation) (Bach, 2010).
∆Pm = Pv − Po = 4
(
vo
2 − vv2
)
(5.1)
where P are pressures and v are velocities at the positions indicated by the suffixes.
The velocities are measured in cm s−1, and pressure in mmHg. The velocities can be
integrated over the systolic phase to eliminate the fluid acceleration terms and give the
mean gradient. The peak velocities yield the peak gradient.
The pressure gradient is dependent on the flow rate through the valve (i.e. velocity and
area), and so the Effective Orifice Area (EOA) is a better measure of valve performance.
EOA represents the cross sectional area of the vena contracta through the valve orifice
and is relatively independent of flow rate (Zoghbi et al., 2009). Inaccurate measurements
arise due to the increased opening of bioprosthetic valves at higher flow rates, changes in
the distribution of velocities through mechanical valves, and sonographer expertise, but
otherwise EOA is a good indicator of valve performance (Otto, 2012). EOA is calculated
based upon velocity measurements as:
EOAv =
Avvv
vo
(5.2)
where v are temporal average velocities mid-systole upstream and in the valve orifice, and
Av is the flow area immediately upstream of the valve (at the annulus/leaflet nadirs).
Although EOA gives a good indication of the resistance to flow regardless of flow rate, it
is highly dependent on the size of the prosthesis, and gives no information about patients’
cardiovascular requirements (Garcia et al., 2004). Losses over the valve as a function of
patients’ cardiovascular requirements can be accounted for by using an indexed effective
orifice area. Body surface area (BSA) is an excellent predictor of a patients’ cardiac
requirement, and can be estimated based on weight and height using the Du Bois formula
(Du Bois and Du Bois, 1916). The indexed effective orifice area (iEOA), is calculated
by taking the ratio of EOA to BSA. It is a remarkably good measure of whether a
Chapter 5. Valve design 135
prosthesis is functioning well in a clinical setting. ≤0.85 cm2 m−2 is used as a cut off to
indicate that a prosthesis is poorly matched for a patient, be that due to undersizing
or dysfunction, both of which are termed prosthesis-patient mismatch. By indexing the
EOA of a valve prior to implantation, clinicians can select more appropriate prostheses
(Bleiziffer et al., 2007; Pibarot and Dumesnil, 2006).
If cardiac catheterisation is performed, or prostheses are assessed in vitro, both pressure
and velocity measurements can be made. The combination of the Bernoulli and continuity
equations forms the Gorlin formula to estimate the EOAG of the valve:
EOAG =
Q
51.6×
√
∆Ps
ρ
(5.3)
Where Q is the temporal average, or root mean square, flow rate in ml s−1; ∆Ps is the
mean pressure drop over the valve (which is measured directly at the upstream and
downstream positions) in mmHg; and ρ is the density of the fluid in g cm−3. Equation
(5.3) is specific to the units mentioned as it is derived from the Bernoulli equation, and
the constant 51.6 is not dimensionless (International Standards Organisation, 2005).
The values of EOA which arise from the formulae in Equations 5.2 and 5.3, as used in
vivo and in vitro, respectively, should not be directly compared due to the difference in
measurement and flow setup (see Section 5.2.3). There do, however, exist internationally
agreed standards for EOAG measured in vitro. For a 23 mm diameter valve, the EOAG
must be at least 1 cm2 (International Standards Organisation, 2005).
The above parameters are good indicators of aortic stenosis, but are less relevant during
regurgitation. Regurgitation is the sum of the retrograde flow during closing (closing
volume, Vc), and when closed (leakage volume, Vl). In vivo regurgitation is assessed
by echocardiographic evaluation of the valve morphology, and the area and velocity of
regurgitant jets (see Section 1.3 and Table 1.2) (Vahanian et al., 2012). The test fluid was
distilled water, as per the ISO standard, although it is not an exact rheological match
for blood.
Regurgitation through an aortic prosthesis is normalised with respect to stroke volume.
The regurgitant fraction (Rf ) can be measured in vitro using an ultrasonic flow meter
as:
Rf =
Vrev
Vsv
(5.4)
Where Vrev is the total volume of retrograde flow, and Vsv is the systolic stroke volume,
over one cycle. According to ISO 5840, for a 23 mm valve, Rf must be less than 10%.
136 Chapter 5. Valve design
As clinicians and engineers have sought to improve the performance of prosthetic heart
valves, they have identified more accurate ways of comparing valves to one another.
Mean pressure gradients and EOA are useful, but they assume steady, lossless flow (as
per Bernoulli’s assumptions) and are not affected by the regurgitation through a valve.
For a more general application than these limited assumptions, Leefe and Gentle (1987)
proposed the use of energy loss to evaluate the performance of prosthetic heart valves.
The energy loss across a valve is a function of the pressure, velocity, and flow rate, before
and after the valve (Akins et al., 2008). It is relatively easy to evaluate pressure and flow
in vitro, and so energy loss can be calculated over one cycle, assuming that there is no
change in height over the valve:
Eloss =
∫
cycle
Q
[
(Pv − Pa) + ρ
2
(
v2v − v2a
)]
dt (5.5)
As it is impossible to measure full pressure and velocity data in vivo without very invasive
procedures, energy loss has been estimated by various groups using combinations of peak
jet velocity, aortic valve area, effective orifice area, mean pressure gradient, left ventricular
stroke work loss, and change in flow rate over time. Most of these estimates do indeed
correlate well with patients suffering from aortic stenosis (Akins et al., 2008; Otto, 2012).
The parameters above are focussed upon the increase in cardiac workload due to a faulty
valve. Although they are correlated with increased shear stresses through the valve, they
are not good markers of the shear-induced thrombogenicity of a valve. Particle image
velocimetry can be used to evaluate the flow and shear field through a valve during a
hydrodynamic test, allowing the platelet shear-activation to be estimated (Leo, 2005).
Recently, Girdhar et al. (2012) and Bluestein et al. (2013b) proposed the in vitro-in silico
Device Thrombogenicity Emulation method (DTE) to measure platelet activation. The
shear histories of platelets passing through a cardiac device are computed, based upon an
fluid-structure interaction (FSI) model. The shear histories are replicated on acetylated
blood in vitro, using either a cone and plate or couette shearing device. In the case of
a prosthetic heart valve, a modified LVAD can be used to simulate the flow of blood
through the valve (Claiborne et al., 2011). Acetylation removes the thrombin feedback
loop in the coagulation cascade and stops the formation of a fibrin network, allowing
the shear stress trajectories to be repeated. The platelet activation state is measured
using a modified prothrombinase assay (Jesty and Bluestein, 1999; Jesty et al., 2003).
The DTE is a valuable and novel method for the evaluation of cardiac devices, but its
computational and biological complexity makes it inappropriate at this stage of valve
design.
Chapter 5. Valve design 137
5.1.2 Valve design
The 1950s and 1960s saw the development of a broad variety of mechanical heart valve de-
signs. Whereas, since their inception, bioprosthetic and other flexible leaflet aortic valves
have successfully relied upon the tricuspid configuration as found in natural valves. As
tricuspid valves are found in nearly all mammals, it would appear that evolution has
optimised this design. Congenital bicuspid and unicuspid valves have a much greater
propensity for aortic stenosis, whereas quadricuspid valves can lead to aortic regurgita-
tion, reinforcing the idea that the tricuspid valve is the optimal design (also see Section
1.3.1).
Instead of merely observing the dominance of tricuspid valves in nature, there might be
a simple geometric argument for the trileaflet’s effectiveness; each leaflet must go from a
closed to open conformation as depicted in Figure 5.1. In a tricuspid valve the free edge
of the leaflet must only strain by 5% in order to open, if it is unstrained when closed. A
bicuspid valve, on the other hand, would need to strain by 55% to open. Conversely, the
closed free edge length in a quadricuspid valve is 27% longer than the open length. Large
strains reduce the durability of synthetic and natural tissues (Thubrikar and Klemchuk,
1990), and so for the aortic and tricuspid valves which are situated in stable annuli, the
tricuspid formation is the most effective for regulating one-way pulsatile flow.
R 2piR
3
Figure 5.1: A axial view onto the leaflet edges of an example tricuspid valve in a cylinder
of radius R. The free edge length of each leaflet in the closed position is equal to 2R (solid
red). To be fully open, the length of the free edge needs to increase to 2piR
3
(dashed red),
requiring the leaflet to strain by 5%. Adapted from Thubrikar and Klemchuk (1990).
138 Chapter 5. Valve design
5.1.3 Sizing, positioning and sewing ring
Surgical implantation of a prosthetic aortic valve gives the surgeon the freedom to decide
precisely how a valve is placed. There are variations in diameter along the axial length
of the root from left ventricle to the arch of aorta. Also, the root is not rotationally
symmetric, most notably due to the presence of the right and left coronary arteries. A
schematic of the aortic root cross section was previously shown in Figure 1.2.
The nadirs of the native aortic valve leaflets are near to the aortic annulus, while the com-
missures approach the level of the coronary arteries. During surgery, the native leaflets
are resected. Early prosthetic valves were positioned intra-annularly, which replicated
the positioning of the native valve and ensured that there was no interference between
the valve and the coronary arteries. However, aortic valve disease is often associated with
patients with small aortic annuli, and so surgeons try to maximise the orifice area of the
prosthesis. By positioning the valve supra-annularly, so that the stent and sewing ring
are accommodated by the sinus, the orifice area may be equivalent to the native orifice
area, improving haemodynamics (Wagner et al., 2007; Zhang and Wu, 2010).
Table 5.1: Aortic root dimensions for a cohort of 80 men and 97 women undergoing TAVR
for severe aortic stenosis, based upon CT data reported in Buellesfeld et al. (2013). The
annulus was defined as a plane through the nadirs of the 3 leaflets, and the left ventricular
outflow tract (LVOT), sinus and ascending aorta were defined as parallel planes proximal
or distal from the annulus. Abbreviated dimensions refer to those in Figure 1.2.
Height (mm)
Virtual derived area
diameter (mm)
LVOT -5 22.9 ± 3.2
Annulus (Laa) 0 23.3 ± 2.2
Right coronary ostium
(Lca−aa)
16.7 ± 3.6
Left coronary ostium
(Lca−aa)
14.4 ± 3.6
Sinus 21.9 ± 3 31.4 ± 14
Sino-tubular junction
(Lstj)
28 ± 3
Ascending aorta 30 29.3 ± 3.5
A prosthesis whose EOA is too low relative to a patient’s requirements will cause excessive
transvalvular gradients, and possible patient-prosthesis mismatch. The EOA at a specific
Chapter 5. Valve design 139
pulse rate and stroke volume is a function of both the leaflet dynamics and the geometric
orifice area of the valve. Geometric orifice area and internal orifice diameter of a valve
are functions of valve design and position. However, the influence of valve diameter in
determining proper function is often overlooked, due to the lack of consistency in sizing
the prosthesis and annulus (Diab et al., 2011; Ruzicka et al., 2009).
International standards dictate that the labelled size of a valve should take a patient-
centred approach. This requires that the labelled size is equivalent to the measured tissue
annulus diameter (TAD), for example, a “size 23” valve will fit a patient with a 23 mm
TAD (International Standards Organisation, 2005).
Figure 5.2: Definition of dimensions for three types of valve positioning: intraannular,
intra-supraannular and complete supraannular. ESRD - external sewing ring diameter,
IOD - internal orifice diameter, and TAD - tissue or aortic annulus diameter.
Although the ISO definition is clear, there are several complicating factors. The tissue
annulus diameter is defined as the minimum aortic diameter of the aortic root. During
imaging measurements, the vessels grow and contract with pressure, leading to a diameter
variation of at least 10% (though this decreases with age) (Kalath et al., 1986). To
compound the problem, conventional echocardiographic measurements are limited by
their 2 dimensional nature, which means that the resulting image may lie out of plane
with the annulus centre resulting in underestimation of the TAD (Cerillo et al., 2012).
During surgery, it should be possible to make static, direct measurements. However,
though the aortic annulus is often spoken of as if there is a distinct ring, in practise
this is far from the case, and simple measurements presuming a uniform diameter can be
unreliable (Ho, 2009). The most consistent method of measurement should use flat-head
candles, as opposed to Hegar dilators, which may overestimate the TAD (Azarnoush
et al., 2013).
140 Chapter 5. Valve design
The sewing ring is used to secure a surgical valve in position. In bioprosthetic and
mechanical valves, the sewing ring is typically fabricated from a Dacron® or Gore-Tex®
mesh. As described in Section 1.8, interfaces between different materials can often be
the weakest link in a valve. The delamination at the interface in the valve designed by
Gallocher et al. (2006) was the cause of valve failure. If possible, it might be preferable
to fabricate the sewing ring and stent cylinder from the same material to avoid joins.
Stasiak et al. (2014) considered the suitability of oriented styrenic BCPs 10× 5× 0.35 mm
for use in the sewing ring by passing a single suture through the sheet, 2 mm from its
edge. When polystyrene cylinders were oriented parallel to the direction of strain, sam-
ples of SI-BS19 could withstand 3 MPa and 500% strain; when oriented perpendicular,
they also supported 3 MPa, but reached 700% strain. A fully sutured sewing ring made
from SI-BS19 could support 75 N. A valve must be able to support approximately 7.5 N.
The strength of styrenic BCP is comparable to other clinically used materials; chemi-
cally treated, sutured, calf pericardium which is used in cardiac bioprostheses, is able to
withstand 2.6 to 7.6 MPa, depending on treatment method (Garc´ıa Pa´ez et al., 2000).
5.1.4 Leaflet design
In Chapter 2 I ignored leaflet shape in my material performance index. Although leaflet
shape may only have higher order effects on the material performance index, it is still an
important part of the overall valve performance. The first descriptions of native leaflet
geometry were made over 2300 years ago, yet they continue to be refined. The shape
of a leaflet can be described using a combination of 3D surfaces, generally consisting
of cylinders, cones, spheres and other real quadric surfaces, or with other mathematical
forms. I searched the literature for descriptions of the native valve shape, and proposals
of more effective shapes. The results of this search are summarised in Table 5.3. For the
nomenclature used to describe the parts of the leaflet, refer to Figure 1.3a.
C
h
ap
ter
5.
V
alve
d
esign
141
Table 5.3: Surfaces used to describe the shape of a trileaflet aortic valve.
Shape
Notes (native description
or proposed design)
Source
Spherical portion - semilunar Native
Philistion of Locir (c.427-c.347 BC) in
(Bezuidenhout and Zilla, 2014)
Sigmoid Native
Erasistratus of Ceos (c.304-c.250 BC) in
(Bezuidenhout and Zilla, 2014)
Hemispheres Native
Leonardo da Vinci (1452-1519) in
(Bezuidenhout and Zilla, 2014)
3 interlocking circles Native
Retzius (1843) in (Bezuidenhout and Zilla,
2014)
‘Cone’ shaped and ‘dome’ shaped (Ghista and Reul, 1983; Roe et al., 1966)
Paraboloids of revolution with foci at
leaflet base
(Leat and Fisher, 1994; Mercer et al., 1973)
Semi-dome shaped / triangular Proposed polymeric (Mohri et al., 1973)
2 radii of curvature over 2 subtended
angles (general Gaussian)
Native Chong et al. (1973),Kim et al. (1991)
Elliptic paraboloid Bioprosthetic (Hamid et al., 1986)
‘Jellyfish’ perforated disk with
polymeric flaps
Polymeric, not a trileaflet (Imachi et al., 1989)
142
C
h
ap
ter
5.
V
alve
d
esign
Table 5.3 continued.
Shape
Notes (native description
or proposed design)
Source
Cylindrical form, conical when open
(see Figure 5.3)
Native, then widely used
in pericardial valves.
(Labrosse et al., 2006; Thubrikar and
Klemchuk, 1990)
Determined by fractal structure of
fibres
Native (Peskin and McQueen, 1994)
Alpharabola Polymeric (Leat and Fisher, 1994)
Conic
Approximated bovine
pericardial, to make
polymeric
(Mackay et al., 1996)
Spherico-hyperbolic Polymeric (Butterfield et al., 2001)
Elliptico-hyperbolic conicoid Polymeric (Butterfield et al., 2001)
Offset hyperboloid of revolution
Native, to make
polymeric valve.
(Jiang et al., 2004)
Set of parabolae
Polymeric, used for
Aortech valve.
(Beith, 2006)
Bezier curves with de Casteljau
algorithm
Polymeric, by in silico
optimisation.
(Mohammadi et al., 2009; Mohammadi and
Mequanint, 2011)
3 intersecting cylinders to define each
leaflet
Polymeric, by in silico
optimisation.
(Burriesci et al., 2010)
C
h
ap
ter
5.
V
alve
d
esign
143
Table 5.3 continued.
Shape
Notes (native description
or proposed design)
Source
Hemicylindrical Proposed polymeric.
Wisman et al. in Bezuidenhout and Zilla
(2014)
Translated hyperbolic curve swept
through quartic radial curve
Polymeric, by in silico
optimisation.
(Gharaie and Morsi, 2015)
144 Chapter 5. Valve design
In Table 5.3, the valve shapes evolve over time from descriptions of the native valve,
through to proposed polymeric trileaflet designs and their optimisation. The most clear
distinction that can be made is between leaflets which are purposefully designed to have
a single degree of curvature (Gaussian curvature = 0) when open, versus those with
positive Gaussian curvature. Notably, the shape described by Thubrikar and Klemchuk
(1990) is designed as a single section of a cylinder when open, thus it can be cut from
a flat sheet and mounted to an appropriate stent (Figure 5.3). The energy required to
reverse the curvature of a cylinder is less than that required for a sphere, which should
reduce the pressure and time required to open a cylindrical form valve. On the other
hand, diastolic stresses are reported to be lower in spherically shaped valves (Liu et al.,
2007b). To date, few designs have been tested against one another in vitro, making direct
comparisons difficult. It is certain that material and form are intrinsically linked. Thus,
if the material is not perfectly biomimetic, a perfectly biomimetic form is unlikely to be
necessary to create an optimal valve.
HS
H
Rb
RC
α
Figure 5.3: Open form of the valve designed by Thubrikar and Klemchuk (1990). Each
leaflet can be cut from a single, flat sheet, and can be described by 5 parameters.
The effect that leaflet design has upon durability and haemodynamics of the valve are
summarised in Table 5.4 by the type of feature. In general, a change in any feature results
in opposite effects on durability and haemodynamics. Each of the 5 features identified
in Table 5.4 could be a parameter used to optimise the valve’s performance. Translation
of each of these features to quantitative parameters would result in a large number of
variables, and an optimisation problem far beyond the scope of this chapter.
The Thubrikar and Klemchuk (1990) shape is attractive due to its simple cylindrical
geometric surface when open, meaning that it should require a reduced pressure to reverse
its curvature and can be formed from a flat sheet. However, the pure cylinder is only
Chapter 5. Valve design 145
the case for the open valve, and upon closure it forms a coaptation region which is
not cylindrical. This coaptation region is necessary for proper functioning of the valve;
without a significant coaptation area the valve will not form a good seal (Loerakker et al.,
2013; Thubrikar and Klemchuk, 1990).
Table 5.4: Design features for heart valve leaflets.
Feature Effect of parameter increase
on valve stresses
Effect of parameter increase on
haemodynamics
Leaflet thickness Reduced stresses Inferior haemodynamics
(Bernacca et al., 2002; Rah-
mani et al., 2012).
Stent height / pro-
file
Reduced valve stresses (par-
ticularly at commissures)
(Chandran et al., 1991;
Hamid et al., 1986).
Inferior haemodynamics, pos-
sible interference with nearby
structures (Chandran et al.,
1991). Reduced likelihood
of prolapse (Labrosse et al.,
2006).
Stent flexibility Reduced valve stresses at
commissures (Chandran
et al., 1991; Christie and
Barratt-Boyes, 1991), but
can cause leaflet bunching if
severe (Vesely, 2003).
Problematic if asymmetry
causes prolapse of a single
leaflet (Vesely, 2003).
Depth/sag of belly
and free edge
Reduced stresses (Vesely,
2003)
Cylindrical valves (zero belly)
require less energy to reverse
curvature (Liu et al., 2007b).
Coaptation area Can increase radial strain
(Loerakker et al., 2013).
Reduced regurgitation.
Angle/distance be-
tween leaflets at
commissure
Opens more easily. Closing and
sealing more difficult.
A polymeric valve can be manufactured in any state between fully open and fully closed.
An advantage of forming a valve in a partially open state (Rottenberg et al., 2000), is to
reduce opening forces and flexural strain between the open and closed positions (Jansen
et al., 1991). However, this can lead to poor closing characteristics; Leo et al. (2006)
performed hydrodynamic and PIV testing of the Aortech valves which had failed in vivo
due to thrombus formation at the commissure. These valves were formed with various
leaflet thicknesses and degrees of openness. While thickness dominated the performance,
the neutral open state of the valve also led to regurgitant flows during closing which
146 Chapter 5. Valve design
contributed to the high shear, thrombus-inducing regions at the commissures. As such,
the valve should be formed with a neutral state near to the closed position, and designed
to have reasonable leaflet coaptation during diastole. This is not possible using a single
cylindrical surface as proposed by Thubrikar and Klemchuk (1990), which lead Burri-
esci et al. (2010) to propose, and optimise, an alternative geometry based purely upon
cylindrical surfaces.
Burriesci et al. (2010) performed in silico optimisation and in vitro evaluation of their
valve formed from cylindrical surfaces, with reference to the much tested and published
Aortech valve (Wheatley et al., 1998). The optimisation considered five geometric pa-
rameters, though it did not include leaflet thickness, which was uniformly set to 100µm,
and used a linear material with a modulus of 45 N mm−2. The optimised valve performed
significantly better than the reference valve in hydrodynamic tests in measures of pres-
sure drop, EOA, energy loss, regurgitation, and leakage. The von Mises stresses during
diastole were also lower for the optimised valve. The optimised cylindrical valve shape
which was proposed, has gone on to be used in the Triskele valve (which is currently in
animal trials). Based upon this study I selected this leaflet shape, which I hereafter refer
to as the cylindrical form valve. Using a 3D CAD package I reproduced the shape and
added a stent and sewing ring to produce the valve shown in Figure 5.4, .
Figure 5.4: Trial prosthetic heart valve design, the “cylindrical form valve”, based upon
the leaflet form proposed by Burriesci et al. (2010).
I compared the cylindrical form valve to the design proposed by Zaffora (2011). This
design was based upon a leaflet belly with a spherical form, and a commissure/leaflet-stent
boundary region with a cylindrical form, whose construction is similar to that described
in Section 5.2.1. A fibre orientation optimisation had been previously performed on
this valve (Serrani et al. Submitted). Prior to commencing this project the Structure
Materials group at the University of Cambridge and Politecnico di Milano produced a
prototype valve based upon the design of Zaffora (2011). This preliminary geometry is
Chapter 5. Valve design 147
shown in Figure 5.5. This valve was characterised by having a particularly high profile,
which resulted in poor hydrodynamics during in vitro testing. A new design, with a
lower profile, was proposed (Figure 5.5). This leaflet form is hereafter referred to as the
spherical form valve.
The cylindrical and spherical form valves will now be manufactured, tested, and com-
pared.
Figure 5.5: The original and updated spherical form valve designs. The valve on the left
is the initial valve, which has a high profile. The valve on the right is the updated low
profile valve design. The red line indicates indicates the height of the nearest commissure
in the updated valve. Both valves were fabricated by compression moulding and have
internal diameters of 22 mm.
5.2 Materials and methods
5.2.1 Valve design
The cylindrical and spherical form valves were designed using Rhino3D® (McNeel, Seat-
tle, WA, USA), a Non-Uniform Rational B-Splines (NURBS) based CAD package. The
visual programming add-on package Grasshopper® was used for parametric manipulation
of the shapes.
Cylindrical form valve
A valve based entirely upon cylindrical surfaces, as described by Burriesci et al. (2010)
was used as a shortlisted design. Dimensions optimised by the authors are indicated in
Figure 5.6. The height of the valve is 15 mm and diameter 22 mm. The bulk of the leaflet
148 Chapter 5. Valve design
is defined by a cylinder of diameter 22 mm inclined at 33.4° to the vertical. The coaption
region is defined by two cylinders also of diameter 22 mm inclined at 10.9° to the vertical.
The intersection between the coaption and main cylinders lies on a plane 0.4 mm from
the leaflet reflection plane and at an angle of 0.63°. The stent cylinder and posts were
designed with a thickness of 2.5 mm, as shown in Figure 5.4. The leaflet thickness was
set to 0.3 mm.
Figure 5.6: Construction dimensions for the ventricular surface of the cylindrical form
valve, based upon the leaflet optimised by Burriesci et al. (2010). The surface is con-
structed with cylinders of 22 mm diameter, which is equivalent to the valve’s internal
diameter.
Spherical form valve
The spherical form valve was proposed by Zaffora (2011) (Figure 5.7). A 120° segment
of a hemisphere is extruded radially, to form a spherical belly and cylindrical sides (for
more details see Appendix C.1). The spherical and cylindrical diameters used for con-
struction were set to 22 mm. Leaflet thickness was set to 0.3 mm; and the commissural
distance (spacing between aortic surfaces of the leaflets, dcom) was minimised to 0.7 mm,
with manufacturing constraints providing the minimum thickness. At the commissures,
the ventricular sides of the leaflets are separated by 0.1 mm. The form of this leaflet
approximates that of a standard bioprosthetic valve.
Chapter 5. Valve design 149
Figure 5.7: Construction dimensions for ventricular surface of the spherical form valve,
based upon the leaflet proposed by Zaffora (2011). The valve has an internal diameter
of 22 mm. A 120° segment of a 22 mm diameter hemisphere (red) is extruded radially, to
form the cylindrical sides (blue). The ventricular sides of the leaflets are separated by
0.1 mm at the commissures.
5.2.2 Prototype valve fabrication
Valve performance is dependent on the triad of valve design, material, and manufacturing.
As the aim of this chapter is to select a valve design for final manufacture, I chose
to use the previously selected material and a preferred means of manufacture during
selection. In Chapter 2 I selected polystyrene-block -polyethylene-polypropylene-block -
polystyrene with 22wt% of polystyrene (SEPS22). Injection moulding is commonly used
to process thermoplastics and so a bench-top injection moulding process was used to form
the prototype valves.
The geometry of the mould to make the spherical form valve was relatively straightforward
to machine. The male part was CNC machined from stainless steel, and the female part
spark eroded (Cambridge Reactor Design Ltd., Cambridgeshire, UK). The geometry of
the mould for the cylindrical form valve was more complex. The prototype tool was
manufactured by rapid additive manufacturing, involving direct metal laser sintering
(EOSINT M280) with an aluminium alloy AlSi10Mg, by 3T RPD Ltd. (Berkshire, UK).
Post-processing involved the addition of mould alignment pins and smoothing the mould
surfaces using a spray paint that was applied by brush.
A bench-top injection moulding apparatus was used to fill the mould cavities. Polymer
pellets were manually compacted after heating in an oven at 160 ◦C for 10 minutes. The
compacted pellets were loaded into the injection moulder’s barrel (φ 25 mm, 100 mm
150 Chapter 5. Valve design
length). The polymer melt passes through 30 mm of connecting tubing before entering
the die. The die is mounted in a purpose-built frame with a 4-point hand clamping system
to clamp the tool without polymer leakage. The barrel and mould were heated with band
heaters and the tube with flexible cable heaters to 160 ◦C. Pressure was recorded at the
barrel exit to indicate when filling of the barrel was complete.
Rigid stent
After hydrodynamic testing was completed, the valves were modified by placing small
holes at the top of the stent posts through which a rigid metal ring was passed. The
metal ring eliminated bending of the stent during diastole. The performance of the
valves was reassessed in a second set of hydrodynamic tests.
Leaflet thickness variation
In another set of hydrodynamic tests the leaflet thickness was modified. The average
leaflet thickness was varied by applying layers of spray-on paint to the male and female
moulds to reduce the thickness of the cavity.
5.2.3 Hydrodynamic performance
The hydrodynamic performances of the valves were assessed using a pulse duplicator.
This consists of a servo-electrically controlled pump, driving a membrane which imitated
the ventricular wall. As illustrated in Figure 5.8, valves were mounted at the inlet and
exit of the ventricular chamber. The test valve was placed at the exit and a bileaflet
mechanical valve at the inlet. The exiting fluid passes through resistance, compliance,
and reservoir units before completing the loop. The system is described in detail by
Lanzarone et al. (2009). The servo-electric drive was programmed such that the systolic
flow rate followed the Swanson and Clark waveform (Swanson and Clark, 1977), and the
diastolic flow rate was reproduced by a modified Talukder and Reul flow rate waveform
(Talukder and Reul, 1978).
The valves are mounted in straight perspex tubing, of diameter 35 mm.
Pressure measurements are made 35 mm upstream and 105 mm downstream from the
test valve (140PC pressure sensors, Honeywell Inc.®,Freeport, IL, USA). Ultrasonic mea-
surements of the flow rate were made 70 mm before the test valve using a transit-time
ultrasound flowmeter (HT110 series, Transonic Systems Inc.®, Ithaca, NY, USA). Data
Chapter 5. Valve design 151
C
R R
Aortic valve
Mitral valve
Reservoir
Diaphragm
Reciprocating piston
Ventricular
chamber
PV PA
Figure 5.8: Schematic of the pulse duplicator set up. The resistance and compliance units
are indicated by ‘R’ and ‘C’, respectively. Pressure measurements were made at PV and
PA.
was recorded at a frequency of 200 Hz using LabView, and analysed using MATLAB
(Mathworks, MA, USA).
Tests were made in accordance with ISO 5840. Measurements to assess EOAG are carried
out at 120 mmHg, 70 BPM at flow rates of 2, 3.5, 5, and 7 l min−1. Regurgitation is
assessed with diastolic pressure differences of 80, 120 and 160 mmHg, at 45, 70 and 120
BPM.
Leaflet dynamics were viewed axially downstream from the valve. Images were collected
at 25 Hz.
5.2.4 Finite element modelling
The geometries were updated after the hydrodynamic tests for the final valve design.
Finite element modelling was used to compare the peak leaflet stresses in the valve before
and after the update. The models were run by Dr Marta Serrani using a previously
validated model.
A uniform pressure load was applied on the aortic surface of the leaflet up to 135 mmHg
to simulate a physiological backpressure during the closing phase. A quasi-static loading
condition was assumed (Haj-Ali et al., 2008; Luo et al., 2003; Sun et al., 2005); the
nonlinear implicit finite element algorithm provided by ABAQUS was used to solve the
numerical problem. The hyperelastic Mooney-Rivlin constitutive model was used to
model the polymer, and is described in detail in Appendix E. The isotropic material
152 Chapter 5. Valve design
parameters which are shown in Table 5.5 were based upon the mechanical properties of
SEPS22 which were collected in Chapter 2.
Table 5.5: Mooney-Rivlin parameters optimised for the tensile strain energy density of
SEPS22.
c1 c2 k4
Isotropic 0.3 0.36 n/a
5.3 Results
Pulsatile testing was performed on both valves at a range of cardiac outputs, diastolic
pressure drops, and pulse rates. Figure 5.9 shows that the EOAG of the cylindrical
form valves was greater than the spherical form valves, except for at a cardiac output of
7 l min−1. The EOAG of both valves increased with cardiac output, as the leaflets are
opened further by higher flow rate.
1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Cardiac output (l min−1)
E
O
A
G
(c
m
2
)
Cylindrical form
Spherical form
Figure 5.9: EOAG measurements of both valve forms at a range of cardiac outputs. The
cylindrical form valves have a greater effective orifice area at low cardiac outputs, but at
7 l min−1 the spherical valve is superior.
Cylindrical form valves were more regurgitant at all pulse rates and diastolic pressure
drops (Figure 5.10). In general, spherical form valve regurgitation increased with in-
Chapter 5. Valve design 153
creasing diastolic pressure, suggesting that there may be some leakage through the closed
valve. Regurgitation through the cylindrical form valve remained constant with increas-
ing diastolic pressure drops, and analysis of the flow profile indicated that regurgitation
is dominated by flow during closing.
60 70 80 90 100 110 120 130 140 150 160 170 180
0
2
4
6
8
10
12
14
16
18
20
Diastolic pressure drop (mmHg)
R
eg
u
rg
it
at
io
n
(%
)
Sphere 45 bpm Sphere 70 bpm
Sphere 120 bpm Cylinder 45 bpm
Cylinder 70 bpm Cylinder 120 bpm
Figure 5.10: The spherical form valves are less regurgitant than the cylindrical form
valves at all diastolic pressure drops.
Although energy loss is rarely used in the clinic, it is a valuable measure for making proper
comparisons between valves. In the data presented here, the cylindrical form valve had
a superior EOAG, but the spherical form valve was superior in terms of regurgitation.
Figure 5.11 shows energy loss, which reconciles EOA and regurgitation to yield an overall
performance measure. The spherical valve had a lower energy loss than the cylindrical
form valve at all flow rates except at 3.5 l min−1. Despite the spherical valve fulfilling
the ISO standard (evaluated at 5 l min−1), 17% of energy imparted on the fluid by the
ventricle was lost passing through the valve. At less efficient flow rates, up to 30% of the
energy was lost.
Energy is lost over the valve during forward flow and when retrograde flow is resisted.
In Figure 5.12 the relative contributions of systole and diastole to energy loss are shown.
Clearly, the high energy loss of the cylindrical form valve at low flow rates is associated
with regurgitation (dashed red line). The cylindrical form valve functions less efficiently
during systole at high flow rates, suggesting that the valve is behaving differently which
could be due to a laminar turbulent transition. For the spherical form, valve diastole and
154 Chapter 5. Valve design
1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8
10
15
20
25
30
35
Cardiac output (l min−1)
F
ra
ct
io
n
al
en
er
gy
lo
ss
(%
)
Cylinder (n=5)
Sphere (n=3)
Figure 5.11: Energy loss over both types of valve tested in the pulsatile flow. There is
a greater fraction of energy lost over the cylindrical form valve than over the spherical
form valve at all flow rates except 3.5 l min−1. The valves are most effective at 5 l min−1.
systole contribute equally to the inefficiencies of the valve.
1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8
0
50
100
150
200
250
Cardiac output (l min−1)
E
n
er
gy
lo
ss
p
er
cy
cl
e
(m
J
)
Regurgitant energy loss - cylinder Regurgitant energy loss - sphere
Systolic energy loss - cylinder Systolic energy loss - sphere
Figure 5.12: Energy loss per cycle for regurgitant (dashed lines) and systolic (solid lines)
cases as a function of cardiac output for both forms of valve. The inefficiencies of heart
valves occur during both diastole and systole. Energy loss in the cylindrical form valve
(red lines) is dominated by the poor closing behaviour, whilst in the spherical valve (blue
lines) the energy loss is equally split between systole and diastole over the range of flow
rates.
Images of the leaflet dynamics of the valves during systole are shown in Figures 5.13 and
5.14. The cylindrical form valve, at 5 l min−1, is more stable when open (images 3 to 5).
In contrast, the spherical form valve appears to be unstable in the open phase, and is
not able to fully reverse the curvature of two of its leaflets. The fully open (lower) leaflet
of the spherical valve also seems to have a “tongue”, where the free edge opens further
than the leaflet belly. This may indicate that, rather than fully reversing, the sphere has
Chapter 5. Valve design 155
formed a saddle.
1 2 3
4 5 6
7 8 9
Figure 5.13: Images taken (at 25 Hz) during the systolic phase for a cylindrical form valve
(5 l min−1, 70 BPM).
The spherical form valve during diastole is shown in Figure 5.15 to demonstrate the
deformation of the stent posts (in images 2 to 10). The bending causes the leaflets to
twist at the free edge centre.
5.3.1 Variation in leaflet thickness
I modified the spherical leaflet form mould to produce a selection of valves with differing
leaflet thicknesses. It was not possible to achieve large uniform changes in leaflet thick-
ness, and I could only achieve a 50µm difference between the maximum and minimum
mean thicknesses. Nonetheless, there was a correlation between leaflet thickness and the
maximum pressure required to open the valve, as shown in Figure 5.16a. There were no
correlations between mean leaflet thickness and other important performance parameters,
such as effective orifice area (EOAG) and energy loss (Eloss); however the sample size was
156 Chapter 5. Valve design
1 2 3
4 5 6
7 8 9
Figure 5.14: Images taken (at 25 Hz) during the systolic phase, for a spherical form valve
(5 l min−1, 70 BPM). There appears to be a flow instability through the orifice causing
the leaflet to oscillate between images.
limited. Transvalvular pressure gradient (TPGmax) varied from 21 to 40 mmHg without
a significant effect on EOAG, suggesting that all valves reached a similar geometric orifice
area (GOA) once the leaflet curvature had reversed (Garcia et al., 2004).
The variation in thickness between leaflets meant that the leaflets did not open simulta-
neously. Furthermore, at low cardiac outputs one, and occasionally two, leaflets did not
open. All three leaflets should open at low, as well as high, cardiac outputs. Figure 5.16b
shows that leaflets with a thickness greater than 325 µm may not open at low cardiac
outputs. This provides an upper bound for leaflet thickness for this form of valve when
made with SEPS22. However, it may be necessary to further decrease leaflet thickness
for all 3 leaflets to open at all cardiac outputs.
The upshot of the non-uniform changes in leaflet thickness was that it was possible
to observe the effect of spatial variations in leaflet thickness on transvalvular opening
pressure. Figure 5.17 shows the results of a covariance analysis on the spatially variant
leaflet thickness measurements and TPGmax.
Chapter 5. Valve design 157
1 2 3 4
5 6 7 8
9 10 11 12
13 14 15 16
Figure 5.15: Images taken (at 25 Hz) during the diastolic phase, for a spherical form
valve. (5 l min−1, 70 BPM).
Covariance, σcov, is a marker of the degree of linear dependence between N random
variables, calculated as:
σcov(X, Y ) =
N∑
i=1
(xi − x¯)(yi − y¯)
N − 1 (5.6)
Where xi and yi are observations of parameters X and Y . x¯ and y¯ are the mean values. In
the covariance matrix, the diagonal components, i = j, are the variance of the variables,
and the off-diagonal components, i 6= j, are the covariances relating one variable to
another. If the X = Y then σcov(X, Y ) = σ
2
X , where σ is variance. A higher covariance
indicates a higher degree of correlation between the two variables, where one is perfect
correlation. X was taken to be TPGmax, and Y to be the leaflet thickness measurements.
Figure 5.17b shows the covariance value for each location on the leaflet. A large, positive
value (≤ 1) indicates a positive correlation between thickness and TPGmax. The thick-
ness of the free edge near to the commissures is the most important leaflet dimension
during opening. The images of the spherical valve when open (Figure 5.14) support this
158 Chapter 5. Valve design
260 280 300 320 340 360 380 400
0
10
20
30
40
50
Leaflet thickness (µm)
T
P
G
m
a
x
(m
m
H
g
)
(a)
250 275 300 325 350 375 400
2
4
6
8
Average leaflet thickness (µm)
M
in
im
u
m
C
O
lm
in
−1
1st leaflet
2nd leaflet
3rd leaflet
(b)
Figure 5.16: (a) The effect of leaflet thickness on the maximum transvalvular pressure
which is reached during opening of the spherical form valve. (b) The minimum cardiac
output (2, 3.5, 5, or 7 l min−1, at 120 mmHg) required to open the 1st, 2nd, and 3rd leaflets.
The thickness measurements are average values taken from 7 points on each leaflet (a)
and opening was determined from videos taken at each cardiac output. The inset shows
the case where two leaflets have opened, and one remained closed during systole.
proposition, as they show the fully open valve to have its tightest bend in the region in
which the “free edge left” and “free edge right” measurements were made. Importantly,
the highest diastolic stresses are also located near to the commissures, indicating that for
a uniformly thick leaflet, the leaflet thickness used in the material performance index in
Chapter 2 applies to this location.
The same analysis was performed for the cylindrical form valve, though there was no cor-
relation between leaflet thickness (which varied over 70 µm) and TPGmax, nor any positive
linear dependencies (covariants) between spatial thickness variation and TPGmax. The
cylindrical form leaflet has a greater GOA than the spherical form leaflet, as a result of
the leaflet curvature fully reversing. As such, the tightest radius of curvature when open
was found along the stent-leaflet boundary. As the thickness measurements were made
using a micrometer (of diameter 8 mm), they were not sensitive to small variations in
thickness along the stent-leaflet boundary.
The overall performance of the cylindrical form valve in terms of energy less was dom-
inated by closing performance. A covariance analysis between leaflet spatial thickness
Chapter 5. Valve design 159
and regurgitation found no significant linear dependencies.
(a)
F
E
le
ft
F
E
ce
nt
re
F
E
ri
gh
t
SE
ri
gh
t
N
ad
ir
SE
le
ft
B
el
ly
−0.1
−0.05
0
0.05
0.1
0.15
0.2
0.25
FE
lef
t
FE
ce
nt
re
FE
rig
ht
S
rig
ht
Na
di
r
SE
lef
t
Be
lly
σ
co
v
(b)
Figure 5.17: (a) shows the locations of measurements for spatial variation in the thickness
of heart valve leaflet. (b) shows the covariance matrix values (i 6= j) of leaflet thickness
and TPGmax (5 valves, 15 leaflets). FE refers to the free edge in this figure, and SE to
the stent-leaflet boundary.
5.3.2 Stent post stiffness
During diastole the posts of all valves were deflected, leading to an increase in coaptation
between the leaflets, but also causing them to assume a very different geometry. To
address this issue I fabricated a new valve in which the stent posts were radially fixed
using a metal ring. This eliminated the deflection of the posts, as shown in Figure 5.18.
Stent post fixing did not have any influence upon the EOAG of the valves (Figure 5.19),
but led to a large improvement in regurgitation, as show in Figure 5.20. It is unlikely
that the ring significantly influenced the flow field downstream of the valve.
A summary of the results, relevant to the fulfilment of the ISO standards, is shown in
Table 5.6. All valves fulfilled the ISO standards for EOAG of being greater than 1 m
2.
With respect to the limit for regurgitation, the cylindrical form valve was only marginally
acceptable once the stent was stiffened (which was accompanied by a 9% drop in EOAG).
The spherical valve was acceptable with both flexible and rigid stent posts. These results
can be directly compared to a clinically available bioprosthetic reference valve which was
previously tested. The bioprosthesis was superior to the polymeric prostheses in measures
160 Chapter 5. Valve design
Figure 5.18: Five images during diastole of the spherical form valve with fixed stent
posts at 5 l min−1 and 70 BPM. There is no difference in the position of the commissures
between the images, indicating that there was no post deflection. The ring appears as a
silver circle.
1 2 3 4 5 6 7 8
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Cardiac output (l min−1)
E
O
A
G
(c
m
2
)
Sphere - Flexible Sphere - Fixed
Cylindrical - Flexible Cylindrical - Fixed
Figure 5.19: Effect of limiting radial deflection on the EOAG of spherical and cylindrical
form valves.
of both EOAG and regurgitation. However the orifice area may partly be aided by the
slightly larger internal diameter of the valve (23 mm vs. 22 mm), though the valve was
mounted with a 22 mm collar such that the geometric orifice area when open was equal.
Chapter 5. Valve design 161
60 80 100 120 140 160 180
0
2
4
6
8
10
12
14
16
18
20
Diastolic pressure drop (mmHg)
R
eg
u
rg
it
at
io
n
(%
)
Fixed 120 bpm Fixed 45 bpm
Fixed 70 bpm Flex 120 bpm
Flex 45 bpm Flex 70 bpm
(a)
60 80 100 120 140 160 180
0
2
4
6
8
10
12
14
16
18
20
Diastolic pressure drop (mmHg)
R
eg
u
rg
it
at
io
n
(%
)
Flex 120 bpm Flex 45 bpm
Flex 70 bpm Fixed 120 bpm
Fixed 45 bpm Fixed 70 bpm
(b)
Figure 5.20: Regurgitation of the spherical (a) and cylindrical (b) form valves, with
(solid) and without (dashed) radial fixation of the stent posts, during pulsatile testing.
The valves were tested at three pulse rates (red, green, blue) and at three diastolic pressure
gradients. The radial fixation of the stent leads to significantly less regurgitation for both
valve forms.
Table 5.6: Summary of hydrodynamic data for the tested valves, under the ISO 5840
standard conditions of 5 l min−1, 70 BPM, 120 mmHg. (n) indicates the number of valves
of each type that were tested.
Valve
TAD
(mm)
EOA (cm2)
Regurgitation
(%)
ISO5840 Standard (n) 23 ≥ 1.00 ≤ 10
Spherical form, unstented (4) 22 1.36 ± 0.05 9.32 ± 0.12
Spherical form, fixed stent (2) 22 1.27 ± 0.07 7.45 ± 0.14
Cylindrical form, unstented (6) 22 1.41 ± 0.08 17.8 ± 0.13
Cylindrical form, fixed stent (2) 22 1.28 ± 0.10 9.71 ± 0.29
Bioprosthesis - 3000 PERIMOUNT
Magna Aortic 25 (previously tested in
system)
23a 1.63 4.66 ± 0.34
Bioprosthesis - 3000 PERIMOUNT
Magna Aortic 25 (in vivo data, (Botzen-
hardt et al., 2005))
21-23 2.09 ± 0.55 3.33 ± 1.15
amounted with 22 mm collar.
162 Chapter 5. Valve design
5.4 Discussion
I proposed two designs of polymeric prosthetic heart valve. One is characterised by being
formed from a spherical surface, the other from cylindrical surfaces. In hydrodynamic
tests the cylindrical form valve achieved a greater EOAG, to the detriment of regurgi-
tation. The cylindrical form valve was based upon the design of Burriesci et al. (2010)
who optimised the design in silico, and verified in vitro, against the Aortech valve. In
their in vitro hydrodynamic tests, using valves of equivalent diameter, at 5 l min−1, 70
BPM and 100 mmHg mean arterial pressure, the EOAG of the design was 3.34 cm
2 and
regurgitation 10.8%, with a 12.1% energy loss. In my experiments, with a flexible stent
I recorded an EOAG of 1.4 cm
2, and once the stent was fixed radially, regurgitation was
reduced to 9.7 %. The performance of the prototype valve tested here is inferior to that
tested by Burriesci et al. (2010) due to the prototype’s higher energy loss. The large
disparity may be attributed to a combination of differences (this report vs. Burriesci
et al. (2010)):
• Testing systems (custom tester vs. Vivitro commercial tester),
• Mode of manufacture (injection moulding vs. dip casting),
• Leaflet thickness (300 µm vs. 100 µm),
• Material (SEPS22 (modulus ∼ 8 MPa) vs. silicone (modulus ∼ 45 MPa)).
Despite these differences, it is interesting to refer the final two points back to the hypoth-
esis made in Section 2.1.5 and Equation (2.2). I related leaflet thickness and modulus to
the force required to open a leaflet. In summary, Pmax
t3E
is equal to some constant Kv. For
the stiffer, thinner silicone valve, Kv is 400 mmHg/(N mm), while for the softer, thicker,
SEPS22 valve, Kv is 300±145 mmHg/(N mm). These are remarkably similar given the
approximations that have been made, indicating that the assumptions made in Section
2.1.5 were reasonable.
The superior EOAG of the cylindrical form valve came at the expense of unacceptable
regurgitation. Regurgitation was dominated by retrograde flow during closing, rather
than leaking. This suggests that the valve had adequate coaptation between leaflets, but
was not able to close quickly enough. In vivo, closing is initiated by the vortices in the
sinuses of Valsalva (Peacock, 1990; Tien et al., 2014; Van Steenhoven and Van Dongen,
1979), which were not present in this in vitro hydrodynamic testing system and could
influence closing dynamics. Contrary to the in vivo results, De Paulis et al. (2005) found
no change in valve behaviour when they compared valved conduits with and without
pseudo-sinuses in vitro.
Chapter 5. Valve design 163
The high regurgitation is likely to be connected to the neutral shape that the valve
assumes. When unloaded, the angle between the leaflets at the commissure (2θc) is 0° for
the spherical form because a finite space is required between the leaflets to form them
without having to separate them after moulding. In the cylindrical valve this angle can
be as much 12°. A small commissural angle has the disadvantage of reduced GOA and
increased leaflet stresses when open, but the advantage of reduced regurgitation. Leo
et al. (2005) performed a thorough velocity field analysis (using laser doppler velocimetry
and particle image velocimetry) of a mechanical valve and three versions of a prototype
Aortech polymeric valve. In vivo trials of the Aortech valve had failed with thrombi
forming on the inside of the stent post, and as such the investigators paid particular
attention to commissure design. The various Aortech prototypes were characterised by
having different θc. Although larger θc aided opening without detriment to regurgitation,
retrograde jets during closing were observed with laser doppler velocimetry, which may
have contributed to the in vivo failure of the design.
The spherical form valve, without modification, would fulfil the ISO 5840 criteria and
this was necessary for progression to an animal and clinical trial. The cylindrical valve
fulfilled the ISO criteria but only marginally once the valve had been modified. As such,
the spherical form valve was selected for further design iterations and testing.
My study has been valuable in demonstrating the advantages and shortcomings of a
valve manufactured from single material. The major shortcoming of the single material
valve was the excessive bending of the stent posts. In general, the quality of the valve
was good as only one valve broke during testing (due to a visible, pre-existing flaw).
The use of direct metal laser sintering to fabricate a mould tool for the production of
prototype, injection moulded, polymeric prosthetic heart valves is the first reported in the
literature. A functioning valve was produced. The surface was considerably rougher than
that of the stainless steel mould so some differences in performance may be attributed to
manufacturing. The rougher surface is most likely to affect durability, as flaws may grow
from the surface.
5.4.1 Design improvements
Regurgitation was much reduced when the stent posts were fixed radially. Furthermore,
significant deflection of the posts led to a markedly different leaflet conformation during
diastole, potentially making the optimised anisotropy irrelevant. Several options existed
to alleviate this issue:
• Increase the radial thickness of the stent.
164 Chapter 5. Valve design
• Increase the circumferential thickness of the stent.
• Increase the stiffness by:
– increasing the PS content (through multi-step injection moulding),
– using a stiffer material (and thus making a valve with material interfaces).
• Reduce the profile (height) of the valve.
• Suture the commissure or stent post to the aortic wall - like the Sorin Freedom Solo
(running suture) or Medtronic 3f (tab at commissure) bioprostheses.
Given the desire to produce a single material valve, which would not need material
interfaces or sutures, the first two options were initially pursued.
Increases in the thickness of the stent brought two design constraints into consideration:
potential interference with the coronary arteries; and a difference between the external
sewing ring diameter (ESRD) and TAD (=internal orifice diameter, IOD). I added the
maximum possible expansion to the stent post’s axial cross section, as shown in Figure
5.21. The construction steps are shown in Appendix C. The sewing ring extends around
the complete circumference of the valve, allowing sutures to be placed at all positions.
The valve continues to be made from a single material, including leaflets, stent, and
sewing ring.
In the hydrodynamic tester the sewing ring was rigidly fixed in place. This may be a
poor representation of the physiological case as during the cardiac cycle the nadirs and
commissures of the native valve are highly mobile. Dagum et al. (1999) placed radio
opaque markers at the commissures and nadirs of the aortic valve of a sheep. They
observed heterogeneous and asymmetric changes in shape. During diastole, 3.7, 10.3
and 11.1% diameter increases at the posterior, right and left sectors, respectively, were
observed. Up to 6° of torsion was also observed, and a diameter increase of between
16 and 40% at the sinotubular junction (Dagum et al., 1999; Thubrikar and Klemchuk,
1990). The authors also stress the importance of matching dynamics of valvular implants
with those of the aortic root. The use of SEPS22 for the stent posts and sewing ring may
allow the valve to flex with the aortic root, in a more physiological manner than current
prostheses with stiff stent posts and sewing rings.
The benefit of the change in stent post design was confirmed using a structural finite
element (FE) model as shown in Figure 5.22. The maximum stress moved from the
middle of the stent-leaflet boundary towards the commissure, and was also reduced by
46%. This reduction in maximum stress is beneficial because small amount of flexibility
Chapter 5. Valve design 165
Figure 5.21: CAD image of the updated valve design consisting of spherical form leaflets,
enlarged stent posts and a sewing ring for supra-annular implantation.
can significantly reduce the stresses in the leaflet during diastole, increasing the durability
of the valve (Krucinski et al., 1993).
(a) (b)
(c)
(d)
Figure 5.22: FE modelling of the old (a and c) and new (b and d) stent post designs
when loaded to 95 mmHg. Images (a) and (b) are viewed axially during diastole. Images
(c) and (d) show the von Mises stresses in a single leaflet of each valve. The new stent
post design reduces the deflection of the posts and deformation of the leaflets compared
to the old. Note that the stress concentrates (σmax = 1.15 MPa) near to the base of the
stent post in (c), but the stress is reduced (σmax = 0.62 MPa) and across the stent-leaflet
boundary in (d). Modelling and images courtesy of Dr Marta Serrani.
166 Chapter 5. Valve design
5.5 Conclusion
I fabricated and hydrodynamically tested two types of valve, referred to as spherical form
and cylindrical form, which had fundamentally different leaflet designs were evaluated.
The spherical form valve has a more conventional leaflet shape, akin to a bioprosthesis,
while the cylindrical form valve uses only cylindrical surfaces to reduce the pressure
required to reverse the leaflet’s curvature. I manufactured both valves from SEPS22 by
injection moulding. The valves were then tested in a hydrodynamic pulsatile tester. The
cylindrical valve had a greater EOAG but suffered from unacceptable regurgitation, due
to its slow closing. The spherical form valve had a lower EOAG, and less regurgitation,
and fulfilled the ISO standard on both counts. Both valves were inferior to a standard
bioprosthesis with respect to hydrodynamic performance.
The original valve designs suffered from excessive deflection of their stent posts and
commissures during diastole, which was improved when the posts were radially fixed in
place. An updated valve design was proposed in which the most notable feature was
an enlargement of the stent posts. Based upon FE modelling, this design resulted in
less deflection of the stent posts and commissures and a more stable deformation of
the leaflets. This design will be taken forward for final manufacture and testing in the
following chapters.
Chapter 6
Valve fabrication
6.1 Introduction
The aim of this chapter is make the valve with a bioinspired orientation. Polymers can
be shaped into nearly any 3D form, and some of the properties of the final product can
be very dependent on the processing of the source polymer. In this chapter I review the
methods which have been used to form polymer-based products, particularly polymeric
heart valves, and I consider alternative means for controlling the orientation of cylindrical
PS domains. I go on to propose a processing method and I experimentally evaluate its
ability to orient domains in a simple geometry. This leads to the development of a model
which can be used to design the process to achieve bioinspired orientation in the polymeric
heart valve leaflets. Finally, a means of processing and tool design is selected, and the
final valve is manufactured.
The most crucial part of a valve is the leaflets, in which the highest diastolic stresses can
be found at the leaflet-stent boundary. Bioinspired orientation can be used to reduce these
stresses and increase the durability of the valve. Zaffora (2011) optimised the orientation
of anisotropy in a polymeric leaflet. The distribution of the optimised orientation was
shown in Figure 1.12. In summary, near to the free edge the optimised orientation is
horizontal and parallel to the free edge. Moving towards the nadir, the oriented fibres
form arcs across the leaflet, with each arc following a similar shape to that of a cable
suspended between the stent-leaflet boundaries.
One criteria upon which polymers were shortlisted in Chapter 2 was on their rubbery
mechanical properties at room temperature and pressure. In order to permanently change
the shape of these polymers the conditions must be changed. If the polymer can be
167
168 Chapter 6. Valve fabrication
dissolved, it becomes a viscous solution for solvent, dip, or film casting. In film casting,
the solvent is evaporated from the polymer solution while on a flat plate to leave a thin
film of polymer. Alternatively, a shaped mandrel can be used, which is dipped into
solution, the solvent is evaporated, and the polymer is then detached from the mandrel.
Depending on the viscosity, mould shape and position, and atmospheric conditions, the
resultant thickness from a single dip can be changed. Repeat dips can then be used to
build up the desired thickness (Daebritz et al., 2004b). This process generally results
in very few flaws in the polymer, which contributes to the good durability results which
have been recorded for dip cast valves (Bezuidenhout and Zilla, 2014). However, it
is unattractive from a manufacturing standpoint due to long processing times, the use
of non-biocompatible solvents, and the difficulty of maintaining consistency (Leat and
Fisher, 1995). Polymer solutions can also be used to coat fabrics or other substrates;
SIBS-toluene solution was used to coat polyester fabric, which was trialled in vivo but
was unsuccessful due to SIBS-polyester dehiscence (Gallocher, 2007).
The absence of major flaws in solvent cast polymers is a significant advantage, so the
issues of consistency in leaflet thickness can be overcome by forming simple flat sheets.
Rather than using a shaped mandrel (in the desired shape), solvent cast films of a constant
thickness can be cut to the desired shape, and solvent bonded to a pre-fabricated frame.
Small changes can then be made to the leaflet geometry by subsequent thermo-forming.
Unfortunately the solvent bonding of the leaflet to the stent is prone to causing defects
which may lead to failure (Leat and Fisher, 1995).
Polymers that are hard to process using temperature or dissolution, ePTFE for example
(which cannot be solvent cast), can also be manipulated as flat sheets which are then
sutured to a stent. ePTFE valves tested in vivo had poor durability and calcified (Nistal
et al., 1990). Alternatively, one end of a cylinder can be split into three flaps which
are folded upon themselves, and sutured to form three leaflets within the cylinder. This
method saw successful clinical use in the less strenuous pulmonary position (Ando and
Takahashi, 2009).
Electrohydrodynamic hot jet plotting could provide a means of directly printing polymers
with exceptional resolution (<10 µm). However, the experience of the UCL group with
electrohydrodynamic atomisation suggests that this technology is yet to reach maturity
(Reade, 2009). The Fraunhofer Institute, in collaboration with the Helmholtz-Institute
for Biomedical Engineering of RWTH Aachen University, fabricated polycarbonate-urethane
venous valves using a dropwise method. Dissolved polymers, of varying hardness, de-
posited in drop volumes of 2 to 60 nl, then allowing the solvent to evaporate in order to
build up a complete valve (Bezuidenhout and Zilla, 2014).
Chapter 6. Valve fabrication 169
Alternatively, polymers can be manipulated at elevated temperatures. Compression
moulding requires a male (core) and female (cavity) mould, between which a block of
polymer is placed. The mould is then compressed, causing the molten polymer to fill the
cavity. The earliest silicone valves made by Roe et al. (1966) were compression moulded
at 180 ◦C, followed by a 4 h cure at 200 ◦C. Later designs used silicone in plaster moulds,
filled by evacuating air or centrifugation followed by room temperature curing (Chetta
and Lloyd, 1980). Claiborne et al. (2013b) also compression moulded uncrosslinked xSIBS
in an evacuated, five part aluminium mould at 260 ◦C, to initiate Diels-Alder crosslinking.
They were successful at producing valves, though further results - especially durability -
have not yet been reported for this valve. Whilst it is relatively easy to reliably reproduce
the temperature conditions during compression moulding, it is hard to control the flow
field within the cavity, and ensure complete filling of complex parts.
Injection moulding can overcome these limitations. Injection moulding involves the com-
pression and heating of polymer pellets in a barrel, before injection into a mould cavity.
The mould can then be split, the part ejected, and the process repeated. Injection mould-
ing tools can have intricate, accurate, and complex geometries, which are reproducibly
and rapidly filled without need for toxic solvents. Various patents suggest that injection
moulding has been used to produce polymeric valves (Bernacca et al., 2009; Bluestein
et al., 2013a; Campbell et al., 2004; Chinn, 2002; Pietsch et al., 1988), though few have
reported the results - if of course the valves were made that way. Mackay (1992) recog-
nised the advantages of injection moulding for polymeric valves, and fabricated moulds
for the Glasgow PU valve. Unfortunately they could not achieve consistently thin leaflets,
due to poor tooling and air bubbles in the mould, limiting the durability of the valve.
Sacristan et al. (2003) produced injection moulded silicone trileaflet valves in a conduit
for use in ventricular assist devices. This has certainly been a successful avenue of use for
injection moulded valves, as the process is very cost effective, and the durability limit of
these valves is rarely approached as the ventricular assist devices are only for short term
use.
6.1.1 Orientation of polystyrene domains
As well as processing the polymer into the form of the prosthesis, the aim of this thesis
is to orient the stiff polystyrene (PS) cylinders of the BCP into a bioinspired orientation.
The selected BCP self-assembles into cylinders at room temperature. During solvent
casting, entropic forces dominate, causing microphase separated cylinders to be randomly
oriented, however, it may be possible to seed the microphase separation, causing it to grow
170 Chapter 6. Valve fabrication
in a quasi-crystalline manner. In the same manner that a bubble of steam will form at a
defect in a pot of water, or a salt will crystallise at a flaw in a substrate, the formation of
microphase separations in BCPs can also be initiated. By chemically or topographically
patterning the surface of a BCP (solution) it is possible to seed a crystalline “flaw” from
which a regular microphase structure can grow. PS cylinder arrays in thin films of SBS
BCP can be oriented perpendicular or parallel to the surface, depending on the strength
of the surface field. Unfortunately, predictions and experiments can only propagate this
orientation up to approximately 10 (parallel oriented cylinder) layers from the surface
(Sakurai, 2008). The use of a selective solvent and fast evaporation can result in cylinders
which are well oriented perpendicular to the surface (Phillip et al., 2010). While these
effects continue to be useful in other applications, they are not a realistic means for
achieving orientation across the leaflets of a BCP prosthetic heart valve.
It is also possible to orient microdomains under the influence of an external electrical or
magnetic field. Microphase separation leads to spatial variation in the dielectric constant,
meaning that domains can order parallel to the electric field. However, the electric field
is limited in power due to dielectric breakdown. To overcome the electric field limitation,
a magnetic field can be used. However, magnetic fields are less efficient at orientation,
and so the lowest strength field which has successfully been used to orient microphases
was 1.8 T (Liedel et al., 2012; Sakurai, 2008).
The most widely studied method of microdomain orientation is using mechanical force
fields. When the temperature is above Tm (melting temperature) but below TODT (order-
disorder transition temperature) the domains can be moved and deformed. The use
of shear and extensional deformations to align microphase domains of BCPs was first
recorded by Keller et al. (1970). Cylindrical BCPs orient in the direction of exten-
sion when in polymer melts (Castelletto and Hamley, 2004; Fang et al., 2009; Stasiak
et al., 2010) and solutions (Hamley, 2000), though high strain rates (20 s−1) decrease
the quality of alignment (Carastan et al., 2013). Continuous shear leads to orientation
perpendicular to the shear gradient - in the direction of flow - while oscillatory shear
causes microdomains to orient perpendicular to the shear gradient, to parallel, and back
to perpendicular as frequency is increased (Chen et al., 1997; Morozov and Fraaije, 2002).
Injection moulding has the potential to reproducibly fabricate complex parts within the
necessary tolerance, and also leads to shear and extensional forces which can be used to
orient PS cylinders. Given these two points, I selected injection moulding as a means of
producing bioinspired, polymeric prosthetic heart valves.
A polymer injection moulding tool with the necessary tolerances to make a prosthetic
heart valve generally costs upwards of £5k, and so, it is unwise to design and develop a
Chapter 6. Valve fabrication 171
tool without a reasonable understanding of the polymer characteristics and mould flow
patterns. Thus I begin this chapter by describing polymers that I injected into a simple
test geometry to determine how shear and stretch influence the cylinder orientation.
Along with these observations, the melt rheology of the polymer was characterised, before
developing a model for the orientation of PS domains during moulding. The model is
validated in a valve test geometry and then the model is used to inform the design of the
final tool. Finally, a valve is injected and the orientation is evaluated.
6.2 Point injection
In an injection moulder used in production (as Figure 6.1) the following steps take place:
1. The solid polymer (in pellet, powder or crumb form) is compacted using a rotating
screw in the barrel.
2. The compacted polymer is melted as it advances along the heated barrel.
3. The screw stops rotating, and is rammed along the barrel, ejecting polymer from the
nozzle of the barrel, into the injection moulding tool.
4. Polymer passes through the sprue bushing, along a runner or runners, and through a
gate or gates, into the cavity.
5. If the tool has been heated intentionally or by the injected polymer, it is cooled and
the part solidifies.
6. The tool is split along the parting line.
7. The (moulded polymer) part is ejected or manually removed from the tool.
For prototyping, I reproduced this process on a bench top scale:
1. Polymer pellets heated in an oven.
2. Heated pellets compacted and cooled.
3. Compacted block placed in barrel of extruder.
4. Barrel and tool heated.
5. A ram used to push the polymer melt from the barrel into the cavity of the tool.
6. Tool removed from the moulder, cooled, and the part removed.
172 Chapter 6. Valve fabrication
Figure 6.1: Schematic of injection moulder. Taken from rutlandplastics.co.uk.
A gate is the inlet for polymer into the cavity. If the gate is oriented perpendicular to a
thin cavity (such as heart valve leaflet) the polymer will diverge in a scenario referred to
as radial flow. To simulate radial flow in a preliminary experiment, one of the shortlisted
polymers considered in Chapter 2, SIS30, was injected at 160 ◦C via a 1 mm diameter
gate into the centre of two stainless steel parallel circular plates each of diameter 80 mm
whose separation could be set in order to achieve a range of sample thicknesses. The
injection was performed using a manual hydraulic press, in which the injection rate was
manual so subject to variation in control. The plates were cooled from 160 ◦C to 40 ◦C
under running water over approximately one minute after the completion of injection.
The time taken to perform the injection was recorded, and the volumetric injection rate
inferred. Two injection rates and three sample thicknesses were tested, as summarised in
Table 6.1.
Table 6.1: Injection rates and sample thickness for point injected specimens of SIS30.
Code Injection rate (m3 s−1) Sample thickness (mm)
A
7±2× 10−8
0.95 ± 0.05
B 0.45 ±0.02
C 0.23 ±0.03
D
2±0.5× 10−8
0.97 ±0.05
E 0.44 ±0.02
F 0.30 ±0.02
Chapter 6. Valve fabrication 173
6.2.1 Structure analysis by small angle X-ray scattering
Cylindrical PS domain orientation leads to anisotropic mechanical properties. In this
chapter I investigate whether these can be harnessed to produce a bioinspired prosthesis,
as described in Section 1.10.1. It was not feasible to perform tensile testing to evaluate
the mechanical anisotropy on small samples, such as parts of the leaflets, so alternative
means of characterising the orientation were needed. The spacing of the PS domains in
my samples is of the order of 10 nm, whose structure can only be analysed using X-rays or
neutrons. Stasiak et al. (2010) showed that mechanical properties are strongly correlated
with small angle X-ray scattering (SAXS) analysis, making it an appropriate means
of assessing orientation. This method utilises Bragg’s law, whereby a crystalline solid
scatters electromagnetic radiation such that different path lengths result in constructive
or destructive interference at a detector. Bragg’s law describes the case of constructive
interference where d is the distance between crystal planes, θ is the scattering angle, λ is
the wavelength, and n is the order of the scattering peak:
2dsinθ = nλ (6.1)
Thus the crystal plane spacing and orientation can be calculated using monochromatic
irradiation and a detector covering a range of angles.
The orientation of the domains in the point-injected samples was evaluated by SAXS
using a synchrotron source of X-rays, at the I22 beamline at the Diamond Light Source,
Harwell Science and Innovation Campus, UK. The energy used was 12.4 keV (0.1 nm)
with 6 m between the sample and the RAPID 2D detector. The sample was irradiated
with a rectangular beam whose full width (at half maximum of the Gaussian distribution)
was 320× 80 µm, allowing spatial mapping of the orientation in small samples to be per-
formed. The integrated SAXS patterns were processed using OriginPro 9.0 (OriginLab,
MA, USA).
As depicted in Figure 6.2, the samples were moved relative to the beam, to collect patterns
at points at various radii, with the beam in the Z-axis direction. A radial section was cut,
of thickness ∼1 mm, and the beam oriented in the plane of the sample. SAXS patterns
were recorded every 1 mm across the sample when the beam was in the Z-axis direction;
and every 0.1 mm for scans through the thickness of the sample, along the Z axis.
174 Chapter 6. Valve fabrication
rf
Injection point
Z
In-plane beam
Z-axis beam
Radial section
Figure 6.2: The beam was oriented in the Z direction for SAXS scanning of the disc
of radius rf along the X and Y axes. A radial section was removed from the disc and
the beam oriented along the Y-axis, for scanning at an array of Z positions through the
sample. Red dots indicate the approximate position of measurements which were made
radially and along the Z axes.
6.2.2 Results
It was a surprise to observe that the cylindrical PS domains of the injection moulded
discs did not orient solely in the direction of flow (Fang et al., 2009; Stasiak et al.,
2010). In fact, a bimodal distribution of orientations was observed; as indicated by
the SAXS pattern and azimuthal integration shown in Figure 6.3. The 1st order peak
corresponds to a domain spacing of 27 nm, and the higher order peaks occur at
√
3 and√
7, indicating that the PS cylinders are arranged in a hexagonal structure (Stasiak et al.,
2010, 2009). Bi-directional orientation is evident across the entirety of all six samples,
and the orientation is symmetric across the sample.
Azimuthal peak broadening at full width half maximum (FWHM) was used as a measure
of the degree of anisotropy for the two reflections at 90° and 180°. In Figure 6.4, the radial
and circumferential cylinder orientation is plotted as vectors, in the X and Y directions
across the sample. The length of the vector is proportional to 1/FWHM. Close to the
injection point (within 5 mm) there is some variation in the degree of anisotropy and
orientation of the domains, which is associated with the development of flow from the gate.
At larger radii, the degree of anisotropy in the radial and circumferential orientations was
high throughout the sample, and did not vary significantly with position.
The samples were scanned with the beam oriented in the plane of the sample, as shown
in Figure 6.5. The pattern varies with position, and is approximately symmetric across
the centre line of the sample, indicating that the bimodal orientation was layered, with
remarkably good segregation between the circumferentially and radially oriented layers.
At some depths, ±0.2 mm for example, radial and circumferential patterns are visible.
Chapter 6. Valve fabrication 175
(a)
(b)
Figure 6.3: (a) SAXS pattern from a Z-axis beam through a 0.95 mm thick sample injected
at 7× 10−8 m3 s−1. The four innermost peaks (indicated by - - - at 2θ = 0.2° = θ1)
correspond to the first order of scattering from a biaxial arrangement of cylinders. The
peaks at 2θ = 0.35° = θ1
√
3 and 2θ = 0.53° = θ1
√
7 correspond to reflections from other
planes of the hexagonally structured cylinders (this arrangement is shown in Figure 6.5c).
(b) shows the azimuthal integration of the diffraction space in (a) in the region marked
by - - -. The amplitude and full width half maximum are marked. These data were
processed by Dr Joanna Stasiak.
(a) 1 mm (b) 0.5 mm (c) 0.25 mm
Figure 6.4: SAXS mapping of the orientation of PS domains in thin discs (1, 0.5, 0.25 mm
thick). The length of each tick mark corresponds to the azimuthal integration of each
peak. The origin of the X and Y axes was located at central injection point of each
sample. This data was processed by Dr Joanna Stasiak.
This is due to the finite thickness of the beam which passed through both layers simul-
taneously. In the centre of the sample there is an isotropic region, as indicated by the
circular scattering pattern at depths of 0 and ±0.1 mm.
176 Chapter 6. Valve fabrication
X-rays are scattered from all the crystal planes in an oriented PS sample. The (110),
(101), and (011) planes have a normalised spacing of 1; the (121), (211), and (112) planes
have a
√
3 spacing; and the (123) and corresponding planes have a
√
7 spacing (Honeker
and Thomas, 2000). These are marked in Figure 6.3a for the pattern from the Z-axis
beam. The corresponding peaks were found in the patterns for the in-plane beam aligned
with the hexagonally arranged PS cylinders (0.3 mm depth in Figure 6.5a). The sum of
the first order peak areas for reflections from each set of planes was compared. The peaks
found at a normalised spacing of 1 accounted for 98% of the first order scatter. Thus,
when the beam is oriented in the Z-direction, the quantity of radial and circumferential
orientation can be compared by looking at the peaks arising due to scattering from the
planes of the unit cell ((110), (101), and (011)).
I measured the thickness of the radial and circumferential layers when the sample was
scanned with the beam oriented in the plane of the sample. This was compared to the
area of the radial and circumferential peaks when the Z-axis beam was scattered. The
magnitude of the radial and circumferential peak areas was a function of the beam’s angle
of incidence on the crystal planes, however the ratio of the radial and circumferential peak
areas from the Z-axis beam matched the thickness of the radial and circumferential layers
when they were measured using the in-plane beam in 9 samples.
Chapter 6. Valve fabrication 177
(a)
(b)
(c)
Figure 6.5: (a) SAXS pattern for an in-plane beam at 13 depths (z positions) in a sample.
(b) Azimuthal integration of the first order peaks for an in-plane beam through a sample
of thickness 0.95 mm and injected at 7× 10−8 m3 s−1. The larger peaks at 0° and 180°
correspond to the radially oriented cylinders in the skin regions, and the six smaller peaks
correspond to the circumferentially cylinders in the core. (c) is a schematic of the bimodal
structure of PS cylinders. The detail shows the hexagonal arrangement of cylinders. This
data was processed by Dr Joanna Stasiak.
178 Chapter 6. Valve fabrication
6.3 Flow characterisation
It is known that deformations during flow can control PS domain orientation. To pre-
dict these deformations I needed to relate the fluid stresses and deformations through a
constitutive model. In this section I progressively test the suitability of more complex
constitutive models to evaluate their accuracy in describing the polymer’s melt behaviour.
Rheological measurements are performed in standard geometries to determine the pa-
rameters of the constitutive models, before being tested in the radial flow scenario. In
the radial flow geometry the equations may be solved analytically for some constitutive
models, otherwise finite element (FE) methods were considered. With an accurate con-
stitutive model, the velocity field under a set of process conditions can be predicted, from
which it should be possible to predict the orientation of PS cylinders. A schematic of the
radial flow geometry, which contains the nomenclature used in the subsequent analyses,
is shown in Figure 6.6.
Q
2b
Pr
rp
rf
z
r
θ
Figure 6.6: Schematic of radial flow from a point injection, using cylindrical coordinates.
Q is the flow rate through the gate, Pr is the pressure reading taken at radius rp, rf = f(t)
is the edge of the disc/flow, and b is the half-thickness of the disc.
6.3.1 Generalised Newtonian Flow
Generalised Newtonian models are the simplest constitutive relations which may be able
to adequately describe the flow field. In this analysis incompressibility, rotational sym-
metry, constant density, and no wall slip are assumed (Winter, 1975):
∇·~v = 0 (6.2)
~v = (vr(r, z), 0, 0) (6.3)
∂
∂θ
= 0 (6.4)
ρ = constant (6.5)
vr(r, θ,±b) = 0 (6.6)
Chapter 6. Valve fabrication 179
where ~v is the velocity vector, ρ is the density, and θ, r, and z are the axes defined in
Figure 6.6. Using assumptions 6.2, 6.3, and 6.4, which correspond to continuity, pure
radial flow, and symmetry, vr varies with z and r as:
vr =
f(z)
r
(6.7)
And so the rate of strain tensor, γ˙, where (1, 2, 3) correspond to (r, θ, z) is, (Winter,
1975);
γ˙ = ∇~v + (∇~v)t =
−2r
−2f 0 r−1f ′
0 2r−2f 0
r−1f ′ 0 0
 (6.8)
where f ′ = df/dz. Given an inflow of Q and half-thickness of b, f is solved to give the
velocity gradients:
∇~v = 3Q
8pirb

−1
r
(
1− z2
b2
)
0 −2
b
(
z
b
)
0 1
r
(
1− z2
b2
)
0
0 0 0
 (6.9)
It is immediately obvious that at the walls (z = b) there are no elongational deformations
(diagonal components), only shear, while at the midplane (z = 0) there is pure extensional
flow, and no shear.
Equation (6.9) can be integrated to yield an expression for vr:
vr =
3Q
8pirh
(
1− z
2
b2
)
, vθ = vz = 0 (6.10)
Using conservation of momentum:
ρ
[
∂~v
∂t
+ ~v · ∇~v
]
= ρg +∇ · σ (6.11)
where σ is the total stress tensor, and σ = −PHI + τ , where PH is the hydrostatic pres-
sure, I is the identity matrix, and τ is the shear stress tensor. The Navier-Stokes equation
in cylindrical coordinates, for the radial component with inertial elements neglected, is
formed:
∂P
∂r
= η
[
1
r
∂
∂r
(
r
∂vr
∂r
)
− vr
r2
+
1
r2
∂2vr
∂θ2
− 2
r2
∂vθ
∂θ
+
∂2vr
∂z2
]
(6.12)
180 Chapter 6. Valve fabrication
Taking assumptions 6.2 to 6.6, flow rate Q, and also assuming a Newtonian fluid of
constant viscosity η,
τ = ηγ˙ (6.13)
where τ is the shear stress and γ˙ is the shear rate. The pressure at radius rp, can be
calculated by integrating the above expression and substituting r for rf , which is the
instantaneous radius of the polymer disc as it expands.
P (rp)− P (rf ) = 2η
[
2
rp
− 2
rf
− 3Q
8pib3
(ln rp − ln rf )
]
(6.14)
However, in most cases real fluids are not Newtonian. The first deviation from the
aforementioned behaviour which I will consider is a variation in viscosity. As well as
the constant viscosity expressions shown above, Winter (1975) also derived an analytical
solution for a power law fluid, where η is a function of shear rate γ˙:
η = mγ˙n−1 (6.15)
where m is the consistency index and n is the power index. This can account for a
monotonically increasing or decreasing viscosity. Assuming that the r-components of the
normal stress components cancel, the radial pressure gradient is a function of r only, and
γ˙ =| γ˙rz |. Then, the velocity field can be obtained:
vr(r, z) =
Q
4pibr
2n+ 1
n+ 1
[
1−
(z
b
) 1+n
n
]
(6.16)
And also the pressure at rp, if the disc has an outer radius of rf :
P (rp)− P (rf ) =
(
2n+ 1
n
Q
pirf4b2
)n
mrf
b(1− n)
[
1−
(
rp
rf
)1−n]
(6.17)
Experimental materials and methods
Rheological characterisation of the selected molten polymer, SEPS22, was carried out us-
ing a stress controlled Bohlin CVO rheometer with both smooth and roughened stainless
steel parallel plates (∅ 25 mm). Steady shear viscometry was performed at 130, 150, 170
and 190 ◦C.
Previously (in Section 6.2), the injection moulding apparatus was incapable of sustaining
a consistent and controllable flow. To validate the orientation model I used a BOY 22D
Chapter 6. Valve fabrication 181
injection moulder (BOY Ltd, Northants, UK) (as shown in Figure 6.7). The moulder
consists of a 32 mm diameter barrel, and screw with 80 mm of travel. The barrel has
four heating regions from hopper to nozzle, which were set to 80, 110, 130 and 150 ◦C. I
designed and constructed a custom riser and plate assembly which could be heated and
cooled using integrated electric cartridge heaters and water cooling channels. The dimen-
sions of the cavity are set by a stainless steel sheet, allowing arbitrary sheet thicknesses
and 2D shapes to be moulded. In these experiments a circular cavity with a radius of
100 mm and thickness of 0.7, 1, and 1.5 mm were used.
Figure 6.7: BOY 22D injection moulder used for making flat discs.
The plates were heated to 150 ◦C and injected using the back pressure control mode, at
various pressures, from 25 to 95% of the maximum 800 bar barrel pressure. A pressure
line connects to the mould cavity at radius rp = 45 mm, and leads to the pressure gauge
(Terwin 2000, 500 bar range). The vertical portion of the line is filled with hydraulic oil,
though the remaining 10 mm cannot be filled with oil as it is horizontal, and so becomes
filled with the melted polymer upon injection.
Experimental results
When stress-controlled rheometric testing was performed with smooth plates the poly-
mer disc slipped, so roughened plates were used for the shear rheology characterisation.
Steady shear rheology is shown in Figure 6.8, with the data fitted to a power law model
(Equation (6.15)) the melt is clearly shear-thinning. The decreasing shear rate sweeps had
182 Chapter 6. Valve fabrication
a lower viscosity than the increasing ones, indicating some thixotropy (time dependence
of viscosity), but this was small compared to the sample-to-sample variation.
I also considered the effect of temperature on the polymer rheology; viscometry data for
steady shear at 130, 150, 170 and 190 ◦C is shown in Figure 6.9. The viscosity is highest at
130 ◦C, decreases to a relatively constant value between 150 and 170 ◦C, before increasing
at 190 ◦C, likely due to oxidation of the polymer.
10−2 10−1 100 101
102
103
104
105
106
Shear rate s−1
V
is
co
si
ty
(P
a
s)
Run 1
Run 2
Run 3
Run 4
Run 5
Figure 6.8: 5 runs of steady shear sweeps of SEPS22 at 150 ◦C, with increasing shear
rates (empty symbols); and 4 decreasing (filled symbols). The increasing and decreasing
shear rates are nearly coincident, so are hard to distinguish. The data is fitted to a power
law model (green), and the 95% confidence interval limits are shown (red). The power
law index is n = 0.094(0.082, 0.106), and consistency factor m = 4696(2430, 9072) (95%
c.i.).
The polymer melt was injected into the cavity at 150 ◦C. Figure 6.10 shows an example
of the screw position and pressure trace that were obtained during injection. An initial
rapid increase in pressure is recorded at the gauge as the polymer melt fills a short section
of the pressure gauge line. Subsequent analyses were based upon the pressure trace in
region (e), which was extrapolated back to an initial time at which the polymer reached
the pressure gauge at rp. Despite operating at 95% of the moulder’s maximum injection
pressure, the high viscosity of the polymer leads to modest injection speeds, and long
processing times. Higher temperatures would reduce processing times, but run the risk
of degradation (Figure 6.9). In future it may be possible to inject under a nitrogen
atmosphere to reduce degradation.
The polymer in the pressure line takes over 2 h to relax - a significant time in comparison
to the time taken to inject the polymer. The ratio of the material’s relaxation time to the
Chapter 6. Valve fabrication 183
10−2 10−1 100 101
102
103
104
105
106
Shear rate s−1
V
is
co
si
ty
(P
a
s)
130◦C
150◦C
170◦C
190◦C
Figure 6.9: Steady shear sweeps of SEPS22 at 130, 150, 170 and 190 ◦C.
processing time, is termed the Deborah number (De). High Deborah numbers indicate
that elastic and solid-like characteristics are important.
In Figure 6.11 the experimental data is compared to the pressure predicted using Equa-
tions 6.10 and 6.17. The experimental data reaches significantly higher pressures than
the power law and Newtonian models predict. In general, the experimental pressure is
approximately double the pressure that is predicted by the power law model. A power
law fluid would lead to a more linear increase in pressure due to the reduced viscosity
at higher shear rates, which we might find in the early part of the injection. Given the
discrepancy, it is appropriate to consider more representative rheological models which
might produce a closer fit to the experimental data, allowing the velocity field to be
predicted, and thus orientation, more accurately.
All of the rheological measurements that have been made thus far were taken in a parallel
plate geometry, which produces shear without extension. A complete description of a fluid
must be valid in both shear and extension; however it is impossible to produce steady
and spatially uniform extensional flows (Petrie, 2006), making extensional rheometry the
ugly cousin of shear rheometry. Instead, the next level of complexity accounts for the
viscoelasticity of polymers.
6.3.2 Viscoelasticity
Newtonian behaviour would have made analysis easier but was clearly not the best de-
scription for this polymer melt. The elastomeric properties of polymers are important at
184 Chapter 6. Valve fabrication
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
·104
0
10
20
30
40
50
A
D
E F
G
Time (s)
P
re
ss
u
re
(b
ar
)
0
20
40
60
80
B
C
Fully retracted
S
cr
ew
p
os
it
io
n
(m
m
)
Pressure
Screw position
Figure 6.10: Example of the pressure and screw position data collected during injection
moulding of a disc. The tool is held at 150 ◦C throughout the procedure. The pressure
gauge is stabilised at A, before the screw is retracted and loaded at B. A constant back-
pressure is applied to the screw C forcing the polymer into the disc cavity. The polymer
melt reaches the pressure gauge line, which it must fill D, before the radial flow contin-
ues E, until the cavity is filled. The screw is halted, though the pressure at the gauge
continues to increase F as the movement is propagated through the nozzle and down the
runner. The tool is left at temperate for 2 h over which time the polymer relaxes G.
melt temperatures, and so these must be accounted for during processing. The simplest
way to incorporate elastic behaviour into the constitutive fluid model is to assume that
the current stress in a fluid is linearly related to its strain history, meaning that the
stress and strain responses are separable. To assess the relative contributions of elastic
and viscous properties oscillatory rheology measurements were performed.
The amplitude sweep data are shown in Figure 6.12a. As amplitude increases, the elastic
and viscous moduli are invariant up to 5.5% deformation at which point the rheometer’s
torque limit is reached and the rheometer cannot complete the test. Linear viscoelastic
measurements can be performed up to 5% strain.
Frequency sweep testing was performed at 1% strain, well within the linear region of
viscoelasticity and torque limit of the machine. The frequency sweep shown in Figure
6.12 demonstrates that at the temperature of interest, 150 ◦C, the elastic contribution to
Chapter 6. Valve fabrication 185
0 200 400 600 800 1,000 1,200
0
5
10
15
20
25
30
35
40
Time (s)
P
re
ss
u
re
(b
ar
)
Power law fluid
Newtonian fluid
Data
Figure 6.11: Experimental data (n=3, 4,×, ◦) vs. predictions based upon generalised
Newtonian models for point injection (region (e) in Figure 6.10). Pressure is measured
at rp = 45 mm, and the predicted pressures are based upon the results of equations 6.10
and 6.17, given the measured experimental flow rates. The power law model is based
upon the rheological data presented in Figure 6.8 and the Newtonian model assumes a
viscosity of 10 kPa. The error bars of the power law model are based upon the confidence
limits of the data shown in Figure 6.8.
viscosity is greater than the viscous contribution, at all measured shear rates. As such the
Weissenberg number (viscous forces
elastic forces
) is large - suggesting solid-like behaviour as associated
with the high Deborah number in Figure 6.10. For reference, the wall shear rate in the
radial flow scenario with a typical Q = 2× 10−8 m3 s−1, reaches approximately 2.5 s−1.
The complex viscosity (Figure 6.12c) exhibited shear thinning behaviour, which followed
a power law at low temperatures, similar to the result of the steady shear experiments
(Figure 6.8). However, comparison of the steady and dynamic data also indicates that
the Cox-Merz rule (Equation (6.18)) is not valid for these flows:
η (γ˙) = η∗ (ω) when γ˙ = ω (6.18)
Indicating that high steady shear rate viscosity cannot be predicted based upon high
frequency oscillatory rheometry. The melt appears more viscous during small oscillations
than during the large deformations found in continuous shear, which may be due to the
breakdown of the self-assembled microstructure (Djabourov et al., 2013).
186 Chapter 6. Valve fabrication
0 1 2 3 4 5 6 7
·10−2
104
105
106
M
o
d
u
lu
s
(P
a)
1
1.5
2
2.5
·104
Strain
C
om
p
le
x
v
is
co
si
ty
(P
a
s)
Elastic
Viscous
Complex viscosity
(a)
10−1 100 101
103
104
105
106
130
150
170
190
210
Frequency (Hz)
E
la
st
ic
/v
is
co
u
s
m
o
d
u
lu
s
(P
a)
Elastic modulus
Viscous Modulus
(b) Viscoelastic moduli.
10−1 100 101
103
104
105
106
130
150
170
190
210
Frequency (Hz)
C
om
p
le
x
v
is
co
si
ty
(P
a
s)
(c) Complex viscosity.
Figure 6.12: Parallel rough plate oscillatory shear of SEPS22 melt; (a) shows an amplitude
sweep at 130 ◦C and 1 Hz. The elastic modulus G′ = σ0
0
cos δ and viscous modulus G′′ =
σ0
0
sin δ, where δ is the phase shift between stress and strain. (b) and (c) are frequency
sweeps, at 1% strain (0), at temperatures from 130 to 210
◦C. Crossover points for
sweeps at 190 and 210 ◦C are marked with •. There is a systematic error in the data
point collected at 6.25 Hz, which can be attributed to a changeover in the rheometer’s
stress transducer at this frequency. Complex viscosity is calculated as η∗ =
√
G′2+G′′2
ω
.
Non-linearity
The polymer melt undergoes significant deformation during injection moulding, and the
Weissenberg and Deborah numbers are both large, so non-linear viscoelastic models may
be more appropriate for modelling the flow of the fluid. A first order fluid model where
Chapter 6. Valve fabrication 187
no elastic effects are considered is Newtonian; adding two elastic terms results in a (non-
linear) third order fluid (Bird et al., 1987):
τ = −α1γ˙ + α2 D
Dt
γ˙ − α11 (γ˙ · γ˙)
− α3 D
2
Dt2
γ˙
+ α12
[
γ˙ ·
(
D
Dt
γ˙
)
+
(
D
Dt
γ˙
)
· γ˙
]
− α1:11 (γ˙ : γ˙) γ˙ + . . .
(6.19)
Where the first line corresponds to the Newtonian case, and in which the substantive/total
derivative is:
D
Dt
γ˙ =
∂γ˙
∂t
+ v ·∇γ˙ + 1
2
(ω · γ˙ − γ˙ · ω) (6.20)
The α constants are related to the viscometric functions during steady shear flow by:
η = −τyx
γ˙
= α1 − 4 (α3 − 2α1:11) γ˙2 (6.21a)
Ψ1 = −τxx − τyy
γ˙2
=
N1(γ˙)
γ˙2
= 2α2 (6.21b)
Ψ2 = −τyy − τzz
γ˙2
=
N2(γ˙)
γ˙2
= −α2 + α11 (6.21c)
The first and second normal stress coefficients, Ψ1 and Ψ2, only arise in viscoelastic fluids.
Ψ1 and Ψ2 are calculated from the normal stress differences N1 and N2. N1 is most easily
obtained with a cone and plate rheometer, as the shear rate is independent of radius
(Macosko, 1994). The normal stress difference is evaluated from the normal plate force
(Fz) as:
N1(γ˙) =
2Fz
piRcp
2 (6.22)
Where Rcp is the radius of the cone and plate. The non-homogeneous shear distribution
of the parallel plate rheometer allows the second normal stress difference to be evaluated:
N1(γ˙)−N2(γ˙) = Fz
piRpp
2
(
2 +
d lnFz
d ln ˙γpp
)
(6.23)
Where Rpp is the radius of the parallel plates and ˙γpp is the shear rate at the edge of the
plates.
Bird et al. (1987) solved the radial flow problem for the constitutive model in Equation
(6.19) using perturbation methods, to yield equations for the radial (with coefficients up
to r−5 term) and axial (up to r−6) velocity profiles during steady, creeping flow.
188 Chapter 6. Valve fabrication
The pressure drop between some point rp and the edge of the disc rf , which is assumed
to be at zero gauge pressure, is given by:
P (Rp)− P (Rf ) =
(
3Rf ln
R
Rf
)
1 + 3
10
(
Rf
R
)2
− 1
Rf ln
(
Rf
R
) [(5A2 − 3A11)− 6
7
Re− 9Rf (A3 − 2A1:11)
]
+ . . .

Where R = r
b
is dimensionless radius, and the constants A2, A3, A11, A12, and A1:11 depend
on the viscometric parameters (Co and Bird, 1977). The terms in [] brackets arise from
the normal stresses, inertia, and non-Newtonian viscosity, respectively.
Experimental Materials and Methods
The viscometric parameters for SEPS22 were evaluated using an Ares strain controlled
rheometer (TA Instruments, DE, USA). A shear rate sweep was performed with the
parallel plates (25 mm diameter), and a steady shear rate step with the cone and plate
fixture (25 mm diameter and 0.1 rad cone), at 150 ◦C. 0.7 to 1.5 mm samples were stamped
from injection moulded discs. The samples were placed in the rheometer for 30 min to
equilibrate and relax.
The velocity and pressure profile equations were solved using MATLAB r2014a (Math-
works, MA, USA).
Experimental results
The viscosity of a 6 constant 3rd order fluid (Equation (6.21a)) consists of a zero-shear
rate viscosity and a (α3 − 2α1:11) γ˙2 term. Unfortunately a second order polynomial
(A+Bγ˙2) to describe viscosity is inappropriate for this polymer melt; no value of B can
produce the correct viscosities. As such, a constant viscosity of 10 kPa s was used, as in
the original Newtonian model. In fact, a γ˙−1 term (as seen in Equation (6.15)) would be
necessary to accurately describe the viscosity.
The 1st normal stresses were collected using the cone and plate rheometer (Figure 6.13).
Unfortunately it was hard to contain the sample within the rheometer plates for long
durations. Given the difficulty of sample containment and normal stress measurement it
is remarkable that a γ˙2 relationship - that may be used in Equation (6.22) - fits so closely
to the 1st normal stresses (Figure 6.13b).
Chapter 6. Valve fabrication 189
0 200 400 600 800 1,000 1,200 1,400
0
0.01
Time (s)
γ˙
(s
−1
)
−2
−1
0
1
2
3
4
5
·107
N
1
(P
a)
(a)
10−2 10−1 100
101
102
103
104
105
106
107
108
γ˙ (s−1)
N
1
(P
a)
(b)
Figure 6.13: 1st normal stresses during pulse shear for SEPS22 at 150 ◦C, calculated using
Equation (6.22). (a) shows the 1st normal stress during a step up and step down in shear
rate with the cone and plate rheometer. Sample containment was lost after approximately
300 s. (b) shows the 1st normal stress during 5 pulse shear rates. The line corresponds to
the relationship N1 = 1000γ˙
−2, for use in Equation (6.21b).
The difference between the 1st and 2nd normal stresses was determined using the parallel
plate rheometer. There is some disparity in normal stress between the upward and
downward sweeps at low shear rates, likely due to memory effects. As such, the upward
sweep gradient was used, d lnFz
d ln γ˙
= 0.5 and Fz = 20γ˙, in Equation (6.23), to yield N1(γ˙)−
N2(γ˙) = 1× 105γ˙. Unfortunately, this does not fit so cleanly with the form of the
2nd normal stress coefficient (Equation (6.21c)), in which Ψ2 = N2/γ˙2. And so assuming
an average shear rate of 1 s−1, Ψ2 = 1× 105. Also, there is a difference in viscosity
measured with this rheometer, and previously measured using the stress controlled Bohlin,
in Section 6.3.1, possibly due to the need to use rough plates in the previous, stress-
controlled experiment.
190 Chapter 6. Valve fabrication
Figure 6.14: The difference be-
tween the 1st and 2nd normal
stresses during shear rate sweeps
(up ◦ and down 3) in the parallel
plate rheometer for 3 runs (red,
green, blue). The data is plotted
as a natural logarithm with a line
of gradient 0.5 fitted to the up-
ward sweep, for use in Equation
(6.23). Sample containment was
lost during the downward sweep
in the second (green) run.
−4.5 −4 −3.5 −3 −2.5 −2 −1.5 −1 −0.5−3
−2
−1
0
1
2
3
ln γ˙
ln
F
z
These values were used to determine the constants α1, α2, α3, α11, and α1:11, equal to
1× 104, 5× 102, −1.5× 10−6, −9.95× 104, and 0, respectively.
The 3rd order viscoelastic model predicts a significantly greater pressure drop than the
equivalent viscosity Newtonian model (Figure 6.15) making it a more accurate model
for this flow. The 3rd order model is only evaluated up to rf = 0.1 m, after which the
cavity should be filled. The initial experimental pressure spike was due to the polymer
filling the horizontal portion of the pressure line. Flashing (overfilling) of the cavity may
explain the higher pressures recorded at the end of the run. If the 3rd order analytical
model were able to take into account the shear thinning viscosity (rather than assuming
a constant viscosity), a more accurate prediction of pressure would be expected.
6.3.3 Computational modelling
Taking account of normal stresses in the analytical model was another step closer to
understanding the behaviour of the SEPS polymer melt, however it still led to a 30%
discrepancy compared to the experimental data. Furthermore, to model the flow in the
complex 3D heart valve geometry a viscoelastic model should be implemented in a finite
element (FE) solver. I considered three options: the (viscoelastic) White-Metzner and
Giesekus models, and (generalised Newtonian) Carreau-Bird model.
The White-Metzner model describes low shear rate viscoelasticity using the following
relation:
σ + λr∇σ = 2ηvγ˙ (6.24)
where σ is the total stress tensor, λr is the relaxation modulus, ηv is the viscosity, and γ˙
Chapter 6. Valve fabrication 191
−100 0 100 200 300 400 500 600 700 800 900 1,0000
5
10
15
20
25
30
35
40
45
50
*
Time (s)
P
re
ss
u
re
(b
ar
)
Power law fluid
Newtonian fluid
Data
3rd order viscoelastic
Figure 6.15: Compari-
son of raw pressure data
with the predicted pres-
sure using a Newtonian,
power law, and a 3rd or-
der viscoelastic model.
The spike in pressure
due to filling of the pres-
sure line is indicated by
*.
is the shear rate. In Equation (6.24), the viscosity and relaxation modulus follow power
law relations:
ηv(γ˙) = (η0 − η∞)(1 +m2vγ˙2)
nv−1
2 + η∞ (6.25)
λr(γ˙) = mλγ˙
nλ−1 (6.26)
Where η0 and η∞ are the zero and infinite shear rate viscosities, respectively. mv and nv
are the viscoelastic viscosity power law consistency factor and index respectively, and mλ
and nλ are the viscoelastic relaxation modulus power law consistency factor and index
respectively. No purely viscous contribution to the stress tensor was used, as this is
associated with very fast relaxation modes in polymer solutions.
The power law viscosity parameters as previously reported (Figure 6.8) were used: mv =
4696 and nv = 0.094.
During steady shear the first normal stress for a White-Metzner fluid is calculated as
(Morrison, 2001):
Ψ1(γ˙) = 2ηv(γ˙)λr(γ˙) (6.27)
And the second normal stress, Ψ2 = 0. Using the shear sweep data from Figure 6.14
yielded mλ = 1.122 and nλ = 2.906.
I also tested the applicability of the Giesekus model (Giesekus, 1982) to the polymer
melt. More realistic constitutive models, such as the one proposed by Giesekus, consider
the polymer chains as dumbbells which are deformed during flow. The upper-convected
192 Chapter 6. Valve fabrication
Maxwell equation considers a dumbbell in a dilute solution, Giesekus extended this equa-
tion to model dumbbells which are surrounded by dumbbells, which impart anisotropic
drag on each other. This results in a non-separable constitutive model, which has a
quadratic stress tensor term, and non-zero normal stresses (Larson, 1988). The constitu-
tive equation of the Giesekus model is:
τ
λ
+
∂τ
∂t
+ v∇τ −
(
(∇v)T τ + τ (∇v)
)
=
η0
λ
γ˙ − a
η0
τ · τ (6.28)
Where a is the mobility parameter, which represents the anisotropy of hydrodynamic
drag on the polymer molecule (Bird and Wiest, 1985).
During steady shear the viscosity η is given by;
η(γ˙) =
η0(1− n2)
1 + (1− 2a)n2 (6.29)
Where n2 and Λ are given by:
n2 =
1− Λ
1 + (1− 2a)Λ (6.30)
Λ =
√√
1 + 16a(1− a)λ2γ˙2 − 1
8a(1− a)λ2γ˙2 (6.31)
The first normal stress difference, N1, during steady shear, is:
N1 = 2λη0
n2(1− an2)
λ2a(1− n2) (6.32)
And the analytical form of the second normal stress difference, N2 is:
N2 =
−aN1
2
(1− n2)
1− an2 (6.33)
When implementing the Geisekus model in an finite element solver, convergence can be
promoted by decomposing the stress tensor into viscoelastic and purely viscous compo-
nents, such that τ = τE +τ V , where τE is equal to τ in Equation (6.28), and the purely
viscous component τ V = 2η0ηrγ˙, where ηr is some viscosity ratio, preferably greater than
0.11 (ANSYS, 2012).
Using the parallel plate rheometry shear sweep data the Giesekus model was fitted using
the 4 parameters shown in Table 6.2. The parameters were optimised for viscosity, using
OriginPro 9 (MA, USA). The viscosity ratio, ηr, was bounded between 0.11 and 1, and
Chapter 6. Valve fabrication 193
the upper limit of α was set to 0.5 due to convergence difficulties in ANSYS when using
higher values.
10−2 10−1 100
103
104
105
106
Shear rate (s−1)
V
is
co
si
ty
(P
a
s)
an
d
N
or
m
al
st
re
ss
d
iff
er
en
ce
(P
a)
Measured viscosity
Giesekus Viscosity
Measured normal Stress Difference
Giesekus Normal Stress Difference
Figure 6.16: Increasing
shear rate sweep in the
parallel plate rheometer
with the results of the
Giesekus model fitted to
the data using equations
6.29, 6.32 and 6.33, with
the parameters as listed
in Table 6.2.
Table 6.2: Parameters used in the Giesekus model (Equation (6.28)) for fitting to shear
viscometry and then subsequently used in the FE model.
Parameter Value Unit
η0 3.2× 105 Pa s
α 0.4 -
λ 30 s
ηr 0.15 -
In contrast to the parallel plate shear rate sweep data which was collected with rough
plates (Figure 6.8), the data shown in Figure 6.16 shows a zero shear rate limit. This is
appropriate for a (Newtonian) Bird-Carreau model fit to this data:
η = η∞ + (ηb0 − η∞)
(
1 + (λbγ˙)
2
)nb−1
2 (6.34)
Where η∞ is the infinite shear rate viscosity, ηb0 is the zero shear rate viscosity, λb and
nb is the power law index for the Bird-Carreau model; which were set as shown in Table
6.3.
194 Chapter 6. Valve fabrication
Table 6.3: Parameters used in the Bird-Carreau model (Equation (6.34)) for fitting to
shear viscometry (Figure 6.8) and then subsequently used in the FE model.
Parameter Value Unit
ηb0 3.5× 105 Pa s
η∞ 1× 102 Pa s
λb 40 s
2
nb 0.61 -
Finite element analysis
ANSYS Polyflow (PA, USA) is a powerful FE solver which is capable of solving the afore-
mentioned differential equations. Large values of the Weissenberg number, which imply
strong memory effects, invoke the use of an evolving function - even during steady simu-
lations - to promote convergence when using the White-Metzner fluid model. Evolution
gradually increases the flow rate, to ensure convergence. The initial flow rate is set such
that the Weissenberg number is less than 0.3 (ANSYS, 2012).
For the FE model, an axisymmetric geometry was assumed, as shown in Figure 6.17. The
segment is 5°× 0.7 mm× 100 mm. A hard, unbiased, tetrahedral mesh was used. The FE
solver, with a Newtonian constant viscosity model, was checked for mesh insensitivity.
The following results are reported for a mesh consisting of 20 elements radially, 10 through
the thickness, and 3 across the segment.
In
Out
Figure 6.17: Mesh of the injection moulding disc segment. A fully developed flow enters
the model from the left, and exits from the right. The top and bottom surfaces have no
slip constraints, and each side surface has a symmetric constraint.
Results
The actual and predicted pressure measurements for the analytical model, FEM and
experimental data are shown in Table 6.4. The constant viscosity of the Newtonian
model is certainly inappropriate for modelling the system. The addition of Newtonian
shear thinning behaviour results in a pressure drop 1/3 the magnitude of the experimental
Chapter 6. Valve fabrication 195
case. As such, viscoelasticity and the addition of normal stresses are necessary to account
for higher pressure. The Giesekus model, however, overestimates the pressure drop;
which may be due to the overestimation of normal stresses when the model parameters
are optimised based upon the viscosity (Figure 6.16), and also due to the constraint
on α in Equation (6.28) which was required for convergence. Furthermore, while the
Giesekus model very often provides an excellent fit to data in shear flow, its performance
in extension is known to be inferior. In particular, Larson (1988) reported that Giesekus
does not predict strain softening in biaxial extension.
Table 6.4: Comparison of pressure drops between P (r = 0.045)−P (r = 0.1) as recorded
and predicted using analytical and FE models.
Model
Shear thin-
ning?
Viscoelastic
/ normal
stresses?
∆P (bar)
Experimental data X X 35
Newtonian analytical 19
Power Law analytical X 12.5
3rd order analytical X 25
FEM Newtonian 19
FEM Power-law X 8
FEM Carreau-Bird X 100
FEM Giesekus X X 67
The velocity profiles are compared at r = 0.045 m, when rf = 0.1 m, as shown in Fig-
ure 6.18. The analytical and FE power law models display characteristic plug-like flow
profiles. The Carreau-Bird and viscoelastic models have profiles which are similar to the
constant viscosity Newtonian model, but the maximum velocity is lower. The Giesekus
and Carreau-Bird models predict a lower maximum velocity than the other models.
The constant viscosity Newtonian model had remarkably similar results in comparison
to the more complex models: it was fortuitous that at the tested shear rates the variable
viscosity was similar to the constant viscosity. This may not occur in future injection
scenarios, so it is necessary to use the variable viscosity models.
The 3rd order fluid model had the closest pressure fit, but is not available in FE software,
so was not used. This left the FE power-law, Carreau-Bird, and Giesekus models. The
introduction of normal stresses in the Giesekus model went some way to accounting for
the pressure drop. However, the Giesekus model had poor stability in the FE solver at
different processing conditions and convergence could not be guaranteed.
The power-law model produced the least accurate prediction of pressure drop, and the
196 Chapter 6. Valve fabrication
velocity profile was also very different to all other models. On the other hand, the Bird-
Carreau led to a a similar velocity profile to the (accurate) Giesekus model, despite its
overestimation of pressure. The stability of the problem in the FE solver was much
better, allowing it to be used to model flows in complex meshes. As a consequence, the
Bird-Carreau model was selected for further modelling. In other instances where normal
stresses are important for describing the behaviour of interest, extrudate swell during
extrusion for example, the Bird-Carreau would not be deemed suitable.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6
·10−4
−4
−3
−2
−1
0
1
2
3
4
·10−4
Velocity (m s−1)
z
p
os
it
io
n
(m
)
Analytical Newtonian
Analytical Power Law
3rd order fluid
FEM Newtonian
FEM Power Law
FEM Giesekus
FEM Bird-Carreau
Figure 6.18: Velocity profiles for r = 0.045 m, when rf = 0.1 m, with Q = 2× 10−8 m3 s−1,
based upon the analytical and FE models.
All of the models considered in this chapter, even Giesekus’, assume an isotropic viscosity.
Given that SEPS22 is moulded below the order-disorder transition temperature, and the
polymers are anisotropic, it is likely that the viscosity is also anisotropic. Particularly
in mixed flow scenarios (elongation and shear), the models are unlikely to provide a
completely accurate description of the flow.
Chapter 6. Valve fabrication 197
6.4 Orientation modelling
6.4.1 Background
I previously selected force fields as the preferred means of orienting the BCP domains.
Preliminary experimentation yielded an unexpected phenomenon during slow point injec-
tion between plates - biaxial orientation of the BCP cylinders. My search of the relevant
literature indicated that although both shear and stretch are widely recorded as mecha-
nisms for the orientation of cylinders in BCPs, the scenario where both deformations are
competing for dominance is rare.
Trebbin et al. (2013) worked with aqueous solutions of amphiphilic BCP micelles, which
are liquid at room temperature and are significantly less viscous than the polymer melts
which I have considered. The BCP micelles form semi-flexible anisotropic particles with
a diameter of 25 nm and length of several µm. Between 5 and 30% wt/wt the particles
have a nematic arrangement, which makes them a model solution for carbon nanotubes,
fibrous proteins, rod-like nanoparticles and DNA.
The authors flowed the solution in a microfluidic channel, which had curved and contraction-
expansion flow features, while simultaneously performing SAXS and polarized optical
microscopy. Generally, the micelles oriented in the direction of flow, due to the shear
gradient. However, during expansion of the flow channel, shear and extensional forces
competed. In the central region of the flow, during expansion, the extensional forces
dominated and the authors observed a region of orientation perpendicular to the flow
direction.
The 2D flow field was obtained using particle image velocimetry, and the flow was also
computationally modelled. Trebbin et al. (2013) compared the extension and shear rates
using the ratio:
Cs =
ε˙
γ˙
(6.35)
here ε˙ is the extension rate, calculated as:
ε˙ = 6
IIIγ˙
IIγ˙
(6.36)
The shear rate is:
γ˙ =
√
2IIγ˙ (6.37)
Where IIγ˙ = tr
(
γ˙2
)
and IIIγ˙ = det (γ˙) are the second and third invariants of the strain
rate tensor respectively.
198 Chapter 6. Valve fabrication
In the experimental setup of Trebbin et al. (2013), the critical value of Cs was determined
to be 0.14, such that regions in which Cs > 0.14 contained micelles oriented parallel to
flow, and where Cs < 0.14, micelles were oriented perpendicular to the direction of flow.
They also noted that the shear thinning nature was an important characteristic of the
Newtonian fluid for accurately equating the modelling and experimental results.
I sought to determine whether a similar shear-extension critical value might also be
applicable to the the BCP melt flow and able to predict the layered structure which I
observed (in Section 6.2.2). Furthermore, Trebbin et al. (2013) made some inroads into
predicting orientation, but their analysis is somewhat limited as it assumes that extension
always results in orientation perpendicular to the direction of flow (though extension may
also be in the direction of flow). As such, an alternative model is proposed.
Finally, while perpendicular orientation has not been previously observed in phase sepa-
rating BCP, it does occur in short fibre polymer composites, for which various modelling
strategies do exist. To reduce computational load, modelling of the flow through thin
cavities has used the Hele-Shaw approximation:
∂
∂x
(
S
∂P
∂x
)
+
∂
∂y
(
S
∂P
∂y
)
= 0 (6.38)
Where S =
∫ h
0
z2/ηdz, and h is the cavity thickness in the z direction, which reduces
the flow equations to 2D. However, such a formulation would be incapable of predicting
a layered structure, as previously observed and which I aim to predict. And so, with
greater computational power, 3D simulations may be pursued (Vlachopoulos and Strutt,
2003).
Thus far, it has been assumed that orientation can be described by a single vector which,
for a single, rigid fibre or cylinder, is indeed the case. However, within a finite fluid
element containing many fibres it is more appropriate to describe orientation with a
probability distribution. However, the probability distribution is cumbersome, and is
hard to mathematically manipulate, particularly as a function of flow during processing.
As such, orientation tensors are often employed.
The components of orientation tensors are equivalent to the coefficients of a Fourier series
expansion of the probability distribution function. Symmetric second and fourth order
tensors are generally used to describe the orientation. Fibres are transported convec-
tively, while shear and stretch reorient fibres. Advani (1987) derived the equations of
change, referred to as the Tucker-Folgar equation, for the orientation tensors during flow.
Theoretically, with a valid flow field and complete fibre information, including the fibre
Chapter 6. Valve fabrication 199
aspect ratio, density, and a connectivity parameter, orientation can be predicted. The
Tucker-Folgar equations require a closure approximation to be solved, and unfortunately
the solution is dependent on the choice of closure method. Models for the prediction
of short-fibre orientation during injection moulding, can be found in commercial soft-
ware. However, in this chapter, attention is focussed upon simple orientation models,
the “scalar” model, previously proposed in Equation (6.35) and Section 6.4.2, and a
“vectorial” model which I propose in Section 6.4.4.
In this chapter there is no substantial motivation to use a probability distribution to
describe orientation, particularly as the anisotropic basis for the hyperelastic material
model used to predict strain energy density in the heart valve requires a definite vector
orientation. Therefore, a definite vector basis is used to describe orientation in this
chapter.
The radial flow geometry is also taken as a simple injection case for polymer compos-
ites. In most commercial scenarios, polymer composites are injected into cold injection
moulding tools (below the glass transition temperature). In these cases the orientation is
dominated by the flow patterns in the advancing polymer front. Hot polymer is convected
from the core layers to the skin layer where it cools, in a process described as fountain
flow. The rolling of the polymer composite has been shown to improve orientation at the
wall through a combination of shear and extension, though also this contributes to built
in stresses (Mavridis, 1988).
6.4.2 Orientation model
It is well known that shear causes cylindrical BCP domains to align in the direction of
flow, and extension causes the cylinders to align in the principal direction of extension.
Thus, the formulation proposed by Trebbin et al. (2013), in Equation (6.35), could be
used to predict orientation. Equation (6.35) determines the dominant mechanism for
orientation. If shear dominates, Cs > CsC , where CsC is a critical valve of the ratio
of shear to stretch rate. Where shear dominates, at the walls of a mould for example,
cylinders are oriented in the direction of flow, and their orientation vector, ~a, is given by:
~a = ~γ =
~v
|~v| (6.39)
In thin, divergent flows stretch occurs perpendicular to the flow direction - as previously
observed in the radial flow scenario and the channels of Trebbin et al. (2013) - and so the
orientation direction when stretch dominates, Cs < CsC , can be calculated based upon
200 Chapter 6. Valve fabrication
the surface normal, n, and velocity vectors:
~a = ~ε =
~n× ~v
|~n× ~v| (6.40)
In all instances the orientation vectors, ~a, are defined over a hemisphere, such that:
~a = −~a (6.41)
The fibres are assumed to be inertia-less, which is a reasonable assumption in a creeping
flow. Convection of fibres is also neglected, so their orientation history is not taken
account of, which may be a more significant shortcoming of the model.
6.4.3 Methods
Injection moulding was performed using the same apparatus and methods as previously
described in Section 6.3.1. Three disc thicknesses, 0.7, 1.0 and 1.5 mm, were injected, at
3 injection rates from 2.5 to 250 mm3 s−1. The final radii of each disc was between 50 and
100 mm.
Previously, in Figure 6.5, I showed that the biaxial SAXS pattern which can be observed
when the X-ray beam passes through the thickness of a sample (Z-axis beam, see Figure
6.2) is due to a layered arrangement of oriented cylinders. The X-ray beam has a finite
thickness, with a Gaussian distribution that has a full width half maximum of less than
230µm, and root mean square of less than 100 µm (Bruker Nanostar SAXS). When the
thickness of the beam is small in comparison to the sample’s thickness (1 mm) distinct
skin and core layers of orientation are visible when viewed in the plane of the sample (see
Figure 6.2).
Radiation passing through a crystalline structure is scattered, leading to interference
at the detector. The scattering angle is related to the crystal spacing by Bragg’s Law
(Equation (6.1)), and the amount of radiation which is scattered is a function of the
crystal quality and thickness. The full width at half maximum of a peak is related
to the size of the crystalline domains (Borsali and Pecora, 2008), while the peak area is
proportional to quantity of crystalline material aligned correctly to give Bragg interference
at the detector (Fratzl et al., 2013).
Two rows of SAXS measurements (at 5 mm spacing) were made on a radial strip cut
from each sample (as indicated in results Figure 6.21a), using a Bruker Nanostar with
CuKα radiation (Materials Science and Metallurgy, University of Cambridge). The az-
Chapter 6. Valve fabrication 201
imuthally integrated patterns were analysed using the MATLAB Peakfit function. A
Gaussian/Lorentzian blended curve with a flat baseline correction was fitted to the peaks.
The peaks were integrated, indicating the amount of orientation in each direction.
6.4.4 Results
I applied the proposed orientation models to the velocity field predicted by the Giesekus
and Carreau-Bird FE models. Figure 6.19a shows how the calculated extension and shear
rates vary through the thickness of the disc. The shear rate is highest at the walls, and
decreases to zero at the midplane (Figure 6.19a). The rate of extension is greater in the
core than in the skin regions of the disc. The ratio of extension to shear, Cs, is shown in
Figure 6.19b.
Cs is large at the wall, and decreases towards the midplane. At the midplane, where both
the shear and extension rates are calculated to be zero, their ratio is undefined. This
singularity arises due to the formulation of the extension rate (Equation (6.36)). In the
skin region, the extension type is uniaxial, whereas at the midplane the extension type
is planar. If the strain rate is decomposed into its components and simplified (Equation
(6.42)) it becomes clear that the conventional formulation of strain rate also depends on
the shearing components (γ˙ij, i 6= j), which are zero at the midplane. As such, planar
stretch is incompatible with a 3D formulation of the stretch rate.
ε˙ = −6 γ˙22γ˙
2
31
γ˙11γ˙22 − γ˙231
(6.42)
Vectorial analysis
In response to the singularity of the standard shear and stretch formulations at z = 0, I
proposed an alternative method. The extensional and shear components of the velocity
gradient tensor can be extracted as follows:
∇~v ·
1 0 00 1 0
0 0 1
 = ∇~v · I =

∂u
∂x
∂v
∂y
∂w
∂z
 = ~εv (6.43)
202 Chapter 6. Valve fabrication
0 1 2 3 4 5 6 7 8
·10−4
0
0.2
0.4
0.6
0.8
1
z position (m)
γ˙
(s
−1
)
0
0.2
0.4
0.6
0.8
1
·10−2
ε˙
(s
−1
)
γ˙ G γ˙ C-B
ε˙ G ε˙ C-B
(a) 0 1 2 3 4 5 6 7
·10−4
0
50
100
150
200
z position (m)
C
s
0.01 G 0.03 G
0.05 G 0.07 G
0.9 G 0.01 C-B
0.03 C-B 0.05 C-B
0.07 C-B 0.09 C-B
(b)
Figure 6.19: (a) The shear (γ˙) and extension (ε˙) rates at a radius of 0.045 m, of a disc of
rf = 0.1 m, and (b) the ratio of extension to shear rate (Cs) at various radii, based upon
the velocity field predicted by the Giesekus (solid lines) and Carreau-Bird (dashed lines)
FE models.
~γv = ∇~v · (1− I) =

∂u
∂y
+ ∂u
∂z
∂v
∂x
+ ∂v
∂z
∂w
∂x
+ ∂w
∂y
 (6.44)
Where ~εv is the extension vector and ~γv is the shear vector, and I is the identity matrix.
As before, the magnitude of the stretch and shear vectors can be compared to determine
the dominant deformation:
Cv =
|~γv|
|~εv| (6.45)
Plots of ~γv, ~εv, and Cv shown in Figure 6.20. All 3 parameters have the same form as
their scalar equivalent in the skin regions. However, at the core of the sample ~ε > 0, and
so Cv is definite. While the vectorial model makes more physical sense at the sample’s
midplane, it only works for uniaxial flows, so cannot be used for modelling the complex
valvular geometries. Away from the midplane, the vectorial and scalar models have a
similar form.
Chapter 6. Valve fabrication 203
0 1 2 3 4 5 6 7 8
·10−4
0
0.2
0.4
0.6
0.8
1
z position (m)
S
h
ea
r
m
ag
n
it
u
d
e
(~γ
,
s−
1
)
0.5
1
1.5
2
2.5
3
·10−3
S
tr
et
ch
m
ag
n
it
u
d
e
(~ε
s−
1
)
~γ G ~γ C-B
~ε G ~ε C-B
(a) 0 1 2 3 4 5 6 7
·10−4
0
50
100
150
200
z position
C
v
0.01 G 0.03 G
0.05 G 0.07 G
0.09 G 0.01 CB
0.03 CB 0.05 CB
0.07 CB 0.09 CB
(b)
Figure 6.20: The vectorial shear (|~γv|) and extension (|~εv|) rates in the disc (rf = 0.1 m),
at a radius of 0.045 m (a), and the ratio of extension to shear rate (Cv) (b) at various radii,
based upon the velocity field predicted by the Giesekus (solid lines) and Carreau-Bird
(dashed lines) FE models.
SAXS orientation maps
Figure 6.21 shows results of a SAXS array scan of a radial segment of a sample. The
bimodal orientation which was reported in Section 6.2.2 is clearly evident in this sample,
though there is a variation in the direction of orientation along the sample (6.21a). The
magnitude of the peak areas change across the samples, possibly due to the angle of
incidence of the X-ray beam with respect to the crystal planes, and inhomogeneities in
the polymer due to variable cooling. When the peak area is considered as a fraction of
the total peak area, the proportion of radially oriented phase remains stable (60±10%)
along the radius of the sample shown in Figure 6.21b.
At the edge of the sample the orientation vectors are not radial and circumferential, which
may be due to the flow not being perfectly circular when it reaches the edge of the cavity,
thus causing it not to flow solely in the radial direction as it fills the cavity.
Ten flow rate and disc thickness combinations were injected for SAXS scanning. There
was some variation in sample size, as not all samples filled the cavity. In these instances,
the flow may have slowed, particularly at larger radii, possibly reducing the quality of the
orientation. The SAXS measurements were grouped, and averaged, at four radii: 0 to 25,
25 to 50, 50 to 75, and 75 to 100 mm. The SAXS patterns were then evaluated against
204 Chapter 6. Valve fabrication
−20 −10 0 10 20 300
10
20
30
40
50
x position (mm)
y
p
os
it
io
n
(m
m
)
(a)
0 10 20 30 40 50 60
0
0.2
0.4
0.6
0.8
1
y position ≡ radius (mm)
R
ad
ia
l
fr
ac
ti
on
(b)
Figure 6.21: (a) SAXS scan of a segment of disc with the cylinder orientation marked.
The length of the vector is proportional to the peak area. The black arc indicates the
outer radius of the disc and the arrows indicate the direction of flow (a). (b) shows
the peak area as a function of radius in a sample (for the sample shown in (a)). The
radial peak is calculated from the average area of the two peaks which corresponds to
the cylinders oriented radially, and Peak 2 to those oriented circumferentially. The radial
peak fraction is calculated as a proportion of the sum of the peak areas at each point. In
this sample the radial orientation dominates.
their orientation predictions based upon the Carreau-Bird flow field and scalar prediction
model (Section 6.4.2). The ratio of stretching deformation to shear deformation (Cs) is
calculated throughout the flow field based upon Equation (6.35). I hypothesised that
orientation is stretch controlled below a critical magnitude of Cs, which I will refer to as
CsC , allowing the orientation to be predicted for each modelled flow field.
Figure 6.22 shows the fraction of points which were predicted incorrectly as a function
of CsC . At low values of CsC the sample is predicted to contain purely circumferential
orientation, while at high values the sample is predicted to contain purely radial orien-
tation. These plots indicate that at a variety of process conditions, and throughout the
radial geometry, a constant value of CsC can be used to predict orientation.
Figures 6.22a to 6.22d are combined to produce a plot where the maximum recorded error
between any single experiment and simulation is shown as a function of CsC . Figure
6.24a also indicates that the optimal value of CsC for the SEPS22 flow at 150
◦C is
3. Furthermore, among the flows tested in this batch of experiments, orientation was
predicted correctly in 85% of the sample (by volume). In the worst case, orientation was
correctly predicted in over 70% of the sample (by volume).
There still remains the question of whether there is a physical significance to CsC = 3.
Figure 6.23 shows the magnitude of extensional and shear deformations at the critical
value of Cs, demonstrating that extension is more efficient at orienting PS cylinders than
Chapter 6. Valve fabrication 205
10−4 10−3 10−2 10−1 100 101 102 103 104
0
0.2
0.4
0.6
0.8
1
CsC
P
ro
p
or
ti
on
in
co
rr
ec
t
(a) r=0 to 25 mm
10−4 10−3 10−2 10−1 100 101 102 103 104
0
0.2
0.4
0.6
0.8
1
CsC
P
ro
p
or
ti
on
in
co
rr
ec
t
(b) r=25 to 50 mm
10−4 10−3 10−2 10−1 100 101 102 103 104
0
0.2
0.4
0.6
0.8
1
CsC
P
ro
p
or
ti
on
in
co
rr
ec
t
(c) r=50 to 75 mm
10−4 10−3 10−2 10−1 100 101 102 103 104
0
0.2
0.4
0.6
0.8
1
CsC
P
ro
p
or
ti
on
in
co
rr
ec
t
(d) r=75 to 100 mm
Figure 6.22: Number fraction of points for which orientation were predicted incorrectly
based upon the given value of CsC . Each line corresponds to a different sample. Radial
orientation dominates in nearly all samples, and so the incorrect proportion is low for
low values of CsC .
shear is. Further work is need to assess the significance of CsC in the Tucker-Folger
equation (Advani, 1987). Although not shown here, I also calculated the critical ratio
(CvC) of vectorial extension (~εv) and shear (~γv) proposed in Section 6.4.4, to be 10.
t1 t0 t1
Figure 6.23: Schematic to illustrate the power of
stretch over shear. Given 2 fluid elements separated
by 1 unit (black  ); after 1 time unit (from t0 to
t1) the elements are separated by stretching (blue ), or by shearing (red  ) deformations. The il-
lustrated deformations are for flow scenarios where
CvC = 10. Stretch is a more efficient means of ori-
entation than shear.
I compared the predicted orientation distribution with the measured orientation. In
206 Chapter 6. Valve fabrication
Figure 6.24b, the proportion of radial orientation is mapped as a function of injection
rate and thickness. Greater disc thicknesses promote circumferential orientation, while
higher flow rates promote radial orientation. In Figure 6.24b, a 100-fold reduction in
flow rate only reduces the fraction of radial orientation by 2%, while a 2-fold change in
thickness leads to a 10% change in radial orientation.
Several samples were injected at a higher temperature (170 ◦C and 190 ◦C). In these
samples there was a lower fraction of radial orientation compared to samples injected at
the same rate, but at lower temperatures. Without further work, it is unclear whether
temperature solely changes the polymer’s rheology and thus flow field, or whether CsC
may also vary with temperature.
10−4 10−3 10−2 10−1 100 101 102 103
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
CsC
M
ax
im
u
m
p
ro
p
or
ti
on
p
re
d
ic
te
d
in
co
rr
ec
tl
y Maximum error
Average error
(a)
0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6
0
20
40
60
80
100
120
140
160
180
200
Thickness (mm)
R
at
e
(m
m
3
s−
1
)
0.78
0.8
0.82
0.84
0.86
0.88
0.9
R
ad
ia
l
fr
ac
ti
on
(b)
Figure 6.24: (a) shows the maximum and average proportion of incorrectly predicted
orientations amongst all samples shown in Figure 6.22. The optimal value of CsC is
marked with , at CsC = 3. (b) is a contour plot of orientation in the thickness-rate
space when injection moulding a disc of SEPS22 at 150 ◦C. The contours are based upon
the scalar orientation model, using the flow field predicted by the Carreau-Bird model.
Scatter marks indicate experimental values whose orientation was evaluated by SAXS.
6.5 Controlling orientation in the injection moulded valve
Thus far, I have shown that flow fields can be used to control the orientation in a simple
injection moulded part. These findings are now to be translated to a prosthetic heart
valve, such that the leaflets have a bioinspired orientation of PS cylinders.
The spherical form valve, as previously described in Section 5.2, was used for testing
the control of orientation in a sample part. For the production of prototype valves, the
Chapter 6. Valve fabrication 207
Table 6.5: Processing conditions for injection of valves.
Gate position Total flow rate (mm3 s−1) Leaflet flow rate (mm3 s−1)
Free Edge 17.1 5.7
Commissure 11.3 3.8
Free Edge Centre 6 2
bench top injection moulding apparatus was adapted from that used in Section 6.2 to
inject discs. A custom aluminium mount was made for the stainless steel valve tool,
consisting of four locking screws for clamping the tool onto the injection plate. Three
injection plates were fabricated such that various gate (inlet) positions could be tested.
The injection scenarios were referred to as Free Edge (FE), Commissural (C), and Free
Edge Centre (FEC), which are illustrated in Figure 6.25. The cavity was vented through
the sewing ring and injection was continued until a steady flow (flash) through the vents
developed. The benchtop injection moulding apparatus has limited control of injection
rate or pressure, nor was I able to record the pressure during injection. Instead, the barrel
position was recorded with respect to time, yielding the flow rates shown in Table 6.5.
After injection, the tool was removed from the injection moulder and cooled by direct
contact with running water.
The leaflets were dissected from the valve and mounted on a frame which flattened them
such that the X-ray beam passed perpendicular to the sample. SAXS was used to assess
the spatial distribution of PS cylinders across the leaflets. SAXS was performed on
beamline I22 at Diamond Light Source, Harwell Science and Innovation Campus, UK.
The peak integration method previously described, was used to assess orientation.
Concurrently, based upon the processing conditions used to inject the valves, the leaflet
orientation was predicted. This provided a test of the orientation model in a complex
geometry. The Carreau-Bird rheological model for the polymer melt, as determined in
Section 6.3.3, was used to model the flow field. Using the scalar orientation model, which
I validated for the simple geometry in Section 6.4.4, the orientation of PS cylinders in
the leaflet was predicted. The cavity was assumed to be axisymmetric with order 3, with
the flow rate into each valve given in Table 6.5. A uniform leaflet thickness of 400 µm
(as measured in the final parts) was used in the simulations. The geometry was meshed
using 184,218 elements. Hard edge sizing was used to produce bricks in the leaflet, with
40 elements through the leaflet’s thickness, and tetrahedral elements in the bulk of the
stent.
To indicate the distribution of orientation across the leaflet and through its thickness,
the orientation fields are shown at 6 depths (0, 40, 80, 120, 160, and 200 µm) within each
208 Chapter 6. Valve fabrication
Figure 6.25: Schematic of inflows and outflows from the valve cavity, used experimentally
and computationally. The inset details the gate (inlet) positions: orange indicates
commissural (C) injection, the green indicates Free Edge Centre (FEC) injection, and
Free Edge (FE) injection occurs through the combination of orange , green and pink .
The outlet is indicated by the cyan region.
Figure 6.26: 184,218 elements were used to model the flow through the valve cavity.
Bricks were used to model the thin surface of the leaflet, while tetrahedral elements were
used to model the stent. There were 40 mesh elements through the thickness of the leaflet
leaflet. The regions near to the wall of the cavity are referred to as skin regions, while
the central regions are referred to as the core.
Chapter 6. Valve fabrication 209
6.5.1 Results
The polymer flows from the commissures in the C injection scenario (Figure 6.27). The
melt then simultaneously flows down the stent posts and grows from the sides of the
leaflet towards the centre. In the centre of each leaflet the flows from the commissures
meet, before flowing towards the nadir, resulting in shear-induced vertical orientation in
the centre of the leaflet. In the core regions of the leaflet there is some stretch-induced
orientation (as indicated by the presence of blue vectors in Figures 6.27e and 6.27f). In
Figure 6.27e stretching arises near to the commissures as the flow diverges to flow along
the free edge and down to the nadir. In Figure 6.27f there is stretching at the base of the
leaflet, presumably due to flow passing out of the leaflet and back into both the stent post
(right and left of image) and stent cylinder (centre of image). Curiously, there are some
stretch vectors to be found in the centre of the leaflet. The high magnitude of stretch
- relative to the magnitude of shear - is likely to arise due to the convergence of flows
from the free edge as they are redirected down the leaflet. As such, it is highly unlikely
that cylinders would actually be oriented horizontally in this region, as the principal
stretch direction is in fact in the direction of flow, highlighting a shortcoming of the
scalar orientation model.
The SAXS map of a representative leaflet injected from the commissures is shown in
Figure 6.27g: in these images biaxial orientation appears as a cross, as opposed to the
layer by layer representation in Figure 6.27. Biaxial orientation is present in the SAXS
map of the leaflet, though rarely are the two orientations perpendicular to one another.
There appears to be a dominance of horizontal orientation at the sides of the leaflet,
which was not the case in the computational predictions. SAXS scanning perpendicular
to the leaflet is unable to determine whether orientation has arisen due to shearing or
stretching deformations. It is interesting to note that the modelling predicts a region of
horizontal orientation near the base of the leaflet (Figure 6.27f), which is also present in
the SAXS map of the leaflet.
The prediction of orientation as a result of injection from the free edge is show in Figure
6.28. In this case the flow travels vertically down the leaflet, gradually fanning out to
enter the stent posts. This fanning does result in stretching deformations which led
to horizontal orientation in the core regions of the leaflet (6.28e and 6.28f). The large
inflow area provided little resistance to flow in comparison to the FEC scenario, and as
control of the injection rate was limited, there was a relatively high flow rate through the
valve (see Table 6.5). Higher flow rates promote orientation in the direction of flow, and
thus at lower flow rates it may be possible to produce more circumferential orientation.
Regardless, the orientation at the sides of the leaflet is dominated by shear at all depths.
210 Chapter 6. Valve fabrication
(a) 0 µm (b) 40 µm (c) 80 µm
(d) 120 µm (e) 160 µm (f) 200 µm
(g)
Figure 6.27: (a) to (f) Predicted orientation at various depths through the leaflet (total
thickness 400 µm) for the C injected valve. The tick marks have been normalised, but
as the plot is a 2D projection of 3D space some appear shorter than expected. The red
marks correspond to cylinders oriented by shear deformations, and blue to orientation
by extension. (g) is a representative SAXS maps of the orientation in injection moulded
leaflets. PS cylinders are predominantly horizontal at the leaflet edge, but also exhibit
significant biaxial orientation near the centre.
Contrary to the predicted orientation, the SAXS map displayed in Figure 6.28g has very
little biaxial orientation. The orientation of the centre of the leaflet is predominantly
vertical, and it fans on either side, in agreement with the orientation of the skin regions,
but the stretch oriented core region does not seem to be present. On the left of the leaflet
some horizontal orientation was recorded, this column of SAXS scans may be due to the
X-ray beam passing through the stent post at a non-perpendicular incidence. The nadir
of the leaflet has a large amount of horizontal orientation, possibly arising due to X-rays
Chapter 6. Valve fabrication 211
passing through the stent cylinder.
(a) 0 µm (b) 40 µm (c) 80 µm
(d) 120 µm (e) 160 µm (f) 200 µm
(g)
Figure 6.28: (a) to (f) shows predicted orientation at various depths within the leaflet
(total thickness 400µm) for the FE injected valve. The tick marks have been normalised,
but as the plot is a 2D projection of 3D space some appear shorter than expected. The red
marks correspond to cylinders oriented by shear deformations, and blue to orientation by
extension. (g) representative SAXS maps of the orientation in injection moulded leaflets.
PS cylinders in the free edge injection results in predominantly vertical orientation in the
centre of the leaflet, fanning to horizontal at the sides.
Locating the gate at the centre of the free edge of the leaflet (FEC, Figure 6.29) led to
a flow most similar to radial flow, where the flow diverges in a thin disc. In contrast to
the perpendicular flow into the disc geometry, here the polymer melt enters the cavity
without changing its direction of flow. Therefore, the flow quickly assumes a steady
profile, producing the characteristic layered biaxial structure with approximately 50%
circumferential orientation near to the gate. As the flow grows radially, the orientation
212 Chapter 6. Valve fabrication
near to the free edge is horizontal. At all points along the leaflet-stent boundary, the
flow crosses perpendicular (though this might not be immediately apparent from the 2D
projection of the 3D leaflet).
The SAXS mapping of the FEC leaflet is shown in Figure 6.29g. The leaflet is dominated
by uniaxial orientation in the horizontal direction. Some biaxial regions are found nearer
to the free edge. The dominance of horizontal orientation may have been aided by
restrictions to the outflow near the nadirs due to the arrangement of the heaters - so
polymer had to flow towards the commissures.
In summary, the orientation model is able to predict PS cylinder orientation in complex
geometries. For each of the injection conditions the orientation of shear aligned vectors
are accurately predicted by the model. Cylinders which are oriented by stretch (and lead
to a biaxial layered structure) are also predicted. However, their exact orientation is not
always correct, nor is their fraction.
As may be apparent from the SAXS maps, referencing the maps to the leaflet was not
a trivial task. There were various confounding factors: the leaflet must be flattened to
ensure that the beam passes perpendicular to the surface. Flattening and mounting the
leaflet required the edges of the leaflet to be gripped, without the grips intersecting the
X-ray beam. To achieve this, the leaflet had to be dissected to include some of the stent
post and stent base; which are much thicker than the leaflet and hard to cut accurately.
Consequently, a reference point cannot be accurately determined on the leaflet.
For mapping of the final valve (as shown at the end of this Chapter) a purpose built
mounting frame was cut from perspex such that the leaflets can be accurately mounted
and referenced.
6.5.2 Discussion
Using the basic assumption that shear results in PS cylinder orientation in the direction of
flow, and stretching leads to orientation perpendicular to flow, the orientation distribution
was predicted for a complex form. In the skin layers (0 to 80 µm), the PS domains orient
in the direction of flow. When injection is slow, stretching dominates in the core of the
thin sections.
In this section the flow fields predicted by the FE solver were assumed to be valid: no
further pressure or velocity field measurements were made. Shear rates were in the previ-
ously tested range of rheometric data. The shortcomings of the scalar orientation model
to determine orientation were highlighted by the predicted orientation in the commissural
Chapter 6. Valve fabrication 213
(a) 0 µm (b) 40 µm (c) 80 µm
(d) 120 µm (e) 160 µm (f) 200 µm
(g)
Figure 6.29: Predicted orientation at various depths within the leaflet (total thickness
400µm) for the FEC injected valve. The tick marks have been normalised, but as the
plot is a 2D projection of 3D space some appear shorter than expected. The red marks
correspond to cylinders oriented by shear deformations, and blue to orientation by ex-
tension. (g) shows a representative SAXS map of the orientation in injection moulded
leaflets. PS cylinders in the FEC injection results in horizontal orientation throughout
the leaflet, with biaxial orientation present near to the centre and free edge.
injection scenario (Figure 6.27). At the centre of the free edge the flows from each com-
missure converge, leading to an acceleration and stretch in the direction of flow (down
the leaflet). The positively extending component of this stretch is not perpendicular to
the flow, as assumed by the scalar model, and is thus unlikely to lead to biaxiality. In
the SAXS mapping (Figure 6.27g) there is some biaxial orientation on the left and right
sides of the leaflet which is caused by the flows from the left and right commissures which
diverge before reaching the centre of the leaflet.
214 Chapter 6. Valve fabrication
Inspection of the SAXS maps indicates that the biaxial orientations do not always lie
perpendicular to one another. This demonstrates that the assumption of orthogonal
orientation is inappropriate for fully describing biaxial orientation. The Advani-Tucker
tensor orientation model (Advani, 1987) would be able to describe such behaviour, as well
as predict the quality of orientation - which I did not consider here. In Figures 6.27 to
6.29, 3D vectors in 3D space are projected to a 2D surface, making it tricky to compare
orientations between the computational prediction and SAXS maps. Future work would
quantitatively evaluate the difference in orientation between cases.
The purpose of testing various gate positions, both computationally and experimentally,
was to inform the design of the bioinspired valve. The most crucial part of a valve is the
leaflets, in which the highest diastolic stresses can be found at the leaflet-stent boundary.
Zaffora (2011) optimised the orientation of anisotropy to reduce these stresses. The
distribution of the optimised orientation was shown in Figure 1.12. In summary, near to
the free edge the optimised orientation is horizontal, and parallel to the free edge. Moving
towards the nadir, the oriented fibres form arcs across the leaflet, with each arc following
a similar shape to that of a cable suspended between the stent-leaflet boundaries.
The optimised orientation shares many characteristics with the natural leaflet, which is
shown in Figure 1.7. The natural leaflet has predominantly horizontal orientation, and
continuous collagen fibres stretch from the leaflet-stent boundary to the centre of the
leaflet where they bundle to form the fibrous Nodule of Arantius. In both the optimised
polymeric and natural valve the fibres approximate to arcs. The natural valve, however,
has another interesting characteristic. Collagen fibres also run between the horizontal
collagen bundles, approximately 45° to the horizontal. This morphology is particularly
evident in the photograph of a stained porcine valve published by Sauren (1981) (page
14). More recently, various authors have used microscope imaging (Marom et al., 2013;
Rock et al., 2014), and small angle light scattering (Sacks et al., 1997; Sellaro, 2003) to
detect the orientation of collagen bundles. Their studies on porcine valves1 also clearly
show the dominance of horizontal fibres, with further fibres passing between bundles at
45°, or oriented vertically from the nadir of the leaflet (particularly evident in the right
coronary leaflet in Rock et al. (2014)). 3 mechanisms of branching were identified (tree-
like, pinnate and fan-like), indicating that the collagen bundles branch in response to a
variety of biomechanical cues.
If the presence of biaxial orientation in the native leaflet has some mechanical purpose, it
1It is worth noting that the main difference between the porcine and human aortic valve is the relative
sizes of the cusps; the non-coronary cusp is largest in humans, while the right coronary cusp is largest
in pigs (Sim et al., 2003). No differences in collagen arrangement have been noted.
Chapter 6. Valve fabrication 215
may be advantageous to reproduce this in the bioinspired BCP valve proposed here. The
FEC gate position (with slow injection) promotes biaxial orientation of PS cylinders in
the leaflet, with the core orientation following the desirable arcs across the leaflet, and the
skin orientation perpendicular. However, there is little consensus among the literature
regarding the purpose of collagen fibre branching and biaxial orientation in the natural
leaflet. In all likelihood, collagen growth and orientation is a result of biomechanical
influences which ‘fine-tune’ the structure (to the individual) over time (Rock et al., 2014).
Given this uncertainty, it would be na¨ıve to design for biaxial orientation on a bioinspired
basis alone.
At high injection rates from a point, when biaxial orientation does not occur, the FEC
gate position would lead to orientation extending radially from the centre of the free
edge. Along the free edge of the leaflet the orientation would be horizontal, while in
the centre of the leaflet the orientation would be vertical. The diastolic load must be
transferred from the leaflet to the stent. As such, it is unsurprising to observe that in
both the optimised and native leaflets the fibres are close to perpendicular to the stent-
leaflet boundary when they cross it. This suggests that even without biaxial orientation
FEC injection would be desirable.
There are considerations beyond orientation to be made when designing an injection
moulding tool. There is an array of guidelines for tool design which often recommend
that the following considerations should be made (Cracknell and Dyson, 1993; Eastman,
2011; Proto Labs Ltd., 2010):
• To reduce sinkage and warpage;
– eliminate thick walls,
– maintain constant wall thickness,
– core out large volumes.
• Avoid wall thicknesses ≤ 0.5 mm.
• Reduce sharp corners.
• Include draft to aid part removal from tool (no vertical sides).
• Gate away from thin edges to avoid break during gate trimming.
• Maintain proper cooling of cores.
Some sources do recommend that gating into thin sections should be avoided as it leads to
high shear rates, higher injection pressures, and difficulty in filling thick sections (East-
216 Chapter 6. Valve fabrication
man, 2011). However, the proposed low processing rate (injection velocity and unit
production) negates the first two reasons, and the small part size counters the final one.
In fact, gating into the thick stent would carry the risk of not completely filling the thin
leaflet.
After injection, cooling, and part ejection the sprue must be removed from the part. Tool
design guidelines warn of the risk of part damage during this trimming step. However, as
this process will be performed by hand, and without time pressure, the risk is reduced.
Furthermore, across the leaflet the greatest stresses are found at the leaflet-stent bound-
ary, while in the leaflet centre they are low. As such, the centre of the leaflet is a less
critical region for stress reduction.
The FE injection scenario is an unconventional means of gating a tool because there
is no thin sprue of polymer which can be trimmed off. The FE injection would require
significant post-processing to remove the ‘runner’ and ‘gate’ material. On the other hand,
the C and FEC gate positions each fill the cavity from 3 locations. Consequently the
melt flow from each gate must meet and join in the part. This meeting line is referred to
as the weld line. The weld line may be a source of flaws from which cracks could grow.
In the C injection scenario the weld lines occur down the centre of each leaflet, whereas
in the FEC scenario welding takes place in the stent posts.
In an alternative injection scenario, which was not considered here, the gates could be
located close to the nadir of each leaflet. Like the FEC scenario, injection at the nadirs
would also allow control of radial and circumferential orientation in the ideal directions.
However, the sprue would be perpendicular to the leaflet’s surface, meaning that a small
nipple would be left after trimming, which may affect the dynamic behaviour of the leaflet.
Also, the radial flow geometries considered earlier also have a perpendicular sprue, which
is likely to cause the less developed orientation close to the gate. Thus, it is preferable
for the inlet flow to be in a similar direction to the cavity flow, as found for the other
gate positions.
In summary, the FEC injection allows for a large degree of control over the orientation
in the leaflet, which can be tuned to mimic the biological and optimised orientations.
The FE scenario was not able to produce the necessary horizontal orientation. The C
gate location can produce appropriate orientation in the critical regions of the leaflet (the
stent boundary), but it led to a weld line in the centre of the leaflet. Therefore, the FEC
gate position was selected for the final tool.
Chapter 6. Valve fabrication 217
6.6 Final valve injection
In Chapter 5 I proposed a final design for the polymeric prosthetic heart valve. Compared
to the prototype valves tested in Chapters 5 and 6 (Figure 6.25 for example), the final
design (Figure 5.21) is characterised by having thicker stent posts and a lower profile
(due to the reduced distance between the nadir and sewing ring). The stent posts were
thickened to reduce their deflection during diastole, and the valve’s profile was lowered
to avoid obstruction of the coronary arteries.
6.6.1 Methods
Modelling
The transient mould filling process was modelled using Autodesk Simulation Moldflow
Adviser 2015. The viscometric data was matched to the previously used Carreau-Bird
model (Table 6.3), as Moldflow adviser uses the Cross model as opposed to the Carreau-
Bird model. The values of the parameters and side-by-side comparison of the two models
is shown in Appendix D. The cavity filling was modelled using an injection rates of
5 mm3 s−1 and 95 mm3 s−1.
Manufacture
The injection moulding tool for the final valve was manufactured by Kelland Precision
Tooling Ltd. (Huddersfield, UK). The injection was performed at the Bradford Centre
for Polymer Micro and Nano Technology using a Fanuc Roboshot S-2000i30B, a servo-
hydraulic moulder, allowing for greater control of the injection rate and reducing the
likelihood of contamination - particularly from hydraulic oil.
Upon consultation with the animal house regarding pig sizes, two valve sizes were man-
ufactured to accommodate the variation in aortic root sizes. The valves were sized for
tissue annulus diameters (TAD) of 21 mm and 23 mm with supra-annular placement, such
that the TAD corresponds to the internal orifice diameter of the valve. Both valves were
designed with a leaflet thickness 0.3 mm. The injection moulding tool contained a cavity
for each valve, meaning that both valves were injected simultaneously.
The parts were injected at a ram velocity of 0.5 mm s−1, corresponding to each cavity
being filled at 95 mm3 s−1.
218 Chapter 6. Valve fabrication
The tool was heated using cartridge heaters, and contained integrated cooling channels,
for the mould to be cooled over approximately 180 s.
The leaflet orientation was mapped with SAXS using a Bruker Nanostar (at Materials
Science and Metallurgy, University of Cambridge), using the same analysis methods as
previously described.
6.6.2 Results
Modelling
The transient injection process is shown as a sequence of images in Figure 6.30. In
comparison to the prototype results presented in Section 6.5.1, the tool is not overfilled
and there is no steady flash from the mould. In comparison to conventional injection
moulding processes, the injection is slow.
(a) (b) (c)
(d) (e) (f)
Figure 6.30: Modelling of the cavity filling process for the final valve at 5 mm3 s−1. The
images are taken at 150 s intervals. The colour corresponds to the polymer’s residence
time in the cavity, with blue for short, green for medium, red for long. First, the leaflets
are filled; second, the stent posts; and finally the sewing ring. There were no differences
in the flow patterns for injection at 5 mm3 s−1 and 95 mm3 s−1.
Chapter 6. Valve fabrication 219
Manufactured valves
The surface finish quality of the final injected valves was excellent (Figure 6.31). Rem-
nants of the weld line were visible in the stent, though with no flaws or cracks visible to
the naked eye.
Figure 6.31: Final injection moulded valve.
The leaflets of the final valves were dissected from the stent body and mapped using
SAXS. The maps of 6 leaflets are shown in Figure 6.32. There was negligible biaxial
orientation in these leaflets as the injection rate was quicker than the prototype valves.
However, compared to the previous FEC injection the shear induced orientation at the
sides of the leaflet is more horizontal. In the 2D projection of the leaflets a horizontal
orientation at the sides matches the optimised distribution well. In the centre of the
leaflet the orientation is vertical. While this is contrary to the designed result, it is not
an immediate worry as the stress in the centre of the leaflet is approximately 20 to 40% of
the stress found at the sides. There would, however, be some second order effects which
would influence the stress distribution. This issue is considered in Chapter 7.
6.7 Conclusion
In this chapter I aimed to manufacture a prosthetic heart valve with bioinspired orien-
tation. In an initial study, discs of BCP were injected from a central point. Although
most authors have observed that PS cylinders orient in the direction of flow, using SAXS
I observed a layered biaxial structure where the skin layers are oriented radially and the
core circumferentially.
I proposed that the biaxial structure forms as a result of the competition between stretch-
ing and shearing deformations. To predict the velocity field during injection moulding I
characterised the melt rheology of the selected polymer, SEPS22. SAXS measurements
220 Chapter 6. Valve fabrication
0 2 4 6 8 10 12 14 16 18 20
−2
0
2
4
6
8
10
12
x position / mm
y
p
os
it
io
n
/
m
m
0 2 4 6 8 10 12 14 16 18 20
−2
0
2
4
6
8
10
12
0 2 4 6 8 10 12 14 16 18 20
−2
0
2
4
6
8
10
12
0 2 4 6 8 10 12 14 16 18 20
−2
0
2
4
6
8
10
12
0 2 4 6 8 10 12 14 16 18 20
−2
0
2
4
6
8
10
12
0 2 4 6 8 10 12 14 16 18 20 22 24
−4
−2
0
2
4
6
8
10
12
Figure 6.32: SAXS mapping of six leaflets from injection moulded valves.
were made upon injection moulded discs to determine the proportion of radial orientation.
Using the velocity field and SAXS measurements, a critical value of the ratio of shear
rate to stretch rate was determined, which could be used to predict biaxial orientation.
A bench-scale injection moulding process was built to test various gate positions from
which to inject the heart valve. Three scenarios were tested: injection along the free edge,
from the commissures, and from a point at the centre of the free edge (FEC). Injection
from the FEC position was most successful at promoting biaxial orientation at low flow
rates.
In the final valve design, the FEC gate position led to the bioinspired orientation of PS
cylinders in the critical side regions of the leaflet, as characterised by Figure 6.32.
Chapter 7
Valve testing
7.1 Introduction
Successful design of a novel heart valve is characterised by lifelong durability of the
valve and without the need for anticoagulant therapy. These characteristics informed
the search for and selection of new materials in Chapter 2. Physiological opening and
closing of the valve is necessary for the avoidance of anticoagulant therapy. In Chapter
5 I used hydrodynamic testing to compare two prototype leaflet designs, resulting in the
selection of the one with a spherical leaflet form. I hypothesised that lifelong durability
might be aided by a bioinspired orientation of stiff PS cylinders in the leaflets of the
prosthetic polymeric valve. In Chapter 6 I designed a novel injection moulding process
which achieved this orientation.
In Chapters 3 and 4 the material (isolated from the valve) was subject to chemical
and biological forces. Chapter 3 showed that the material’s biocompatibility was good.
Chapter 4 showed that the material calcified less than a biological reference, but that the
combination of calcification and mechanical fatigue may be a failure route, which should
be considered in further testing.
A point has now been reached at which it is appropriate to test the final valve, for the
characteristics of lifelong durability without the need for anticoagulant therapy. In this
chapter hydrodynamic and durability testing will be performed.
221
222 Chapter 7. Valve testing
7.2 Hydrodynamics
In general, adequately sized trileaflet valves with good hydrodynamics and made from
haemocompatible materials do not require anticoagulation. As the haemocompatibility of
the coated and uncoated materials is adequate (Chapter 3), the hydrodynamic parameters
of the final valve design will be tested with respect to ISO 5840 in this section.
7.2.1 Materials and methods
The same pulse duplication methods as described in Chapter 5 were used to assess the
hydrodynamic performance of the valve. These tests were performed by Mr de Gaetano
at Politecnico di Milano, Milan, Italy.
Valves were manufactured from SEPS22, as described in Section 6.6. Unfortunately, the
injection moulding tool resulted in valves with too thin leaflets. In particular, the leaflets
were specified to have a uniform thickness of 300 µm. Using a micrometre the minimum
leaflet thickness was 240µm at one of the stent-leaflet boundaries. Using a microscope
the minimum leaflet thickness at the commissure was 150 µm. Nevertheless the valves
were made sufficiently well for meaningful hydrodynamic testing. Over ten valves were
made and three were used for hydrodynamic testing.
7.2.2 Results
The effective orifice area, EOAG (Equation (5.3)); the mean systolic pressure difference,
MSPD; and regurgitation of the valve were evaluated, over a minimum of 10 cycles for 3
valves. The results are reported in Table 7.1.
The final valve design exceeded the hydrodynamic ISO 5480 standard for a 23 mm valve.
Compared to the prototype valves previously tested in Chapter 5, the final valve design
had a greater EOAG and lower regurgitation than both the spherical and cylindrical form
prototype valves.
The hydrodynamic performance was probably aided by the unintentionally thinner leaflets
which allowed the leaflets to open and close more easily. At the conclusion of Chapter 5,
the valve design had been updated by increasing the thickness of the stent posts. This
change resulted in less deflection of the posts.
Chapter 7. Valve testing 223
Table 7.1: Valve performance and minimum ISO performance parameters, at a cardiac
output of 5 l min−1, 70 BPM, and 120 mmHg. Results are reported as mean ± range.
Parameter
23 mm polymeric
valve (n=3)
ISO 5480
standard
EOAG (cm
2) 1.59± 0.02 ≥ 1
MSPD (mmHg) 6.83± 0.18
Regurgitant volume (ml/stroke) 5.06± 0.61
Regurgitation (%) 8.71± 0.99 ≤ 10
7.3 Valve fatigue testing
The long term survival of a prosthetic heart valve can be tested under in vitro conditions.
In accordance with the ISO 5840 standard, rigid valves must survive 400 million cycles,
and flexible valves must survive 200 million cycles, equivalent to approximately 11 and
5.5 years, respectively, in a human.
As discussed in Section 2.1.5, it is possible to reduce the time for fatigue testing by
increasing the frequency. However, the maximum testing frequency is dependent on the
type of valve and tester. In particular, the valve should open and close completely, leaflet
dynamics should be physiological, and viscoelastic materials should not be tested above
the frequency at which they undergo a β or glass transition (20 or 80 Hz for SIBS30
when tested in Section 2.1.5). Testing performed at a frequency between 60 and 200
BPM is referred to as quasi-real time durability testing, and a pulsatile tester (as used
for hydrodynamic testing) can be used. Accelerated fatigue testing takes place at higher
frequencies (300 to 1800 BPM) using specialised accelerated testers.
7.3.1 Materials and methods
For this valve, quasi-real time durability testing was performed using the pulsatile tester
as previously described (Section 5.2.3). Tests were performed using the pulse duplicators
by me at the Department of Chemical Engineering and Biotechnology, University of
Cambridge, UK; and by Francesco de Gaetano at LaBS, Politecnico di Milano, Milan,
Italy. The tests were performed at frequencies of 135 and 200 BPM, and at a cardiac
output of 5 l min−1, as detailed in Table 7.2.
ISO 5840 recommends that sinusoidal oscillations are used in an accelerated fatigue tester.
In the quasi-real time test, a modified flow curve was used such that the diastolic period
lasted for 50% of the cycle, and the pressure variation was approximately sinusoidal at
224 Chapter 7. Valve testing
the valve.
Survival of a valve is defined as the valve supporting a transvalvular pressure gradient
during diastole (TPGd) of at least 100 mmHg, for at least 10% of the cycle period and
for at least 95% of the number of cycles survived.
7.3.2 Results
The pulse duplicator flow loops (particularly the Cambridge setup) were not well suited
for use at 200 BPM, which led to a diastolic overpressure. I recorded the maximum
transvalvular pressure that was reached during diastole, (TPGd)MAX , as well as the
transvalvular pressure supported for 10% or more of the cycle, (TPGd)10, according to
ISO 5840. The large diastolic overpressure led to the large difference between (TPGd)MAX
and (TPGd)10. The large distance between the mitral valve and ventricle probably con-
tributed to the overpressure.
All valves failed having performed fewer than 200 million cycles, thus failing ISO 5840.
Failure occurred when the valves were unable to support the diastolic pressure difference,
due to tears in the leaflets, or holes forming in the leaflets, as shown in Figure 7.1. The
diastolic overpressure differed between testing systems. Higher pressures were associated
with reduced valve lifetimes.
(a) (b) (c) (d)
Figure 7.1: Images of failed valves. Valves (a), (b), and (d) failed at the leaflet stent
boundary, and valve (c) failed with a hole at the centre of one leaflet. Images (a) to (d)
correspond to valves B, C, D, and F in Table 7.2.
7.3.3 Discussion
The injection moulding tool, and thus the valves which it produced were not within
tolerance. In particular, the leaflets were specified to have a uniform thickness of 300 µm.
Using a micrometre and light microscopy, the actual leaflet thickness was measured.
With the micrometre, the minimum thickness was 240 µm, on one of the stent-leaflet
Chapter 7. Valve testing 225
Table 7.2: Quasi-real time durability testing results for SEPS22 valves. TPGd is the
transvalvular pressure gradient during diastole, where 10 indicates that TPGd is sustained
for at least 10% of the cycle, as per the ISO standard, and MAX is the maximum TPGd
in a cycle.
Valve
Testing
Location
Size
(mm)
(TPGd)10
(mmHg)
(TPGd)MAX
(mmHg)
Frequency
(BPM)
Lifetime (106
cycles)
1 Milan 23 92 163 200 2.0
2 Milan 23 100 190 200 4.6
3 Milan 23 111 172 200 3.5
B Cambridge 23 120 150 135 0.39
C Cambridge 23 145 190 200 0.58
D Cambridge 23 130 185 200 2.0
F Cambridge 21 145 373 200 1.7
boundaries. By light microscopy the minimum thickness was 150µm. The microscope
measurement is likely to be a more accurate value, assuming that the minimum thickness
was found at the commissures. The micrometre was less sensitive to fine spatial variations
in thickness as it had a diameter of 8 mm.
Testing was performed in distilled water, so calcification did not accelerate failure. The
diastolic overpressure may have contributed to the acceleration of failure.
The remainder of this chapter will analyse the cause of failure, and possibilities for future
improvements.
7.4 Failure analysis
In Chapter 2, fracture mechanics methods were used for the comparative assessment of
material fatigue. Previously, and in parallel, Zaffora (2011) and Serrani et al. (submitted)
have used FE modelling to evaluate valve designs and to optimise PS cylinder orientation.
In this section, experimental fatigue and FE modelling will be combined to help determine
the cause of failure. Materials, processing and leaflet form shall be considered as I seek
to determine whether the durability of the valve could achieve a level of acceptability.
7.4.1 FE modelling
The FE model was constructed and run by Dr Serrani. The FE model was previously
validated by comparing the deflection of the stent posts to pulsatile testing results. The
226 Chapter 7. Valve testing
modelling of the valve was performed using an anisotropic hyperelastic model without
viscoelasticity. The base of the sewing ring was constrained vertically and the outside
edge was constrained radially.
Dr Serrani used the Mooney-Rivlin model for strain energy, ψ. A derivation of this model
is shown in Appendix E. Assuming incompressibility:
ψ = c1(I1 − 3) + c2(I2 − 3) + k4 log2
√
I4, (7.1)
where c1, c2, and k4 are material parameters, and k4 is associated with the orientation.
I1, I2, and I4 are the 1
st, 2nd, and pseudo-4th (Equation (E.5)) invariants of the Cauchy-
Green stress tensor, C. The material parameters were fitted to the experimental data,
which was previously presented in Section 2.3.3 for SEPS22, to give the values shown in
Table 7.3.
Table 7.3: Mooney-Rivlin parameters optimised for the tensile strain energy density of
SEPS22
c1 c2 k4
Anisotropic 0.15 0.15 1.62
A comparison of the strain energy density as a function of strain for the computational
and for the fatigue model is shown in Figure 7.2a. Unfortunately, the Mooney-Rivlin
formulation has no damage, Mullins, or memory components, meaning that it is not a
good fit to the experimental data.
The stress strain data for the fatigue model was collected using oscillations at increasing
steps of maximum strain. Higher strains probably cause breakage of the PS cylinders,
so that the cyclic stress strain curves that reach higher maximum strains do not pass
through the same points as those that reached lower maximum strains. This “strain
softening” or breakage is reversible, but only once samples are left for at least an hour to
recover.
Although Mooney-Rivlin is widely used for elastomeric materials, it is impossible for it to
fit the experimental data shown here. The differential of the strain energy with respect
to strain should give the stress-strain relationship. However, the best fit line to the
experimental data for strain energy density when stretched parallel to the cylinders (solid
red) is nearly linear with respect to strain, indicating that stress would be independent
of strain.
The leaflet stress is approximately equal regardless of the constitutive model that is used,
because the leaflets must support the same pressure load (assuming that the leaflets are
Chapter 7. Valve testing 227
the same thickness). The effect of the difference between the modelled and actual strain
energy density in the valve will be that regions of low stress will, in reality, be stiffer
and have a higher strain energy density than the model predicts, as shown by the plot of
strain energy density against stress (Figure 7.2b). Future work will need to implement
damage or memory components to account for this effect.
The geometry of the valve in the model was modified to account for the tooling inaccu-
racies. In the actual valve geometry the leaflet thickness reached a minimum of 150 µm
at the stent leaflet boundary, as indicated in Figure 7.2c. The actual orientation of PS
cylinders was used, as mapped using SAXS and shown in Figure 6.32. The designed
valve geometry, where leaflet thickness was a constant 300 µm, was also modelled with
the actual and optimised orientations, for comparison.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
0
0.2
0.4
0.6
0.8
1
1.2
1.4
Strain
S
tr
ai
n
E
n
er
gy
D
en
si
ty
(M
P
a)
Model L
Model R
Experimental L
Experimental R
(a)
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Stress (MPa)
S
tr
ai
n
E
n
er
gy
D
en
si
ty
(M
P
a)
Model L
Model R
Experimental L
Experimental R
(b)
(c)
Figure 7.2: Comparison of the Mooney-Rivlin formulation for strain energy density and
experimental data with a quadratic fit. (a) shows strain energy density against strain,
and (b) against stress. L indicates stretching parallel to the PS cylinder orienation; R,
perpendicular. (c) Axial view of 3D CAD model of valve marked with the measured
leaflet thicknesses (mm) at the commissures from the actual valve geometry.
228 Chapter 7. Valve testing
Results
The results of the FE simulation are shown in Figure 7.3. The von Mises and strain
energy density maxima were located at the leaflet-stent boundary, and at the centre of
the leaflet which was folded. These maxima support the failure locations of the actual
valves shown in Figure 7.1.
Figure 7.3: Results of the FE simulation showing von Mises stress during diastole in the
final valve when loaded to 60 mmHg. The maximum stress is marked for each leaflet,
1.29 MPa in the lower left leaflet, 1.22 MPa in the lower right, and 0.87 MPa in the upper
right. The maximum stress is located at the leaflet-stent boundary in the right leaflets,
and in the centre of the left leaflet. This figure was produced by Dr Serrani.
The maximum predicted strain energy density in the valve as a function of the applied
pressure is shown in Figure 7.4. The actual valve FE model did not converge up to
120 mmHg, and so higher pressure results could only be obtained by extrapolation. A
quadratic fit was used for the extrapolation, which is appropriate given that stress in the
leaflet is directly proportional to the transvalvular pressure (if there are no large leaflet
deformations), and there is an approximately quadratic relation between stress and strain
energy density for these materials.
Chapter 7. Valve testing 229
0 10 20 30 40 50 60 70 80 90 100
0
0.05
0.1
0.15
0.2
Applied backpressure (mmHg)
S
tr
ai
n
en
er
gy
d
en
si
ty
(M
P
a)
Real O, Actual G, Centre
Real O, Designed G
Optimised O, Designed G
Real O, Actual G, Side
Figure 7.4: Maximum strain energy density in the valve as a function of applied pressure
in the FE simulation. Orientation is abbreviated to O, geometry is abbreviated to G.
Centre and side refer to the elements containing the maximum strain energy density at
the centre and at the side of leaflet, respectively, as shown in Figure 7.3. Simulations
were performed by Dr Serrani.
7.4.2 Fracture mechanics to crack nucleation
In Chapter 2, fracture mechanics was used for the comparative assessment of material
fatigue, but for absolute predictions of fatigue other parameters must also be considered.
Mechanical fatigue failure in polymers occurs through the growth of flaws to become
cracks which then cause catastrophic tearing. When strained uniaxially, crack growth
rate is determined by the size of the crack, the material properties, and strain energy
density. Lifetime prediction based on fracture mechanics methods assumes that:
1. Flaws grow independent of one another,
2. The growth of naturally occurring flaws is determined by the same parameters that
determine the growth of artificial cracks in test pieces,
3. The flaws have a high enough population density such that a flaw above the critical
size is likely to be found in a region of high stress (Kingston and Muhr, 2012).
The physical validity of the assumptions has been assessed using micro computed to-
mography imaging (Le Gorju Jago, 2008). Flaws initially grow independently, before
coalescence causes rapid increases in the size of cracks. Furthermore, stress concentra-
tions may not necessarily occur in the same region as the largest flaws. As it would be
impossible to know the precise 3D description of all flaws and compute their growth in a
230 Chapter 7. Valve testing
FE model, fatigue life predictions use an effective flaw size parameter.
In Chapter 2 an industry standard effective flaw size of 40 µm was used. To determine
the effective flaw size in SEPS22 samples, a crack nucleation test is used. This consists
of cyclically straining tensile bar samples without any pre-cracks until they fail.
First, I investigated the role of processing and orientation on material fatigue. The crack
growth tests were performed on isotropic (solvent cast) samples, and perpendicular to the
PS cylinder orientation to supplement the measurements made parallel to the PS cylinder
orientation in Section 2.3.5. Second, I determined the actual effective flaw size in SEPS22,
by performing crack nucleation tests on oriented samples (parallel and perpendicular),
and solvent cast samples. Third, the crack growth parameters were evaluated when the
polymer was submerged in buffer solution. These experiments were used to construct an
absolute fatigue model, which was compared to the FE modelling and experimental valve
fatigue data.
Materials and methods
Samples of oriented SEPS22 were manufactured via compression moulding, as previously
described (Section 2.2). SEPS22 sheets were also formed using solvent casting. SEPS22
pellets were dissolved in toluene at 10w/v% before drying in a flat-bottomed Teflon®
coated vessel.
The macroscopic crack growth rates for SEPS22 R (perpendicular) and SEPS22 SC (sol-
vent cast - isotropic) were determined as previously described (Section 2.2.4).
Tensile bar samples (with dimensions shown in Figure 2.10a) were cut from the polymer
sheets parallel (L) and perpendicular (R) to the PS cylinder orientation. PS cylinders
were randomly oriented in the solvent cast (SC) samples.
The tensile bars were then cyclically strained using the Texture Analyser (Stable Mi-
crosystems, Surrey, UK). Nine samples could be simultaneously tested using wide, rubber
coated grips. Initial flaws grew to form cracks and caused failure in the tensile bars.
For the assessment of crack growth rate in representative environmental conditions, the
crack growth setup was modified by Eugenia Biral, under my supervision. The new setup
allowed samples to be strained and crack position monitored whilst immersed. The crack
growth rate was measured for solvent cast samples of SEPS22 in tris-buffered solution
(Fisher Scientific, Loughborough, UK).
Chapter 7. Valve testing 231
Results
The power law model was fitted to the crack growth data for SEPS22 L, R, and SC. The
power law coefficients, B, and exponents, F , are reported in Table 7.4.
Equation (2.14) is used to connect the macroscopic fracture mechanics model with crack
nucleation in the tensile bars. Assuming a final crack size for failure of 300 µm (the
tensile bar and leaflet thickness), the effective initial flaw size for use in Equation (2.14)
was calculated. This is reported in Table 7.4.
The effective flaw sizes of SEPS22 ranged between 253 and 291 µm. Natural rubber
is reported to have an effective flaw size of approximately 70 µm (Zarrin-Ghalami and
Fatemi, 2012).
A comparison of the fracture mechanics and crack nucleation lifetime predictions is shown
in Figure 7.5. This figure demonstrates that failure is not isotropic for the range of strain
energy densities shown. At a given strain energy density, a BCP strained parallel to
the PS cylinders has a greater lifetime than when strained perpendicular. This means
that optimal orientation in a polymeric prosthetic heart valve has two benefits: the strain
energy density is reduced (Section 1.10.1) and the lifetime at a given strain energy density
is greater.
The fracture mechanics lifetime prediction for the parallel samples has a steeper gradient
in the log-log plot than the other material. This is likely to be a result of the strain-
softening of the sample. The samples have a “memory” of the maximum strain that they
have reached, consistent with the Mullins effect (Diani et al., 2009). The lifetime of the
isotropic samples is greater than expected, probably due to the low flaw size associated
with this processing method.
Table 7.4: Fracture mechanics crack growth parameters and calculated effective flaw
sizes, based upon a af = 300 µm. NR refers to natural rubber, with the values shown
below reported by Zarrin-Ghalami and Fatemi (2012). The crack growth parameters for
SEPS22 SC in buffer were determined by Eugenia Biral.
Material F B (a0)af=300 (µm)
SEPS22 L 3.83 8.3× 10−6 291
SEPS22 R 1.8 1.84× 10−5 253
SEPS22 SC 2.51 9.87× 10−6 260
SEPS22 SC
(in buffer)
2.55 1.35× 10−5 -
NR 2 4× 10−5 70
232 Chapter 7. Valve testing
10−1 100
102
103
104
105
106
107
Strain energy density (MPa)
C
ra
ck
n
u
cl
ea
ti
on
li
fe
(c
y
cl
es
)
SEPS22 L
SEPS22 R
SEPS22 SC
NR
SEPS22 L Nucleation
SEPS22 R Nucleation
SEPS22 SC Nucleation
Figure 7.5: Comparison of the fracture mechanics and crack nucleation lifetime pre-
dictions as a function of strain energy density in a uniaxial tensile bar. Crack growth
parameters and flaw sizes are those listed in Table 7.4, and the NR model is based upon
the data collected by Zarrin-Ghalami and Fatemi (2012) for natural rubber. Each crack
nucleation point corresponds to the mean lifetime and standard deviation of 9 samples.
A comparison of crack growth rates for SEPS22 SC in air and buffered solution is shown
in Figure 7.6. There was no significant difference between the crack growth rates in the
different environments. The values of B and F are reported in Table 7.4.
The lifetime of the valves which underwent fatigue testing was then compared to the result
of the FE modelling and fatigue prediction, as shown in Figure 7.7. The predicted lifetime,
when the maximum strain energy density is calculated based upon the valve being loaded
(TPGd)10, is less than an order of magnitude different from all the experimental results.
The stress-strain model for the material does not contain strain softening terms, which
accounts for some of the discrepancy in experimental and model data in Figure 7.2a.
Furthermore, the fatigue crack model assumes that a critically sized flaw can be found
in the high stress region. The small region of high stress decreases the likelihood of this
occurring, though future analyses could make use of Weibull analysis to determine the
probability of a critically sized flaw occurring in the volume of interest.
Chapter 7. Valve testing 233
100 100.1 100.2 100.3 100.4 100.5 100.6 100.7 100.8 100.9 101
10−5
10−4
10−3
10−2
Strain energy release rate (kJ m−2)
C
ra
ck
gr
ow
th
ra
te
(m
m
/c
y
cl
e)
Air
Buffer
Figure 7.6: A comparison of crack growth rates for SEPS22 when in air and when sub-
merged in tris-buffered solution (data collected by Eugenia Biral).
Discussion
With reference to Table 7.4, there is a large disparity in the initial flaw size between
the value for natural rubber reported by Zarrin-Ghalami and Fatemi (2012), and that
calculated for SEPS22. Furthermore the effective flaw size in these experiments did vary
depending on the value of the failure crack size (af ), to which it was similar. Zarrin-
Ghalami and Fatemi (2012) on the other hand reported that the results were relatively
unaffected by the value of af , however they performed their crack nucleation testing with
1 mm thick samples. Natural rubber is the best characterised elastomer, but it can behave
quite differently to other elastomers. In particular, natural rubber strain crystallises
which improves the fatigue life. It also has a good fatigue life without requiring filler
particles.
Kingston and Muhr (2012) performed macroscopic crack growth and crack nucleation
studies on natural and styrene-butadiene rubber containing filler particles. In their exper-
iments, unfilled natural rubber had a small and invariant effective flaw size (10− 50 µm),
as also observed by (Choi and Roland, 1996; Lake and Thomas, 1967), whereas the ef-
fective flaw size of the filled natural and styrene-butadiene rubbers varied with applied
strain. At large strains (> 200%) the effective flaw size was greater than 200 µm, though
at low strains it was similar to the invariant natural rubber flaw size, 10− 15 µm. They
concluded that when fatigue life is low, steady-state crack growth rate may not be a valid
234 Chapter 7. Valve testing
10−1.6 10−1.4 10−1.2 10−1 10−0.8 10−0.6 10−0.4
104
105
106
107
108
109
1010
Strain energy density (MPa)
S
u
rv
iv
al
(c
y
cl
es
)
Experimental valve life (W from FE model)
Power law fracture mechanics SEPS22 L
Theoretical threshold W
ISO 5840 - flexible leaflet min. life
Figure 7.7: The power law SEPS22 fatigue model plotted together with the predicted
maximum strain energy density in the failed valves as predicted from the FE model.
Each valve consists of two points: the strain energy density was calculated based upon
(TPGd)10 and upon (TPGd)MAX . The ISO 5840 standard for the minimum valve lifetime
is plotted, as well as the theoretical threshold strain energy density, based upon T0 =
0.02 kJ m−2, kg = 2.5, and a constant effective flaw size of 250µm (see Section 7.4.2 for
more information).
assumption for filled materials (Kingston and Muhr, 2012).
Despite not being a “filled” rubber - in the conventional sense of the word - the PS
BCP tested here have large effective flaw sizes. Such behaviour has been associated
with conventional filled elastomers at high strains. A full fatigue model would require
a non-steady state crack growth, and a greater range of strain energy densities to be
considered.
It was not possible - in a reasonable time period (less than 1 month) - to achieve statisti-
cally significant measures of fatigue life greater than 5× 105 cycles in a crack nucleation
study. Thus it was not possible to validate whether the BCPs had a smaller effective flaw
size at lower strain energy densities. Further work would require the testing equipment
to be able to operate at higher frequencies.
In future, the crack growth rate and crack nucleation tests will need to be run at lower
strain energy densities. This would move out of the power law region of polymer crack
growth, and into region (2), according to Figure 2.8. When the tearing energy is less
Chapter 7. Valve testing 235
than T0, and in the absence of ozone, the crack growth rate tends to zero, and fatigue
life becomes infinite. In region (2), crack growth follows:
dc
dN
= A
(
Tmax − T0
Tref
)
(7.2)
where Tref is generally taken to be 1 kJ m
−2 (Mars and Fatemi, 2003), and A is a material
parameter. T0 is thus the maximum tearing energy for theoretical infinite fatigue life.
Physically, T0 corresponds to the energy required to break monomer-monomer bonds
at the crack tip, and is controlled by the number of monomers between crosslinks and
the monomeric bond strength. Experimental data found T0 to be between 0.02 and
0.07 kJ m−2 for a variety of rubbers. The similarity between materials was probably due
to the monomer bonds being of the same type (C−C) (Lake and Thomas, 1967). If the
threshold tearing energy, T0 = 0.02 kJ m
−2, is converted to a uniaxial strain energy density
using Equation (2.12), the minimum threshold strain energy density can be calculated.
Assuming that the geometric parameter kg = 2.5 and the effective flaw size was constant,
250µm, gave W0 approximately equal to 0.03 MPa. This value provides a theoretical
lower limit for the strain energy density for infinite fatigue life, and is plotted on Figure
7.7.
In Figure 7.7, the experimental valve lifetimes are all greater than the predicted lifetimes,
regardless of whether (TPGd)10 or (TPGd)MAX is used to predict the strain energy
density in the valve. This difference cannot be attributed to the difference between the
constitutive relation used in the FE model and experimental data, as shown in Figure
7.2b, as this disparity predicts that the strain energy density in the FE model should
underestimate the actual strain energy density.
The disagreement is likely to be a result of the difference in volume between the tensile
bars used to determine the effective flaw size and the region of high stress along the stent-
leaflet boundary. The likelihood of finding a flaw above the critical size is much greater
for the tensile bars as they had a volume of approximately 60 mm3, where as the elements
along the stent-leaflet boundary occupied a maximum of 3 mm3. I also attributed the
longer-than-expected valve lifetimes to the smooth finish of the valve moulding tool in
comparison to the mould used to form the compression moulded tensile bars for fatigue
testing.
The power law model predicts that the ISO standard may be reached for this valve
design with a strain energy density greater than T0. The predicted maximum strain
energy density for the designed valves with actual and optimised orientation predicts a
lifetime of 3× 107 and 1× 108, respectively. Though closer to, these do not exceed the
236 Chapter 7. Valve testing
ISO standards, so indicate that, if the fatigue models are correct, further optimisation of
leaflet form and thickness are necessary. Generally, the fatigue curves diverge (upwards)
from the power law model before reaching T0, increasing the likelihood that the valves
may survive.
In Chapter 2 material cyclic fatigue was compared based upon the power law model.
Figure 7.7 shows that the power law model may be valid when the ISO standard is
reached (noting that the model has not yet been validated past 5× 106 cycles). This
indicates that it was appropriate to use the power law model to compare fatigue life to
select materials to produce for successful heart valves.
I found it surprising to discover that the effective flaw size was so large for the BCP. In
conventional rubbers, such as natural rubber and styrene-butadiene rubber, the largest
flaws arise due to the presence of fillers. The BCP does not contain carbon black filler
which would normally be approximately 200µm, from which flaws can often grow to
cracks (Mars and Fatemi, 2002), and so I hypothesised that BCP flaws were due to
non-homogeneous orientation and frozen-in stresses during cooling.
To investigate these hypotheses I performed a high resolution SAXS mapping of a tensile
bar sample and polarised light stress analysis, which I followed by cyclic fatigue testing.
Using the same methods previously described, tensile bar samples with parallel orienta-
tion were scanned using the Bruker Nanostar SAXS. Peak width was used as a measure
of the quality of orientation.
SAXS is based on scattering of X-rays from crystal planes. As such, the orientation of
planes with respect to the beam’s angle of incidence can affect the amount of scatter-
ing. Nevertheless, all three samples shown in Figure 7.8 (d, e, and f) were found to be
inhomogeneous in their orientation quality.
I also took photoelastic images of the same samples when they were strained. When
polarised light is passed through some materials the electromagnetic wave components
are resolved along the two principle stress directions and each component experiences a
different refractive index. The two components interfere, leading to a photoelastic fringe
pattern. The relative retardation of the components is a function of the stress in the
material.
The photoelastic images are shown in Figure 7.8 (a, b, and c). There is a variation in
colour across the tensile bar samples, indicating that the stress state also varies. After
imaging, the samples were cyclically strain until failure. The point at which they failed
is marked upon the images. In sample (b) the failure point is certainly at a stress
concentration. In samples (a) and (c) there may be stress concentrations associated with
Chapter 7. Valve testing 237
the failure point. Stress and orientation quality do not appear to match one another.
Future work should involve annealing the polymer to reduce frozen in stresses.
(a) (b) (c)
0 2 4
2
4
6
8
10
12
14
16
18
20
22
24
26
28
P
ea
k
w
id
th
x (mm)
y
(m
m
)
(d)
0 2 4
P
ea
k
w
id
th
x (mm)
(e)
0 2 4
P
ea
k
w
id
th
x (mm)
0
5
10
15
20
25
30
35
40
45
50
55
60
(f)
Figure 7.8: (a), (b), and (c): Birefringence of strained SEPS22 samples marked (-) with
the point at which fatigue failure occurred. Failure occurred for samples (a), (b), and
(c) after 2809, 6977, and 8104 cycles, respectively. (d), (e), and (f) are the SAXS peak
width maps of the same SEPS22 samples, unstrained and before fatigue testing. The
peak width is measured in degrees, so the maximum peak width was 60°.
Dr Stasiak performed light microscopy of solvent cast samples, and samples injection
moulded in the aluminium moulds in Chapter 6. A selection of these images are shown
in Figure 7.9. Both the compression (a) and injection (b) moulded samples have visible
surface features with dimensions of approximately 10µm. The parallel features on the
injection moulded samples probably originate from the machining of the mould. The
solvent cast material (c) appears to have a much smoother surface. The large surface
flaws on the moulded samples are likely to have increased the effective flaw size.
7.5 Conclusion
The final valve was tested for hydrodynamic and fatigue performance in vitro. The valve
had acceptable hydrodynamic performance, in accordance with ISO 5840, and was supe-
rior to the prototype valves tested in Chapter 5. The effective orifice area of the 23 mm
valve was 1.59± 0.02 cm2, and regurgitation was 8.71± 0.99%. The fatigue performance
238 Chapter 7. Valve testing
(a) (b)
(c)
Figure 7.9: Light microscopy of the surface of SEPS22 processed into sheets by (a)
compression moulding, (b) injection moulding, and (c) solvent casting. The red scale bar
indicates 100 µm.
was unacceptable for a flexible leaflet valve, as the valves only survived for approximately
4× 106 cycles at 100 mmHg.
Tears in the leaflets of the valves were concentrated at the leaflet stent boundary, where
inaccuracies in tool manufacture had led to a thinning of the leaflets. A FE model of
the designed and actual valve geometries, with the actual PS cylinder orientation used in
both cases, was used to assess the strain energy density when loaded during diastole. The
strain energy density was compared to an experimental fatigue model for the material.
Based upon a power law for the growth of cracks, I constructed fatigue models for SEPS22
strained parallel and perpendicular to the PS cylinders, and isotropic SEPS22. The
effective flaw size was determined using crack nucleation tests. The fatigue of oriented
PS BCP was anisotropic because fatigue life varies according to whether strained parallel
or perpendicular to the PS cylinders. The effective flaw size of SEPS22 was considerably
larger than that of natural rubber, possibly indicating that microphase-separating BCPs
Chapter 7. Valve testing 239
behave like filled elastomers.
Combining the absolute fatigue model and FE simulations led to a predicted fatigue
life of 5× 105 cycles for the actual valve at 100 mmHg, which was significantly lower
than recorded value, although in remarkably good agreement given the simplifications
inherent in the model. The maximum strain energy density in the designed valve was
considerably lower, and the predicted fatigue life of a correctly moulded valve could reach
1× 108 cycles. This, however, is beyond the validated range for the fracture mechanics
model.
240 Chapter 7. Valve testing
Chapter 8
Concluding remarks
8.1 Conclusions
The ideal prosthetic heart valve has lifelong durability without the need for anticoagulant
therapy. A prosthesis must contend with a complex biological, chemical, and mechanical
environment, and for a valve design to be successful, the intrinsically connected triad of
materials, processing, and valve form must work in harmony.
Since the early years of cardiac surgery polymers have been candidate materials for use
in the ideal prosthetic heart valve, but none have yet succeeded. The choice of material
informs the design, and so design optimisation must start with being able to select optimal
materials. Acceptable durability and haemodynamics in a prosthetic heart valve invokes
the use of materials that are both durable and flexible.
I presented a set of criteria based upon chemical stability and mechanical properties for
selecting materials for use in a polymeric prosthetic heart valve. A material performance
index uses the leaflet’s thickness as the design variable, while the material properties of
modulus and fatigue stress are combined in the ratio
Eflex
σ3f
, such that superior materials
have a lower performance index. For the rapid evaluation of new materials I proposed
the use of fracture mechanics methods for the prediction of the cyclic fatigue stress, σf ,
which enables evaluation of the mechanical properties in up to two days.
The method is a general and quantitative means of rapidly assessing materials for use in
a polymeric prosthetic heart valve, and could also be used for other applications requiring
both flexibility and durability. Although valve design is driven by the choice of material,
this method allows design and material to be de-coupled, making screening quicker and
more efficient.
241
242 Chapter 8. Concluding remarks
The native heart valve has a collagen architecture which contributes to its durability. The
anisotropy of collagen has inspired other researchers to embed strong fibres into polymer
leaflets, but these have failed due to delamination and dehiscence. I proposed that a set of
block copolymers (BCPs) which self-assemble at the nano-scale to form stiff PS cylinders
in a rubbery bulk and can be processed to yield mechanically anisotropic materials which
could mimic the anisotropic properties of the native heart valve.
I tested the stability of a shortlist of BCPs under accelerated degradation conditions
(ISO 10993:13). This indicated that the unsaturated BCPs (SIS18, SIS30, and SI/BS19)
were more susceptible to degradation. The saturated BCPs (SIBS30, SEBS20, SEBS29,
and SEPS22) met the in vitro, qualitative requirements of ISO 10993:13 for long term
implantable materials.
I evaluated the mechanical properties of the polymers to generate the material perfor-
mance index. The saturated BCPs could be ranked in increasing order of preference
as: SEBS20, SEBS29, SEPS22, and SIBS30. Unfortunately, SIBS30 suffered from severe
shrinkage during moulding which led to visible flaws, and so it had to be eliminated from
the shortlist. When PS cylinders were well oriented SEPS22 and SIBS30 had superior
performance indices relative to a reference polyurethane material that had previously
been tested for prosthetic heart valve leaflets. As a consequence, SEPS22 was selected
for further testing and use in valve design.
A review of the literature showed a wide range of leaflet designs for polymeric prosthe-
ses, two were selected for detailed evaluation. One of these was based upon cylindrical
surfaces (Burriesci et al., 2010), and the other based upon spherical surfaces (Zaffora,
2011). Prototype valves were injection moulded from SEPS22 and their hydrodynamics
tested in vitro. The cylindrical valve had a greater effective orifice area but unacceptable
regurgitation. The spherical valve had a lower effective orifice area, but less regurgitation
and, in total, a lower energy loss over the valve. Only the spherical form valve complied
with the prosthetic heart valve International Standard for hydrodynamics (ISO 5840),
and so this design was selected for use in the final valve. The design was improved by
increasing the thickness of the stent posts which reduced their deflection during diastole.
In terms of hydrodynamics, I acknowledge that the polymeric prostheses were inferior to
a similarly sized Perimount Magna reference bioprosthesis, which could be addressed in
future.
In the experiments that I performed, when the thickness of leaflets made from SEPS22
(without optimised orientaiton) was greater than 325 µm, all three leaflets did not consis-
tently open at flow rates greater than or equal to 2 l min−1, so the leaflets were designed
with a thickness of 300µm. The hydrodynamic results for the spherical valve also in-
Chapter 8. Concluding remarks 243
dicated that the maximum transvalvular pressure gradient during opening was strongly
correlated with the thickness of the leaflets near to the commissures. Although varia-
tions in thickness across the leaflet were beyond the scope of this thesis, this provides an
interesting direction for further optimisation of the leaflet design.
Injection moulding was proposed as a means to orient the stiff PS cylinders and thus
manufacture the prostheses with a bioinspired orientation. Thin discs of cylinder forming
BCPs were injection moulded isothermally. Using small angle X-ray scattering (SAXS), a
novel structure in which the PS cylinders formed a layered biaxial structure, was observed.
I hypothesised that this structure was formed by the force field during injection, and so
determined a constitutive and orientation model for the BCP.
I characterised the rheology of the polymer melt, and made pressure measurements on the
discs during injection moulding. The Giesekus and 3rd order fluid viscoelastic constitutive
models provided the best description of the flow and pressure field during injection. The
generalised Newtonian Carreau-Bird model also provided an acceptable description of
the velocity field. The third order model was not available in the FE software, and the
Giesekus model was not stable over the range of processing conditions, so I selected the
Carreau-Bird model to describe the velocity field. I proposed that the ratio of shear-to-
extensional deformations (Cs =
γ˙
ε˙
) could be used to predict the layered biaxial orientation
during point injection of discs, with shear resulting in orientation in the direction of flow
and extension resulting in orientation perpendicular to flow.
Nine discs were injection moulded and the ratio of radial to circumferential orientation
was assessed using SAXS. Based upon a comparison of the flow field and the SAXS
data, a constant critical value of CsC = 3 correctly predicted orientation in 85% of the
samples (by volume). In more complex shapes, such as the heart valve, the model was
an acceptable predictor of biaxial orientation.
A combination of modelling (using the Carreau-Bird model and shear-extension ratio)
and bench-scale injection moulding was used to determine a point at which to inject the
valves. Three injection scenarios were considered: injection from the commissures, along
the full length of the free edge, and from points at the centre of the free edge of each
leaflet. Injection from the centre of the free edge of each leaflet allowed a bioinspired
orientation to be produced (which was mapped using SAXS). Low flow rates led to a
layered biaxial structure which mimicked the major bundles of collagen across the leaflet
and interconnections between the bundles. At high flow rates, the critical regions of the
leaflet (at the stent-leaflet boundary) contained PS cylinders in the optimal orientation
which, according to the FE model, reduced the average and maximum stresses in the
valve relative to the isotropic case.
244 Chapter 8. Concluding remarks
The haemocompatibility of BCPs of types SI/BS19, SIS30, SIBS30 and SEPS22 were as-
sessed in comparison to expanded polytetrafluoroethylene (ePTFE), polyester (PES), and
bovine pericardium (BP) graft materials. Haemocompatibility was assessed by rotating
them in modified Chandler loops at physiological shear rates, while in contact with fresh
human blood. In measures of coagulation and thrombogenicity, the BCPs were better
than PES, but worse than ePTFE and BP. In measures of inflammation, the BCPs and
ePTFE were better than PES and BP. A heparin coating was applied to SEPS22, that
improved the material and made it superior to all of the reference materials in terms of
coagulation, thrombogenicity, and inflammation.
The durability of the heparin coating was tested by dynamically straining the material
in buffered solution for 3 weeks. XPS and water contact angle measurements indicated
that there was only a small loss of heparin from the surface. However, immersion in
an oxidising solution did lead to destruction of the coating. The direct contact viability
of murine fibroblasts (L929) upon SEPS22 and heparin coated SEPS22 was superior to
PTFE, PES, BP, and the control samples. With respect to haemocompatibility and cell
viability, uncoated SEPS22 would be satisfactory, but coating with heparin would make
it an excellent material for use in the prosthetic heart valve.
The crystallisation of calcium-phosphate on the native heart valve, bioprostheses, and
polymeric prostheses can cause their failure. The candidate BCPs were tested using a
device which flexed the materials while they were submerged in calcifying solution. The
BCPs calcified less than the glutaraldehyde-fixed pericardium reference, but calcification
was shown to increase the flexural stiffness of statically calcified polymers, and accelerate
failure of the BCPs when they were dynamically flexed.
Surprisingly, heparin coating improved the durability of SEPS22 in both calcifying and
non-calcifying solutions. Despite the high electronegativity of heparin, it also led to a
small reduction in calcification. The higher surface roughness of uncoated SEPS22 may
account for this difference. Contrary to earlier hypotheses (Imachi et al., 2001), I found
that calcium-phosphate crystals did not localise to surface cracks.
Two sizes of aortic prosthesis were manufactured by injection moulding, for use with
aortic root annulus diameters of 21 and 23 mm. The valves were tested in vitro ac-
cording to ISO 5480 with a sample size of three. The effective orifice area of the valve
was 1.59 ± 0.02 cm2, which was greater than the standard 1 cm2. Regurgitation was
8.71± 0.99 which was less than the 10% maximum. Seven valves survived approximately
4× 106 cycles at 100 mmHg, which was less than the minimum standard of 2× 108 cy-
cles. Therefore hydrodynamics were satisfactory, but durability was unacceptable. The
leaflets of the valve were thinner than intended (due to inaccurate tooling), which was a
Chapter 8. Concluding remarks 245
major contributor to the premature fatigue failure.
My study of crack nucleation indicated that the effective flaw size was several orders
of magnitude larger than expected. Large effective flaw sizes have also been recorded
for filled rubbers, suggesting that the phase-separated PS domains have an effect simi-
lar to that of filler particles. Frozen-in stresses were identified in compression moulded
polymers, and non-homogeneous orientation was also observed. Both contributed to the
concentration of stresses and growth of flaws.
By extrapolating the fatigue prediction and using the results of FE analysis, I predicted
that correctly manufactured polymeric valves could theoretically reach 3× 107 cycles.
Furthermore, if the orientation could be optimised, 1× 108 cycles could be reached. These
values are short of the ISO 5840 standard for flexible leaflet valves but, crucially, this
prediction is based upon a logarithmic extrapolation of the fracture mechanics model
so lifetimes may be greater than predicted as the strain energy density approaches the
threshold valve for crack growth. Understanding of the failure mechanisms is still at an
early stage, indicating that further work is required, first, to reduce the strain energy
density in the leaflet, and second, to validate the fatigue model at lower strain energy
densities. This work has definitely established that there is still potential for bioinspired
polymeric prostheses to overcome the issues of durability and need for anticoagulation
with further development.
8.2 Further work
The next stages in development of a polymer prosthetic heart valve should focus on
increasing durability to a level equivalent to ten or more years operation in a person.
In the following sections, I present strategies for improving durability of the valve and
SEPS22 BCP material. I will go on to detail several necessary, interesting, and potentially
fruitful avenues for investigation.
In parallel with the valve development, future work should monitor the progression of
other solutions to heart valve disease. In particular, the arrival of new oral anticoagu-
lants could reduce the drawbacks associated with mechanical heart valves. Some of the
latest oral anticoagulants have been licensed by the FDA with the potential to supersede
warfarin. The thrombin inhibitor, dabigatran; and factor Xa inhibitors, rivaroxaban,
apixaban, and edoxaban. In the case of dabigatran, it is contraindicated for use in pa-
tients with mechanical heart valves after a trial of 252 patients was halted due to an
excess of ischaemic strokes and bleeding events (Yeh et al., 2015).
246 Chapter 8. Concluding remarks
To overcome the durability issues associated with bioprostheses, new processing methods
were developed and implemented in the early 2000s. In vivo durability data is still being
collected on these valves. Improvements are anticipated and the recommended age for a
bioprosthesis could fall further below the current 60 years (Vyavahare et al., 1997a).
The greatest shift in prosthetic heart valve technology in the past decade has been to
transcatheter implantation. Most recently, TAVI has shown benefits in comparison to
surgical replacement in low-risk, as well as high-risk patients older than 70 years (Thyre-
god et al., 2015). Once safety is shown to be sufficient for younger patients to receive
TAVI, then data can be collected on the actual durability of valves. Modelling suggests
that crimping the tissue reduces the lifetime by 50% (Martin and Sun, 2015). Poly-
meric materials have the potential to overcome the issues of durability in transcatheter
bioprostheses, but the effects of crimping will need to be tested.
8.2.1 Valve design
I concluded Chapter 7 with the prediction that a valve manufactured with uniformly
thick leaflets, could allow the valve to reach 3× 107 cycles at 100 mmHg for the actual
orientation and 1× 108 cycles if the optimised orientation of PS domains were achieved.
Accelerated testing should be used to verify such long lifetimes.
In order to reach the ISO 5840 minimum lifetime, I predict that the maximum strain
energy density in leaflets made from oriented SEPS22 will need to be reduced to below
0.04 MPa. This assumes that the power law model provides a minimum limit to the
lifetime, which is a good assumption when the threshold strain energy release rate, T0,
is approached. The maximum strain energy density should also occur at a point in the
valve where the principal strain direction is aligned with the PS cylinder orientation.
In order to achieve the predicted minimum lifetime a number of aspects warrant further
investigation. The threshold strain energy density has only been calculated theoretically.
To determine this experimentally will require tensile and fatigue testing equipment that
can reach frequencies higher than 1 Hz. Further fatigue testing of the material will also
reveal whether the effective flaw size varies with strain for phase separating BCP, and
thus improve the accuracy of the fatigue model.
During my testing of the BCP in aqueous solutions, I observed under SEM a surface
mud-cracking pattern. Further investigation of this effect is recommended as it may play
a role in the formation of flaws in the material.
Strain energy density in the valve was evaluated using FE simulations, but the material
Chapter 8. Concluding remarks 247
model lacked a memory component. A more accurate lifetime prediction will be possible
once strain memory (a form of the Mullins effect) is incorporated. Leaflet thickness is the
primary determinant of strain energy density in the leaflet, and the pressure required to
open the valve. In this thesis, a maximum leaflet thickness was determined by extrapola-
tion from hydrodynamic testing, but this thickness did not result in an adequate lifetime
for the valve. Incorporating fluid-structure interactions into a hyperelastic, viscoelastic,
anisotropic FE model is a large step for modelling, but would allow the maximum leaflet
thickness to be determined.
Alternatively, improving the accuracy of the material lifetime model and valve strain
energy density constitutive equation would allow the current FE model of diastolic loading
to determine a minimum leaflet thickness, which could then be tested in vitro.
In Section 5.3.1 I evaluated the effect of variations in thickness across the leaflet on the
pressure required to open the valve. The thickness at the commissures of the spherical
valve was found to have the most influence on the valve opening. There are likely to
be strong benefits from optimising spatial variations in the leaflet thickness, which could
be performed using FE modelling, but would require a fluid-structure interaction model
needed to simulate the opening of the valve.
8.2.2 Materials and processing
In Chapter 2, the BCP SIBS30 could not be selected because of cracking and shrinkage
which occurred during melt processing. However it had the best material performance
index so its deficiencies should be investigated. If higher temperatures could be used
during injection moulding, possibly by using a nitrogen atmosphere to reduce oxidation,
it may be possible that a lengthy annealing process could reduce the shrink-associated
cracking. A longer annealing process would also be likely to benefit the SEPS22 material,
in which stress concentrations were evident.
In Chapter 6 I reported a novel biaxial orientation phenomenon. While this was not pur-
sued for the final injection, it would be of considerable interest to investigate the fatigue
properties of the layered material. Furthermore, I drew many analogies between the PS
cylinder BCP and conventional short-fibre filled composite polymers. In particular, the
biaxial orientation may also be modelled using the Tucker-Folgar equation. The Tucker-
Folgar equation contains parameters used to describe the aspect ratio of fibres and the
strength of interactions between the fibres, which I would expect to be quite different
for the BCP in comparison to conventional composites (Advani, 1987; Chung and Kwon,
2002; Park and Park, 2011).
248 Chapter 8. Concluding remarks
To achieve mechanical anisotropy in the materials used in this thesis, I employed a BCP
which phase separated. Further improvement of the materials might be possible by
investigating the association between material performance and molecular architecture.
In particular, molecular weight is a strong determinant of polymer mechanical properties,
and may influence the modulus and fatigue life differently, thus allowing the material
performance index to be improved. It would also be worthwhile to assess the performance
index of the BCP materials once they have been chemically crosslinked (Ducrot et al.,
2014).
Microphase separation in BCP leads to a form of physical crosslinking, in this case be-
tween PS domains. Crosslinking is greater in a triblock copolymer relative to a diblock
copolymer, as the links are domain-bulk-domain, rather than just domain-bulk. Consid-
ering an ABA triblock copolymer, where A forms cylinders or spheres in a bulk of B;
each single chain starts in a cylinder, enters the bulk, and may either bridge to another
cylinder, or loop to the same cylinder, as illustrated in Figure 8.1. The proportion of
chains which display bridging is approximately 37 to 50% (Karatasos et al., 2000).
Drolet and Fredrickson (2001) investigated whether the bridging fraction could be in-
creased in pentablock copolymers (ABABA) to improve BCP toughness. They per-
formed a theoretical optimisation of the bridging fraction when the size of the central
block (ABABA) is varied while keeping the total ratio of block A to B constant. As the
ratio of central block to end blocks was increased, the bridging fraction increased from
0.25 to 0.67. If SEPSEPS could be synthesized (instead of SEPS), it may be possible to
improve durability.
If crosslinking is used to improve durability, it is important to check whether the mod-
ulus is changed, and thus whether the material performance index is improved. It is
certainly possible that when some materials are crosslinked, the toughness can be in-
creased with no change in stiffness. Li et al. (2014) tuned the chain length of hybrid
alginate-polyacrylamide hydrogels to produce materials with a higher fracture energy
without changing the elastic modulus.
Much of my thesis has been based upon mimicking the anisotropy of the native valve.
In a slightly different approach, the strain dependent mechanical properties could be
mimicked. In particular, opening and closing a valve only requires small leaflet strains,
but loading a valve during diastole leads to greater strains. The native valve has a highly
non-linear stress-strain response, as shown in Figure 1.8. Once the strain reaches 8%, the
leaflet’s stiffness increases up to 300-fold. Copying this non-linear stress-strain behaviour
could produce an optimal valve.
Chapter 8. Concluding remarks 249
L
B
L
B
L
Figure 8.1: Bridging (B) and looping (L) in ABA triblock copolymers.
Textiles have been proposed for prosthetic valves as they are easy to bend, but strong
when loaded in tension. Unfortunately they failed due to textile wear and tissue ingrowth
(Heim et al., 2014). A possible alternative could be the use of liquid crystal elastomers,
which could be tuned to be isotropic when unstrained and become nematic upon strain.
For example, with the aim of making general tissue replacement materials, Agrawal et al.
(2013) coupled liquid crystal mesogens to poly(hydrogenmethylsiloxane), the materials
hardened under dynamic compressive strain.
250 Chapter 8. Concluding remarks
Appendix A
Valve development flow chart
251
252
A
p
p
en
d
ix
A
.
V
alve
d
evelop
m
en
t
fl
ow
ch
art
Figure A.1: Flow chart for the selection of a polymeric material for use in a prosthetic heart valve.
A
p
p
en
d
ix
A
.
V
alve
d
evelop
m
en
t
fl
ow
ch
art
253
Figure A.2: Flow chart for the design of a polymeric prosthetic heart valve.
254 Appendix A. Valve development flow chart
Appendix B
Derivation of performance indices for material
selection
δ
L
a
b
Figure B.1: Schematic of a uniformly loaded simple beam, with a fixed end. The beam
has length L, width a, and depth b. The deflection when loaded is δ.
Given the task of designing an aeroplane wing. The wing might be thought of as a simple
beam which is fixed at one end, and is uniformly loaded along its length, as indicated
in figure B.1. As a designer (and a passenger!) my primary concern is that the tip of
the wing does not deflect too much. A large deflection would be likely to disrupt the
aerodynamics, and in the worst case might even cause the wing to fail. The deflection
of the wing might be expressed using equation B.1, where δ is the deflection at the
unsupported end, Pw is the load which is uniformly distributed across the wing, l is the
length of the wing, E is the Young’s Modulus of the material and I is the second moment
of area of the wing, which, for a simple rectangular wing (for the sake of simplicity) would
be ab
3
12
where a is the width of the wing and b is the depth of the wing.
255
256 Appendix B. Derivation of performance indices for material selection
δ =
Pwl
3
8EI
(B.1)
Now, given the above equation, there might be several strategies available to minimise
the deflection. Let us assume that the load to be supported (our plane and passengers),
the length (maybe by the size of the hangar) and the width of the wing (aerodynamics)
are set, which leaves us with the material stiffness and the wing’s depth to alter. Our
discerning designer, and passenger, would be quick point out that the weight of the wing
is also quite important; too heavy and the wing must then support even more of its own
weight. The weight of a simple rectangular wing is:
W = ablρg (B.2)
where ρ is the density of the material and g is the constant of gravitational acceleration.
Like baggage weight limits, there is a wing weight limit and so when combined with
our previously constrained variables, the product bρ becomes equal to some constant.
Equations B.1 and B.2 can now be combined in equation B.3, with the set parameters
and variable parameters grouped:
δ =
2Pwl
6a2g3
2W 3
ρ3
E
(B.3)
Equation B.3 tells us that if compare 2 materials, if one is more dense, then a wing made
from this material must be thinner. For the thinner wing to achieve the same deflection
that material must be stiffer. If we group together the set parameters and the deflection
we have a bundle of ‘design’ terms which amount to a performance index, IW .
IW =
ρ3
E
(B.4)
Minimising IW would reduce deflection, or alternatively allow us to make our wing longer,
or support a greater load. Based upon a physical and a mechanical property of a ma-
terial we can now compare the likelihood that they would make a good aeroplane wing.
Furthermore, if we wanted to search the complete material space, we could plot all ma-
terials on a graph of logE versus log ρ. On such a graph, with logarithmic scales, we
could draw lines of gradient 3 upon which materials with equal IW would lie. The free
searching method is delightfully elegant, but we must be mindful that design can rarely
be bundled into 2 or 3 material properties, and other, often qualitative measures must
also be considered.
Appendix C
Construction of the final valve
Table C.1: Steps in the construction of the final spherical valve geometry
Sphere of radius equal to TAD
Sphere centre located on
rTAD + sin(dcom)
StentID = TAD,
stentOD = rTAD + 3
Sewing ring design, giving
maximum ESRD=33.66 mm
257
258 Appendix C. Construction of the final valve
Table C.1 continued.
The outer edge of the sewing
ring skirts the outside of the
stent with a 1 mm of clearance,
and the leaflet nadirs with 2 mm
of clearance. This minimises the
sewing ring area whilst allowing
space for suturing.
Half stent cross section design.
Leaves 1 mm radial space to the
edge of the SR. Each stent
occupies 42.6°. A minimum of
space is left between the stent
edge and the commissure to
promote opening whilst
providing a tight seal during
closing (0.1 mm).
Fillets: 0.02 at commissure, to
4 mm, to 2 mm at stent post
centre. 1 mm down radial corners
The aortic surface of the valve is
mirrored, repeated rotationally,
and then closed. The body is
then shelled to produce a
constant thickness leaflet which
extends to the stent.
Appendix D
Cross model
The Cross model was matched to the Carreau-Bird model for SEPS22. The Cross model
viscosity is expressed as:
η =
η0W
1 +
(
η0W γ˙
τW
)1−nW (D.1)
Where η0W , is the zero shear rate viscosity; τW , is the critical stress level for transition
to shear thinning; and nW , is the power law index in the Cross model. The Cross model
predicts a zero-shear viscosity and transition region, and further parameters can also be
implemented to account for the pressure and temperature variation of viscosity. To fit to
the previously used Carreau-Bird model, Cross-WLF parameters were selected as shown
in Table D.1. The Carrea-Bird and Cross model viscosities are shown in Figure D.1. For
the range shown, they match well.
Table D.1: Cross model fitting parameters for SEPS22 at 150 ◦C
η0W 3.6× 105 Pa s
nW 0
τW 5× 104 Pa
259
260 Appendix D. Cross model
10−3 10−2 10−1
0
0.5
1
1.5
2
2.5
3
3.5
·105
Shear rate (s−1)
V
is
co
si
ty
(P
a
s)
Carreau-Bird
Cross-WLF
Figure D.1: Comparison of the Carreau-Bird and Cross model viscosities.
Appendix E
Mooney-Rivlin model
Mooney-Rivlin is a hyperelastic model which can be used to describe elastomer behaviour.
The strain energy, ψ, is defined as follows:
ψ = ψ(C)− p(J − 1) (E.1)
Where C is the Cauchy-Green tensor, p is the hydrostatic pressure, and J is the volume
ratio, which is assumed to equal to 1 for an incompressible material.
The strain energy can be separated into an isotropic and anisotropic portion:
ψ(C) = ψiso(C) + ψaniso(C) (E.2)
For the isotropic part of the potential, the strain energy can be expressed in terms of its
invariants:
ψiso = ψiso(I1, I2) = c1(I1 − 3) + c2(I2 − 3) (E.3)
Where c1 and c2 are material parameters.
The anisotropic contribution of ψ is expressed as:
ψaniso = ψaniso(I4) = k4 log
2
√
I4 (E.4)
Where k4 is a material parameter, and the 4th invariant, I4 takes into account the cylin-
ders’ orientation (vector a):
I4 = a ·Ca (E.5)
261
262 Appendix E. Mooney-Rivlin model
In summary, the strain energy function is:
ψ = c1(I1 − 3) + c2(I2 − 3) + k4 log2
√
I4 − p(J − 1) (E.6)
The parameters were optimised for SEPS22 to yield the values shown in Table E.1
Table E.1: Mooney-Rivlin parameters optimised for the tensile strain energy density of
SEPS22
c1 c2 k4
Anisotropic 0.15 0.15 1.62
Isotropic 0.3 0.36 n/a
Bibliography
Advani, S. G. (1987). The Use of Tensors to Describe and Predict Fiber Orientation in
Short Fiber Composites. Journal of Rheology, 31(8):751.
Agrawal, A., Chipara, A. C., Shamoo, Y., Patra, P. K., Carey, B. J., Ajayan, P. M.,
Chapman, W. G., Verduzco, R., and Rafael (2013). Dynamic Self-Stiffening in Liquid
Crystal Elastomers Aditya. Nature Communications, 23(4):1739–51.
Ainslie, K., Thakar, R., Bernards, D., and Desai, T. (2009). Inflammatory Response to
Implanted Nanostructured Materials. In Biological Interactions on Materials Surfaces,
chapter 18, pages 355–371. Springer US.
Akat, K., Borggrefe, M., and Kaden, J. J. (2009). Aortic valve calcification: basic science
to clinical practice. Heart, 95(8):616–23.
Akins, C. W., Travis, B., and Yoganathan, A. P. (2008). Energy loss for evaluating heart
valve performance. The Journal of thoracic and cardiovascular surgery, 136(4):820–33.
Alexander, K., Clarkson, J., Bishop, D., and Fox, S. (2001). Good Design Practice
for Medical Devices and Equipment: A Framework. Technical report, University of
Cambridge and Cambridge Consultants.
Alferiev, I., Stachelek, S. J., Lu, Z., Fu, A. L., Sellaro, T. L., Connolly, J. M., Bianco,
R. W., Sacks, M. S., and Levy, R. J. (2003). Prevention of polyurethane valve cusp
calcification with covalently attached bisphosphonate diethylamino moieties. Journal
of biomedical materials research. Part A, 66(2):385–95.
Allison, B. C., Applegate, B. M., and Youngblood, J. P. (2007). Hemocompatibility of
hydrophilic antimicrobial copolymers of alkylated 4-vinylpyridine. Biomacromolecules,
8(10):2995–2999.
Amado, E. and Kressler, J. (2011). Triphilic block copolymers with perfluorocarbon moi-
eties in aqueous systems and their biochemical perspectives. Soft Matter, 7(16):7144.
Amann, K. (2008). Media calcification and intima calcification are distinct entities
263
264 Bibliography
in chronic kidney disease. Clinical Journal of the American Society of Nephrology :
CJASN, 3(6):1599–605.
Amiji, M. and Park, K. (1993). Surface modification of polymeric biomaterials with
poly(ethylene oxide), albumin, and heparin for reduced thrombogenicity. Journal of
biomaterials science. Polymer edition, 4(3):217–234.
Anderson, G. H., Hellums, J. D., Moake, J., and Alfrey, C. P. (1978). Platelet response
to shear stress: changes in serotonin uptake, serotonin release, and ADP induced ag-
gregation. Thrombosis research, 13(6):1039–1047.
Anderson, H. (1981). Normal and Abnormal Mineralization in Mammals. ASAIO Trans-
actions, 27:702–708.
Anderson, J., Cook, G., Costerton, B., Hanson, S., Hensten-Pettersen, A., Jacobsen, N.,
Johnson, R., Mitchell, R. N., Pasmore, M., Schoen, F. J., Shirtliff, M., and Stoodley, P.
(2004). Host reactions to biomaterials and their evaluation. In Ratner, B. D., Hoffman,
A. S., Schoen, F. J., and Lemons, J. E., editors, Biomaterials Science, chapter 4, pages
293–354. Elsevier, 2 edition.
Anderson, J. M. (2001). Biological responses to materials. Annual review of materials
research, 31:81–110.
Andersson, J. and Andersson, J. (2003). Complement Activation Triggered by Biomate-
rial Surfaces. Medicine.
Ando, M. and Takahashi, Y. (2009). Ten-year experience with handmade trileaflet poly-
tetrafluoroethylene valved conduit used for pulmonary reconstruction. Journal of Tho-
racic and Cardiovascular Surgery, 137(1):124–131.
Andrews, R. K., Lopez, J. a., and Berndt, M. C. (1997). Molecular mechanisms of
platelet adhesion and activation. International Journal of Biochemistry and Cell Biol-
ogy, 29(1):91–105.
ANSYS (2012). ANSYS POLYFLOW User’s Guide.
Ashby, M., Bre´chet, Y., Cebon, D., and Salvo, L. (2004). Selection strategies for materials
and processes. Materials & Design, 25(1):51–67.
ASTM International (2002). Flexural Properties of Unreinforced and Reinforced Plastics
and Electrical Insulating Materials.
Azarnoush, K., Pereira, B., Duale´, C., Dorigo, E., Farhat, M., Innorta, A., Dauphin, N.,
Geoffroy, E., Chabrot, P., and Camilleri, L. (2013). Comparison between three types
Bibliography 265
of stented pericardial aortic valves (Trivalve trial): study protocol for a randomized
controlled trial. Trials, 14:413.
Bach, D. S. (2010). Echo/Doppler Evaluation of Hemodynamics After Aortic Valve
Replacement. Principles of Interrogation and Evaluation of High Gradients. JACC:
Cardiovascular Imaging, 3(3):296–304.
Balachandran, K., Sucosky, P., and Yoganathan, A. P. (2011). Hemodynamics and
mechanobiology of aortic valve inflammation and calcification. International journal
of inflammation, 2011:15.
Bar-Cohen, Y. (2005). Biomimetics: mimicking and inspired-by biology. Smart Structures
and Materials, 5759:1–8.
Bates, F. S. and Fredrickson, G. H. (1990). Block copolymer thermodynamics: theory
and experiment. Annual review of physical chemistry, 41:525–57.
Beith, J. G. (2006). Valve.
Belanger, M.-C. and Marois, Y. (2001). Hemocompatibility, Biocompatibility, Inflam-
matory and in Vivo Studies of Primary Reference Materials Low-Density Polyethylene
and Polydimethylsiloxane : A Review. Journal of Biomedical Materials Research,
58(5):467–477.
Berlin, D. B., Davidson, M. J., and Schoen, F. J. (2014). The power of disruptive tech-
nological innovation: Transcatheter aortic valve implantation. Journal of Biomedical
Materials Research Part B: Applied Biomaterials, pages n/a–n/a.
Bernacca, G., Fisher, A., Wilkinson, R., Mackay, T., and Wheatley, D. (1992a). Calcifi-
cation and stress distribution in bovine pericardial heart valves. Journal of Biomedical
Materials Research, 26(7):959–66.
Bernacca, G., Mackay, T., and Wheatley, D. (1992b). A dynamic in vitro method
for studying bioprosthetic heart valve calcification. Journal of Heart Valve Disease,
1(1):115–30.
Bernacca, G., Mackay, T., Wilkinson, R., and Wheatley, D. (1995). Calcification and
fatigue failure in a polyurethane heart valve. Biomaterials, 16(4):279–285.
Bernacca, G. M., Haworth, W. S., O’connor, B., and Wheatley, D. J. (2009). Method of
manufacture of a heart valve prosthesis.
Bernacca, G. M., O’Connor, B., Williams, D. F., and Wheatley, D. J. (2002). Hydrody-
266 Bibliography
namic function of polyurethane prosthetic heart valves: influences of Young’s modulus
and leaflet thickness. Biomaterials, 23(1):45–50.
Bernacca, G. M. and Wheatley, D. J. (1999). Surface modification of polyurethane heart
valves: effects on fatigue life and calcification. International Journal of Artificial Or-
gans, 21(12):814–9.
Bertazzo, S., Gentleman, E., Cloyd, K. L., Chester, A. H., Yacoub, M. H., and Stevens,
M. M. (2013). Nano-analytical electron microscopy reveals fundamental insights into
human cardiovascular tissue calcification. Nature materials, 12(6):576–83.
Bezuidenhout, D. and Zilla, P. (2014). Flexible Leaflet Polymeric Heart Valves. Cardio-
vascular and Cardiac Therapeutic Devices: Studies in Mechanobiology, Tissue Engi-
neering and Biomaterials, 15:93–129.
Bird, R., Armstrong, R., and Hassager, O. (1987). Dynamics of Polymeric Liquids.
Volume 1. Fluid Mechanics. Wiley-Interscience, 2 edition.
Bird, R. B. and Wiest, J. M. (1985). Anisotropic Effects in Dumbbell Kinetic Theory.
Journal of Rheology, 29(5):519.
Bleiziffer, S., Eichinger, W. B., Hettich, I., Guenzinger, R., Ruzicka, D., Bauernschmitt,
R., and Lange, R. (2007). Prediction of valve prosthesis-patient mismatch prior to
aortic valve replacement: which is the best method? Heart (British Cardiac Society),
93(5):615–20.
Bluestein, D., Claiborne, T. E., and Slepian, M. (2013a). Polymeric Heart Valve.
Bluestein, D., Einav, S., and Slepian, M. J. (2013b). Device thrombogenicity emulation: a
novel methodolgy for optimizing thromboresistance of cardiovascular devices. Journal
of Biomechanics, 46:338–344.
Bodnar, E. and Frater, R. (1992). Replacement Cardiac Valves. McGraw-Hill.
Boloori Zadeh, P., Corbett, S. C., and Nayeb-Hashemi, H. (2013). Effects of fluid flow
shear rate and surface roughness on the calcification of polymeric heart valve leaflet.
Materials Science & Engineering. C, Materials for Biological Applications, 33(5):2770–
5.
Borsali, R. and Pecora, R. (2008). Soft Matter Characterization. Springer Netherlands.
Botzenhardt, F., Eichinger, W. B., Bleiziffer, S., Guenzinger, R., Wagner, I. M., Bauern-
schmitt, R., and Lange, R. (2005). Hemodynamic Comparison of Bioprostheses for
Bibliography 267
Complete Supra-Annular Position in Patients With Small Aortic Annulus. Journal of
the American College of Cardiology, 45(12):2054–2060.
Bourantas, C. V., Van Mieghem, N. M., Farooq, V., Soliman, O. I., Windecker, S.,
Piazza, N., and Serruys, P. W. (2013). Future perspectives in transcatheter aortic
valve implantation. International Journal of Cardiology, 168(1):11–18.
Braune, S., Grunze, M., Straub, A., and Jung, F. (2013). Are there sufficient standards
for the in vitro hemocompatibility testing of biomaterials? Biointerphases, 8(1):30.
Brockbank, K. G. M. and Song, Y. C. (2003). Mechanisms of bioprosthetic heart valve
calcification. Transplantation, 75(8):1133–5.
Bronzino, J. D. (2000). The Biomedical Engineering Handbook. Springer, 2 edition.
Bruck, S. (1991). Biostability of materials and implants. Journal of Long-term Effects of
Medical Implants.
Brydson, J. (1995). Thermoplastic Elastomers: Properties and Applications. iSmithers
Rapra Publishing, volume 81 edition.
Buellesfeld, L., Stortecky, S., Kalesan, B., Gloekler, S., Khattab, A. a., Nietlispach,
F., Delfine, V., Huber, C., Eberle, B., Meier, B., Wenaweser, P., and Windecker, S.
(2013). Aortic root dimensions among patients with severe aortic stenosis undergoing
transcatheter aortic valve replacement. JACC: Cardiovascular Interventions, 6(1):72–
83.
Burriesci, G., Marincola, F. C., and Zervides, C. (2010). Design of a novel polymeric
heart valve. Journal of medical engineering & technology, 34(1):7–22.
Butcher, J. T., Simmons, C. A., and Warnock, J. N. (2007). Review : Mechanobiology
of the Aortic Heart Valve. Journal of Heart Valve Disease, 17:62–73.
Butterfield, M., Wheatley, D., Williams, D., and Fisher, J. (2001). A new design for
polyurethane heart valves. Journal of heart valve disease, 10(1):105–10.
Cacciola, G., Peters, G. W., and Baaijens, F. P. (2000). A synthetic fiber-reinforced
stentless heart valve. Journal of biomechanics, 33(6):653–8.
Campbell, L. A., Moe, R., and Sarnowski, E. J. (2004). Method of manufacturing a
prosthetic heart valve.
Carabello, B. a. (2002). Evaluation and management of patients with aortic stenosis.
Circulation, 105(15):1746–1750.
268 Bibliography
Carastan, D. J., Amurin, L. G., Craievich, A. F., Gonc¸alves, M. D. C., and Demarquette,
N. R. (2013). Morphological evolution of oriented clay-containing block copolymer
nanocomposites under elongational flow. European Polymer Journal, 49(6):1391–1405.
Castelletto, V. and Hamley, I. (2004). Morphologies of block copolymer melts. Current
Opinion in Solid State and Materials Science, 8(6):426–438.
Cerillo, A. G., Mariani, M., Berti, S., and Glauber, M. (2012). Sizing the aortic annulus.
Annals of cardiothoracic surgery, 1(2):245–56.
Chakraborty, S., Bag, S., Pal, S., and Mukherjee, A. K. (2006). Structural and mi-
crostructural characterization of bioapatites and synthetic hydroxyapatite using X-ray
powder diffraction and Fourier transform infrared techniques. Journal of Applied Crys-
tallography, 39(3):385–390.
Chandran, K. B., Kim, S. H., and Han, G. (1991). Stress distribution on the cusps
of a polyurethane trileaflet heart valve prosthesis in the closed position. Journal of
biomechanics, 24(6):385–95.
Charnley, J. (1963). Tissue reactions to polytetrafluoroethylene. Lancet, Decem-
ber(28):1379.
Chen, Z. Z., Kornfield, J., and Smith, S. (1997). Pathways to Macroscale Order in
Nanostructured Block Copolymers. Science, 277(5330):1248–1253.
Chenoweth, D. E. (1987). Complement Activation in Extracorporeal Circuits. Annals of
the New York Academy of Sciences, 516:306–313.
Cherepanov, G. (1967). Crack propagation in continuous media. Journal of Applied
Mathematics and Mechanics, 31(3):503–512.
Chetta, G. and Lloyd, J. (1980). The Design, Fabrication and Evaluation of a Trileaflet
Prosthetic Heart Valve. Journal of Biomechanical Engineering, 102(1):8.
Chinn, J. A. (2002). Polymeric heart valve fabricated from
polyurethane/polysiliconeurethane blends.
Cho, H. H.-j. H.-j. and Kim, H.-s. H. (2009). Osteopontin: a multifunctional protein
at the crossroads of inflammation, atherosclerosis, and vascular calcification. Current
Atherosclerosis Reports, pages 206–213.
Cho, K., Jang, W. J., Lee, D., Chun, H., and Chang, Y. W. (2000). Fatigue crack growth
of elastomers in the swollen state. Polymer, 41(1):179–183.
Bibliography 269
Choi, I. S. and Roland, C. M. (1996). Intrinsic Defects and the Failure Properties of
cis-1,4-Polyisoprenes. Rubber Chemistry and Technology, 69(4):591–99.
Chong, K. P., Wieting, D. W., Hwang, N. H. C., and Kennedy, J. H. (1973). Stress
Analysis of Normal Human Aortic Valve Leaflets during Diastole. Artificial Cells,
Blood Substitutes and Biotechnology, 1(2):307–321.
Christensen, K., Larsson, R., and Emanuelsson, H. (2001). Heparin coating of the stent
graft effects on platelets, coagulation and complement activation. Biomaterials, 22:349–
355.
Christie, G. W. and Barratt-Boyes, B. G. (1991). On stress reduction in bioprosthetic
heart valve leaflets by the use of a flexible stent. Journal of Cardiac Surgery, 6(4):476–
81.
Christie, G. W. and Barratt-Boyes, B. G. (1995). Age-dependent changes in the radial
stretch of human aortic valve leaflets determined by biaxial testing. The Annals of
Thoracic Surgery, 60:S156–9.
Chung, D. H. and Kwon, T. H. (2002). Fiber orientation in the processing of polymer
composites. Korean-Australia Rheology Journal, 14(4):175–188.
Claiborne, T., Xenos, M., Sheriff, J., Chiu, W., Soares, J., Alemu, Y., Gupta, S., Judex,
S., Slepian, M., and Bluestein, D. (2013a). Toward Optimization of a Novel Trileaflet
Polymeric Prosthetic Heart Valve via Device Thrombogenicity Emulation. ASAIO
Journal, 59(3):275–83.
Claiborne, T. E., Girdhar, G., Gallocher-Lowe, S., Sheriff, J., Kato, Y. P., Pinchuk, L.,
Schoephoerster, R. T., Jesty, J., and Bluestein, D. (2011). Thrombogenic potential of
Innovia polymer valves versus Carpentier-Edwards Perimount Magna aortic biopros-
thetic valves. ASAIO journal (American Society for Artificial Internal Organs : 1992),
57(1):26–31.
Claiborne, T. E., Sheriff, J., Kuetting, M., Steinseifer, U., Slepian, M. J., and Bluestein,
D. (2013b). In vitro evaluation of a novel hemodynamically optimized trileaflet poly-
meric prosthetic heart valve. Journal of Biomechanical Engineering, 135(2):1–8.
Clark, M. A., Arnold, S. V., Duhay, F. G., Thompson, A. K., Keyes, M. J., Svensson,
L. G., Bonow, R. O., Stockwell, B. T., and Cohen, D. J. (2012). Five-year clinical
and economic outcomes among patients with medically managed severe aortic stenosis:
Results from a medicare claims analysis. Circulation: Cardiovascular Quality and
Outcomes, 5(5):697–704.
270 Bibliography
Co, A. and Bird, R. B. (1977). Slow viscoelastic radial flow between parallel disks. Applied
Scientific Research, 33(5-6):385–404.
Cohn, L. H., Collins, J. J., DiSesa, V. J., Couper, G. S., Peigh, P. S., Kowalker, W., and
Allred, E. (1989). Fifteen-year experience with 1678 Hancock porcine bioprosthetic
heart valve replacements. Annals of surgery, 210(4):435–442; discussion 442–443.
Corden, J., David, T., and Fisher, J. (1995). In vitro determination of the curvatures
and bending strains acting on the leaflets of polyurethane trileaflet heart valves during
leaflet motion. Proceedings of the Institution of Mechanical Engineers, 209(4):243–53.
Cracknell, P. and Dyson, R. (1993). Handbook of Thermoplastics Injection Mould Design.
Springer, 1 edition.
Cribier, A., Eltchaninoff, H., Bash, A., Borenstein, N., Tron, C., Bauer, F., Derumeaux,
G., Anselme, F., Laborde, F., and Leon, M. B. (2002). Percutaneous transcatheter
implantation of an aortic valve prosthesis for calcific aortic stenosis: First human case
description. Circulation, 106(24):3006–3008.
Daebritz, S. H., Fausten, B., Hermanns, B., Franke, A., Schroeder, J., Groetzner, J.,
Autschbach, R., Messmer, B. J., and Sachweh, J. S. (2004a). New flexible polymeric
heart valve prostheses for the mitral and aortic positions. The Heart Surgery Forum,
7(5):E525–32.
Daebritz, S. H., Fausten, B., Hermanns, B., Schroeder, J., Groetzner, J., Autschbach,
R., Messmer, B. J., and Sachweh, J. S. (2004b). Introduction of a flexible polymeric
heart valve prosthesis with special design for aortic position. European Journal of Car-
diothoracic Surgery : Official Journal of the European Association for Cardiothoracic
Surgery, 25(6):946–52.
Dagum, P., Green, G. R., Nistal, F. J., Daughters, G. T., Timek, T. A., Foppiano, L. E.,
Bolger, A. F., Ingels, N. B., and Miller, D. C. (1999). Deformational Dynamics of
the Aortic Root Modes and Physiologic Determinants. Circulation, 100(Supplement
2):54–62.
D’Arcy, J., Prendergast, B., and Chambers, J. (2011). Valvular heart disease: the next
cardiac epidemic. Heart, 97(2):91–93.
Dasi, L. P., Simon, H. A., Sucosky, P., and Yoganathan, A. P. (2009). Fluid mechanics of
artificial heart valves. Clinical and experimental pharmacology & physiology, 36(2):225–
37.
De Hart, J., Peters, G., Schreurs, P., and Baaijens, F. (2004). Collagen fibers reduce
Bibliography 271
stresses and stabilize motion of aortic valve leaflets during systole. Journal of Biome-
chanics, 37(3):303–311.
De Paulis, R., Schmitz, C., Scaffa, R., Nardi, P., Chiariello, L., and Reul, H. (2005). In
vitro evaluation of aortic valve prosthesis in a novel valved conduit with pseudosinuses
of Valsalva. Journal of Thoracic and Cardiovascular Surgery, 130(4):1016–1021.
Demer, L. L. and Tintut, Y. (2008). Vascular calcification: pathobiology of a multifaceted
disease. Circulation, 117(22):2938–48.
DeYoreo, J. J. and Vekilov, P. G. (2003). Principles of Crystal Nucleation and Growth.
Reviews in Mineralogy and Geochemistry, 54(1):57–93.
Diab, M., Peter, J., Faerber, G., Bothe, W., Lemke, S., Breuer, M., and Doenst, T.
(2011). Sizing strategy is a major determinant for postoperative pressure gradients in
commonly implanted stented tissue valves. Circulation, 124:A16654.
Diani, J., Fayolle, B., and Gilormini, P. (2009). A review on the Mullins effect. European
Polymer Journal, pages 601–612.
Djabourov, M., Nishinari, K., and Ross-Murphy, S. B. (2013). Physical Gels from Bio-
logical and Synthetic Polymers. Cambridge University Press.
Doehring, T. C., Kahelin, M., and Vesely, I. (2005). Mesostructures of the Aortic Valve.
Journal of heart valve disease, 14(5):679–686.
Dorozhkin, S. V. (2009). Calcium Orthophosphates in Nature, Biology and Medicine.
Materials, 2(2):399–498.
Drolet, F. and Fredrickson, G. H. (2001). Optimizing chain bridging in complex block
copolymers. Macromolecules, 34(15):5317–5324.
Du Bois, D. and Du Bois, E. F. (1916). A Formula To Estimate The Approximate
Surface Area If Height And Weight Be Known. Archives of Internal Medicine (Chic),
XVII(6 2):863–871.
Ducrot, E., Chen, Y., Bulters, M., Sijbesma, R. P., and Creton, C. (2014). Toughening
elastomers with sacrificial bonds and watching them break. Science (New York, N.Y.),
344(6180):186–9.
Eastman (2011). Eastman polymers processing and mold design guidelines. Technical
report, Eastman Chemical Company, Kingsport, TN.
Ebewele, R. (2000). Polymer Science and Technology. CRC Press.
272 Bibliography
Edwin, F., Tettey, M., Aniteye, E., Tamatey, M., Sereboe, L., Entsua-Mensah, K., Kotei,
D., and Baffoe-Gyan, K. (2011). The development of cardiac surgery in West Africa–the
case of Ghana. The Pan African medical journal, 9:15.
Egan, K. P., Kim, J. H., Mohler, E. R., and Pignolo, R. J. (2011). Role for circulating
osteogenic precursor cells in aortic valvular disease. Arteriosclerosis, Thrombosis, and
Vascular Biology, 31(12):2965–2971.
Ehrenstein, M. R., Jury, E. C., and Mauri, C. (2005). Statins for atherosclerosis - as
good as it gets? The New England Journal of Medicine, 352(1):73–5.
El Fray, M. and Altsta¨dt, V. (2003a). Fatigue behaviour of multiblock thermoplas-
tic elastomers. 1. Stepwise increasing load testing of poly(aliphatic/aromatic-ester)
copolymers. Polymer, 44(16):4635–4642.
El Fray, M. and Altsta¨dt, V. (2003b). Fatigue behaviour of multiblock thermoplastic
elastomers. 2. Dynamic creep of poly(aliphatic/aromatic-ester) copolymers. Polymer,
44(16):4643–4650.
El Fray, M. and Altsta¨dt, V. (2004). Fatigue behaviour of multiblock thermoplastic elas-
tomers. 3. Stepwise increasing strain test of poly(aliphatic/aromatic-ester) copolymers.
Polymer, 45(1):263–273.
El Fray, M., Prowans, P., Puskas, J. E., and Altsta¨dt, V. (2006). Biocompatibility and fa-
tigue properties of polystyrene-polyisobutylene-polystyrene, an emerging thermoplastic
elastomeric biomaterial. Biomacromolecules, 7(3):844–50.
El-Hamamsy, I. and Poirier, N. C. (2013). What is the role of the Ross procedure in
today’s armamentarium? The Canadian journal of cardiology, 29(12):1569–76.
Engberg, A. E., Rosengren-Holmberg, J. P., Chen, H., Nilsson, B., Lambris, J. D.,
Nicholls, I. a., and Ekdahl, K. N. (2011). Blood protein-polymer adsorption: Im-
plications for understanding complement-mediated hemoincompatibility. Journal of
biomedical materials research. Part A, pages 74–84.
Erlenko¨tter, A., Endres, P., Nederlof, B., Hornig, C., and Vienken, J. (2008). Score
model for the evaluation of dialysis membrane hemocompatibility. Artificial organs,
32(12):962–9.
Fang, L., Wei, M., Shang, Y., Jimenez, L., Kazmer, D., Barry, C., and Mead, J. (2009).
Surface morphology alignment of block copolymers induced by injection molding. Poly-
mer, 50(24):5837–5845.
Bibliography 273
Farzaneh-Far, A., Proudfoot, D., Shanahan, C., and Weissberg, P. (2001). Vascular and
valvar calcification: recent advances. Heart, 85(1):13–7.
Fisher, L. and Termine, J. (1985). Noncollagenous proteins influencing the local mecha-
nisms of calcification. Clinical Orthopaedics and Related Research, 200(November):362.
Fratzl, P. (2008). Collagen: Structure and Mechanics. Springer Science and Business
Media, LLC.
Fratzl, P., Dunlop, J. W. C., and Weinkamer, R. (2013). Materials Design Inspired by
Nature: Function Through Inner Architecture, volume 1. Royal Society of Chemistry.
Gallegos, R., Rivard, A., Suwan, P., Black, S., Bertog, S., Steinseifer, U., Armien, A.,
Lahti, M., and Bianco, R. (2006). In-vivo experience with the Triflo trileaflet mechan-
ical heart valve. Journal of heart valve disease, 15(6):791–9.
Gallocher, S. L. (2007). Durability Assessment of Polymer Trileaflet Heart Valves. PhD
thesis, Florida International University.
Gallocher, S. L., Aguirre, A. F., Kasyanov, V., Pinchuk, L., and Schoephoerster, R. T.
(2006). A Novel Polymer for Potential Use in a Trileaflet Heart Valve. Journal of
Biomedical Materials Research. Part B Applied Biomaterials, 79(2):325–334.
Garcia, D. and Kadem, L. (2006). Orifice Area, Effective Orifice Area, or Gorlin Area?
Journal of Heart Valve Disease, 15(5):601–8.
Garcia, D., Pibarot, P., Landry, C., Allard, A., Chayer, B., Dumesnil, J. G., and Durand,
L.-G. (2004). Estimation of aortic valve effective orifice area by Doppler echocardiog-
raphy: effects of valve inflow shape and flow rate. Journal of the American Society of
Echocardiography, 17(7):756–765.
Garc´ıa Pa´ez, J. M., Herrero, E. J., San Mart´ın, A. C., Garc´ıa Sestafe, J. V., Te´llez, G.,
Milla´n, I., Salvador, J., Cordo´n, A., and Castillo-Olivares, J. L. (2000). The influence
of chemical treatment and suture on the elastic behavior of calf pericardium utilized
in the construction of cardiac bioprostheses. Journal of Materials Science: Materials
in Medicine, 11(5):273–277.
Gardner, R. (1974). An examination of the fluid mechanics and thrombus formation time
parameters in a Chandler rotating loop system. Journal of Laboratory and Clinical
Medicine, 84(4):494–508.
Gent, A. N., Lindley, P. B., and Thomas, A. G. (1965). Cut Growth and Fatigue of
Rubbers. I. The Relationship between Cut Growth and Fatigue. Rubber Chemistry
and Technology, 38(2):292–300.
274 Bibliography
Ghanbari, H., Kidane, A. G., Burriesci, G., Ramesh, B., Darbyshire, A., and Seifalian,
A. M. (2010). The anti-calcification potential of a silsesquioxane nanocomposite poly-
mer under in vitro conditions: potential material for synthetic leaflet heart valve. Acta
Biomaterialia, 6(11):4249–60.
Gharaie, S. H. and Morsi, Y. (2015). A novel design of a polymeric heart valve. Interna-
tional Journal of Artificial Organs, 38(5):259–70.
Gharib, M., Kremers, D., Koochesfahani, M. M., and Kemp, M. (2002). Leonardo’s vision
of flow visualization. Experiments in Fluids, 33(1):219–223.
Ghista, D. and Reul, H. (1983). Advances in Cardiovascular Physics, Part 1 Modelling.
In Cardiovascular Engineering, page 230. S. Karger.
Giachelli, C. M., Speer, M. Y., Li, X., Rajachar, R. M., and Yang, H. (2005). Regulation
of vascular calcification: roles of phosphate and osteopontin. Circulation Research,
96(7):717–22.
Giesekus, H. (1982). A simple constitutive equation for polymer fluids based on the
concept of deformation-dependent tensorial mobility. Journal of Non-Newtonian Fluid
Mechanics, 11(1-2):69–109.
Girdhar, G., Xenos, M., Alemu, Y., Chiu, W.-C., Lynch, B. E., Jesty, J., Einav, S.,
Slepian, M. J., and Bluestein, D. (2012). Device thrombogenicity emulation: a novel
method for optimizing mechanical circulatory support device thromboresistance. PloS
one, 7(3):e32463.
Goffin, Y., Van Hoeck, B., Jashari, R., Soots, G., and Kalmar, P. (2000). Banking
of Cryopreserved Heart Valves in Europe: Assessment of a 10-year Operation in the
European Homograft Bank (EHB). Journal of Heart Valve Disease, 9(2):207–14.
Golander, C.-G. and Larsson, R. (1986). Surface modified solid substrate and a method
for its preparation.
Goldbarg, S. H., Elmariah, S., Miller, M. a., and Fuster, V. (2007). Insights into degener-
ative aortic valve disease. Journal of the American College of Cardiology, 50(13):1205–
13.
Golomb, G. and Wagner, D. (1991). Development of a new in vitro model for studying
implantable polyurethane calcification. Biomaterials, 12(May):397–405.
Gonzalez-Alvarez, a., Arellano, M., Diat, O., Legrand, J. F., and Piau, J. M. (2004).
Elongational properties and crystallization of poly (isobutylene) melts probed by syn-
chrotron radiation. Revista Mexicana de Fisica, 50(5):506–514.
Bibliography 275
Goosen, M. F. A. and Sefton, M. V. (1979). Heparinized Styrene-Butadiene-Styrene
Elastomers. Journal of Biomedical Materials Research, 13:347–364.
Granta Design Ltd. (2003). Materials Database.
Griffith, A. (1921). The Phenomena of Rupture and Flow in Solids. Philosophical
Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences,
221(582-593):163–198.
Gupta, M. (2001). Simulation of planar entrance flow using strain-rate-dependent shear
and elongational viscosities. Journal of reinforced plastics and composites.
Haj-Ali, R., Dasi, L. P., Kim, H. S., Choi, J., Leo, H. W., and Yoganathan, A. P.
(2008). Structural simulations of prosthetic tri-leaflet aortic heart valves. Journal of
Biomechanics, 41(7):1510–1519.
Hamid, M., Sabbah, H., and Stein, P. (1986). Influence of stent height upon stresses on
the cusps of closed bioprosthetic valves. Journal of biomechanics, 19(9):759–769.
Hamley, I. W. (2000). Effect of shear on ordered block copolymer solutions. Current
opinion in Colloid & Interface Science, 5:342 – 350.
Hamley, I. W. (2004). Developments in block copolymer science and technology. John
Wiley & Sons, Inc.
Hamon, R. F., Khan, a. S., and Chow, a. (1982). The cation-chelation mechanism of
metal-ion sorption by polyurethanes. Talanta, 29(4):313–26.
Han, D. K., Park, K., Park, K. D., Ahn, K. D., and Kim, Y. H. (2006). In vivo bio-
compatibility of sulfonated PEO-grafted polyurethanes for polymer heart valve and
vascular graft. Artificial Organs, 30(12):955–959.
Harada, N. S., Oyama, H. T., Bartoli, J. R., Gouvea, D., Cestari, I. a., and Hui Wang,
S. (2005). Quantifying adsorption of heparin on a PVC substrate using ATR-FTIR.
Polymer International, 54(1):209–214.
Harasaki, H., Moritz, A., Uchida, N., Chen, J.-F., McMahon, J. T., Richards, T. M.,
Smith, W. A., Murabayashi, S., Kambic, H. E., Kiraly, R. J., and Nose´, Y. (1987). Ini-
tiation and Growth of Calcification in blood pumps. ASAIO Transactions, 33(3):643–
648.
Hastings, G. (1992). Cardiovascular Biomaterials. Springer-Verlag London Limited,
London, UK.
276 Bibliography
Hathcock, J. J. (2006). Flow effects on coagulation and thrombosis. Arteriosclerosis,
thrombosis, and vascular biology, 26(8):1729–37.
Haworth, W. W. S. (1978). Testing of Materials for Artificial Heart Valves. British
Polymer Journal, 10(4):297–301.
Hayes, M., Edwards, D., and Shah, A. (2015). Fractography in Failure Analysis of Poly-
mers. William Andrew.
Heim, F., Gasser, B., Khoffi, F., and Blondel, P. (2014). Textile heart valve: first in-vivo
experiment in the aortic position. Journal of heart valve disease, 23(3):316–8.
Helske, S. and Kupari, M. (2007). Aortic valve stenosis: an active atheroinflammatory
process. Current Opinion in Lipidology, 18(5):483–491.
Hermans, H., Herijgers, P., Holvoet, P., Verbeken, E., Meuris, B., Flameng, W., and
Herregods, M.-C. (2010). Statins for calcific aortic valve stenosis: into oblivion after
SALTIRE and SEAS? An extensive review from bench to bedside. Current Problems
in Cardiology, 35(6):284–306.
Heron, M., Hoyert, D. L., Murphy, S. L., Xu, J., Kochanek, K. D., and Tejada-Vera, B.
(2009). Deaths: final data for 2006. National vital statistics reports : from the Centers
for Disease Control and Prevention, National Center for Health Statistics, National
Vital Statistics System, 57(14):1–134.
Hilbert, S., Ferrans, V., Tomita, Y., Eidbo, E., and Jones, M. (1987). Evaluation of
explanted polyurethane trileaflet cardiac valve prostheses. Journal of Thoracic and
Cardiovascular Surgery, 94(3):419–29.
Ho, S. Y. (2009). Structure and anatomy of the aortic root. European journal of echocar-
diography : the journal of the Working Group on Echocardiography of the European
Society of Cardiology, 10(1):i3–10.
Hoerstrup, S. P., Sodian, R., Sperling, J. S., Vacanti, J. P., and Jr, J. E. M. (2000a).
Engineered Heart Valves. Tissue Engineering, 6(1):75–85.
Hoerstrup, S. S. P., Sodian, R., Daebritz, S., Wang, J., Bacha, E. A., Martin, D. P.,
Moran, A. M., Guleserian, K. J., Sperling, J. S., Kaushal, S., Vacanti, J. P., Schoen,
F. J., and Mayer, J. E. (2000b). Functional living trileaflet heart valves grown in vitro.
Circulation, 102:Iii–44–Iii–49.
Hoffman, D., Sisto, D., Yu, L., Dahm, M., and Kolff, W. (1991). Evaluation of a Stented
Polyurethane Mitral Valve. ASAIO Transactions, 37(3):M354–5.
Bibliography 277
Holzapfel, G. a., Gasser, T. C., and Ogden, R. W. (2000). A new constitutive framework
for arterial wall mechanics and a comperative study of material models. J. Elasticity,
61(8):1–48.
Honeker, C. C. and Thomas, E. L. (2000). Perpendicular deformation of a near-single-
crystal triblock copolymer with a cylindrical morphology. 2. TEM. Macromolecules,
33(25):9407–9417.
Hong, J., Larsson, A., Ekdahl, K. N., Elgue, G., Larsson, R., and Nilsson, B. (2001).
Contact between a polymer and whole blood: Sequence of events leading to thrombin
generation. Journal of Laboratory and Clinical Medicine, 138(2):139–145.
Hopkins, R. a. (2006). Bioprosthetic valves and laudable inflammation? Circulation,
114(4):261–264.
Hukins, D. W. L., Mahomed, A., and Kukureka, S. N. (2008). Accelerated aging for
testing polymeric biomaterials and medical devices. Medical engineering & physics,
30(10):1270–4.
Hulme, A. and Cooper, J. (2012). Life prediction of polymers for industry. Sealing
technology, (September):8–12.
Imachi, K., Abe, Y., Matsuura, H., Iwasaki, K., Son, Y.-p., and Kouno, A. (2001). A
new hypothesis on the mechanism of calcification formed on a blood-contacted polymer
surface. Journal of Artificial Organs, 4(1):74–82.
Imachi, K., Mabuchi, K., Chinzei, T., Abe, Y., Imanishi, K., Yonezawa, T., Maeda, K.,
Suzukawa, M., Kouno, A., Oho, T., Fujisama, I., and Alsumi, K. (1989). In Vitro and
In Vivo Evaluation of a Jellyfish Valve for Practical Use. ASAIO Transactions, 35(3).
International Standards Organisation (2005). Cardiovascular implants - Cardiac valve
prostheses.
ISO (2010). Biological evaluation of medical devices - Part 13: Identification and quan-
tification of degradation products from polymeric medical devices (ISO 10993-13:2010).
Iung, B. (2003). A prospective survey of patients with valvular heart disease in Eu-
rope: The Euro Heart Survey on Valvular Heart Disease. European Heart Journal,
24(13):1231–1243.
Iung, B., Cachier, A., Baron, G., Messika-Zeitoun, D., Delahaye, F., Tornos, P., Gohlke-
Ba¨rwolf, C., Boersma, E., Ravaud, P., and Vahanian, A. (2005). Decision-making in
elderly patients with severe aortic stenosis: Why are so many denied surgery? European
Heart Journal, 26(24):2714–2720.
278 Bibliography
Iung, B. and Vahanian, A. (2011). Epidemiology of valvular heart disease in the adult.
Nature reviews. Cardiology, 8(3):162–172.
Jansen, J., Willeke, S., Reiners, B., Harbott, P., Reul, H., Lo, H. B., Daebritz, S. H.,
Rosenbaum, C., Bitter, A., Ziehe, K., Rau, G., and Messmer, B. J. (1991). Advances
in design principle and fluid dynamics of flexible polymeric heart valve. ASAIO Trans-
actions, 37(3):M451–3.
Jesty, J. and Bluestein, D. (1999). Acetylated prothrombin as a substrate in the measure-
ment of the procoagulant activity of platelets: elimination of the feedback activation
of platelets by thrombin. Analytical biochemistry, 272(1):64–70.
Jesty, J., Yin, W., Perrotta, P., and Bluestein, D. (2003). Platelet activation in a cir-
culating flow loop: combined effects of shear stress and exposure time. Platelets,
14(3):143–149.
Jiang, H., Campbell, G., Boughner, D., Wan, W. K., and Quantz, M. (2004). Design
and manufacture of a polyvinyl alcohol (PVA) cryogel tri-leaflet heart valve prosthesis.
Medical Engineering and Physics, 26(4):269–277.
Jones, J. M., O’Kane, H., Gladstone, D. J., Sarsam, M. a., Campalani, G., MacGowan,
S. W., Cleland, J., and Cran, G. W. (2001). Repeat heart valve surgery: risk factors for
operative mortality. The Journal of thoracic and cardiovascular surgery, 122(5):913–
918.
Jono, S., McKee, M. D., Murry, C. E., Shioi, a., Nishizawa, Y., Mori, K., Morii, H.,
and Giachelli, C. M. (2000). Phosphate regulation of vascular smooth muscle cell
calcification. Circulation research, 87(7):E10–E17.
Kalath, S., Tsipouras, P., and Silver, F. H. (1986). Non-invasive assessment of aortic
mechanical properties. Annals of biomedical engineering, 14(6):513–24.
Kapolos, J., Mavrilas, D., Missirlis, Y., and Koutsoukos, P. G. (1997). Model experimen-
tal system for investigation of heart valve calcification in vitro. Journal of biomedical
materials research, 38(3):183–90.
Karatasos, K., Anastasiadis, S. H., Pakula, T., and Watanabe, H. (2000). On the loops-to-
bridges ratio in ordered triblock copolymers: An investigation by dielectric relaxation
spectroscopy and computer simulations. Macromolecules, 33(2):523–541.
Keller, A., Pedemonte, E., and Willmouth, F. M. (1970). Macro lattice from segregated
amorphous phases of a three block copolymer. Nature, 225(5232):538.
Kelly, B. S., Heffelfinger, S. C., Whiting, J. F., Miller, M. A., Reaves, A., Armstrong, J.,
Bibliography 279
Narayana, A., and Roy-Chaudhury, P. (2002). Aggressive venous neointimal hyperpla-
sia in a pig model of arteriovenous graft stenosis. Kidney International, 62(6):2272–
2280.
Kester, R. C. (1984). The thrombogenicity of Dacron arterial grafts and its modification
by platelet inhibitory drugs. Annals of the Royal College of Surgeons of England,
66(4):241–246.
Kheradvar, A., Groves, E. M., Dasi, L. P., Alavi, H. S., Tranquillo, R. T., Grande-Allen,
K. J., Simmons, C. A., Griffith, B., Falahatpisheh, A., Goergen, C. J., Mofrad, M.
R. K., Baaijens, F. P., Little, S. H., and Canic, S. (2015). Emerging Trends in Heart
Valve Engineering: Part I. Solutions for Future. Annals of biomedical engineering,
43(4):833–43.
Kidane, A. G., Salacinski, H., Tiwari, A., Bruckdorfer, K. R., and Seifalian, A. M. (2004).
Anticoagulant and antiplatelet agents: their clinical and device application(s) together
with usages to engineer surfaces. Biomacromolecules, 5(3):798–813.
Kim, I. Y., Kim, J. W., Jung, D. Y., Kim, C. S., and Min, B. G. (1991). Development
of Polymer Prosthetic Heart Valve - Fabrication and In Vitro Test. The Seoul Journal
of Medicine, 32(1):35–42.
Kingston, J. and Muhr, A. (2012). Determination of effective flaw size for fatigue life
prediction. In Constitutive Models for Rubber, volume VII, pages 337–342. CRC Press.
Kini, V., Bachmann, C., Fontaine, A., Deutsch, S., and Tarbell, J. M. (2000). Flow
visualization in mechanical heart valves: occluder rebound and cavitation potential.
Annals of biomedical engineering, 28(4):431–41.
Kiraly, R. J. and Hillegass, D. V. (1980). Polyolefin blood pump components. Synthetic
biomedical polymers concepts and applications, pages 59–71.
Kobayashi, J. (2011). Stentless aortic valve replacement: an update. Vascular health and
risk management, 7:345–351.
Kokubo, T. (1990). Surface chemistry of bioactive glass-ceramics. Journal of Non-
Crystalline Solids, 120:138–151.
Kolh, P., Kerzmann, A., Honore, C., Comte, L., and Limet, R. (2007). Aortic valve
surgery in octogenarians: predictive factors for operative and long-term results. Euro-
pean Journal of Cardio-thoracic Surgery, 31(4):600–606.
Krings, M., Kanellopoulou, D., Mavrilas, D., and Glasmacher, B. (2006). In vitro pH-
280 Bibliography
controlled calcification of biological heart valve prostheses. Materialwissenschaft und
Werkstofftechnik, 37(6):432–435.
Kristensen, E. M. E., Larsson, R., Sa´nchez, J., Rensmo, H., Gelius, U., and Siegbahn,
H. (2006). Heparin coating durability on artificial heart valves studied by XPS and
antithrombin binding capacity. Colloids and surfaces. B: Biointerfaces, 49(1):1–7.
Krucinski, S., Vesely, I., Dokainish, M. A., and Campbell, G. (1993). Numerical simu-
lation of leaflet flexure in bioprosthetic valves mounted on rigid and expansile stents.
Journal of biomechanics, 26(8):929–43.
Ku¨tting, M., Roggenkamp, J., Urban, U., Schmitz-Rode, T., and Steinseifer, U. (2011).
Polyurethane heart valves: past, present and future. Expert Reviews of Medical Devices,
8(2):227–233.
Kyrle, P. A., Eichler, H. G., Ja¨ger, U., and Lechner, K. (1987). Inhibition of prostacyclin
and thromboxane A2 generation by low-dose aspirin at the site of plug formation in
man in vivo. Circulation, 75(5):1025–1029.
Labrosse, M. R., Beller, C. J., Robicsek, F., and Thubrikar, M. J. (2006). Geometric
modeling of functional trileaflet aortic valves: Development and clinical applications.
Journal of Biomechanics, 39(14):2665–2672.
Lach, R., Schneider, K., Weidisch, R., Janke, a., and Knoll, K. (2005). Application of
the essential work of fracture concept to nanostructured polymer materials. European
Polymer Journal, 41(2):383–392.
Lake, G. (1983). Aspects of fatigue and fracture of rubber. Progress of Rubber Technology,
45:89.
Lake, G. J. and Thomas, A. G. (1967). The Strength of Highly Elastic Materials. Pro-
ceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences,
300(1460):108–119.
Lanzarone, E., Vismara, R., and Fiore, G. B. (2009). A new pulsatile volumetric device
with biomorphic valves for the in vitro study of the cardiovascular system. Artificial
Organs, 33(12):1048–1062.
Larson, R. G. (1988). Constitutive equations for polymer melts and solutions. Butterworth
Publishers, Stoneham, MA, USA.
Le Gorju Jago, K. (2008). Fatigue life of rubber components: 3D damage evolution from
X-ray computed microtomography. In Constitutive Models for Rubber V, page 173.
Taylor & Francis/Balkema, London.
Bibliography 281
Leat, M. E. and Fisher, J. (1994). A synthetic leaflet heart valve with improved opening
characteristics. Medical Engineering and Physics, 16(6):470–476.
Leat, M. E. and Fisher, J. (1995). The influence of manufacturing methods on the function
and performance of synthetic leaflet heart valve. In Proceedings of the Institution of
Mechanical Engineers, Part H: Journal of Engineering in Medicine, pages 65–70.
Leefe, S. E. and Gentle, C. R. (1987). Theoretical evaluation of energy loss methods in
the analysis of prosthetic heart valves. Journal of biomedical engineering, 9(2):121–7.
LeFan, J. (2011). Liquid Silicone Rubber Injection Molding.
Leiblert, L. and Leibler, L. (1980). Theory of Microphase Separation in Block Copolymers.
Macromolecules, 13(6):1602–1617.
Leo, H. L. (2005). An In Vitro Investigation of the Flow Fields Through Bileaflet and
Polymeric Prosthetic Heart Valves. Phd, Georgi Institute of Technology.
Leo, H. L., Dasi, L. P., Carberry, J., Simon, H. a., and Yoganathan, A. P. (2006). Fluid
dynamic assessment of three polymeric heart valves using particle image velocimetry.
Annals of biomedical engineering, 34(6):936–52.
Leo, H.-L., Simon, H., Carberry, J., Lee, S.-C., and Yoganathan, A. P. (2005). A Com-
parison of Flow Field Structures of Two Tri-Leaflet Polymeric Heart Valves. Annals of
Biomedical Engineering, 33(4):429–443.
Leopold, J. a. (2012). Cellular mechanisms of aortic valve calcification. Circulation:
Cardiovascular Interventions, 5(4):605–614.
Levy, R. J., Schoen, F. J., and Howard, S. L. (1983). Mechanism of calcification of
porcine bioprosthetic aortic valve cusps: role of T-lymphocytes. The American Journal
of Cardiology, 52(5):629–31.
Li, J., Illeperuma, W. R. K., Suo, Z., and Vlassak, J. J. (2014). Hybrid hydrogels with
extremely high stiffness and toughness. ACS Macro Letters, 3(6):520–523.
Liedel, C., Pester, C. W., Ruppel, M., Urban, V. S., and Bo¨ker, A. (2012). Beyond
Orientation: The Impact of Electric Fields on Block Copolymers. Macromolecular
Chemistry and Physics, 213(3):259–269.
Lindroos, M., Kupari, M., Heikkila¨, J., and Tilvis, R. (1993). Prevalence of aortic valve
abnormalities in the elderly: an echocardiographic study of a random population sam-
ple. Journal of the American College of Cardiology, 21(5):1220–1225.
282 Bibliography
Liu, A. C., Joag, V. R., and Gotlieb, A. I. (2007a). The emerging role of valve intersti-
tial cell phenotypes in regulating heart valve pathobiology. The American journal of
pathology, 171(5):1407–18.
Liu, Y., Kasyanov, V., and Schoephoerster, R. T. (2007b). Effect of fiber orientation on
the stress distribution within a leaflet of a polymer composite heart valve in the closed
position. Journal of biomechanics, 40(5):1099–106.
Liu, Z.-X., Wang, X.-M., Wang, Q., Shen, X.-C., Liang, H., and Cui, F.-Z. (2012).
Evolution of calcium phosphate crystallization on three functional group surfaces with
the same surface density. CrystEngComm, 14(20):6695.
Loerakker, S., Argento, G., Oomens, C. W. J., and Baaijens, F. P. T. (2013). Effects
of valve geometry and tissue anisotropy on the radial stretch and coaptation area of
tissue-engineered heart valves. Journal of Biomechanics, 46(11):1792–1800.
Long, M. and Urry, D. (1981). On the molecular mechanism of elastic fiber calcification.
ASAIO Journal, 27:690–696.
Lu, G. and Xu, T. X. (2003). Inversion of a spherical shell. In Energy Absorption of
Structures and Materials, chapter Cylindrica, pages 258–261. Elsevier.
Luo, X. Y., Li, W. G., and Li, J. (2003). Geometrical stress-reducing factors in the
anisotropic porcine heart valves. Journal of biomechanical engineering, 125(5):735–
744.
Ma, B., Wang, X., Wu, C., and Chang, J. (2014). Crosslinking strategies for prepara-
tion of extracellular matrix-derived cardiovascular scaffolds. Regenerative Biomaterials,
1(1):81–89.
Mackay, T. G., Wheatley, D. J., Bernacca, G. M., Fisher, a. C., and Hindle, C. S.
(1996). New polyurethane heart valve prosthesis: design, manufacture and evaluation.
Biomaterials, 17(19):1857–63.
Mackay, T. G. T. (1992). Towards a Tri-Leaflet Polyurethane Heart Valve Prosthesis.
Phd, University of Strathclyde.
Macosko, C. W. (1994). Rheology Principles, Measurements and Applications. Wiley-
VCH, USA.
Mai, Y. and Eisenberg, A. (2012). Self-assembly of block copolymers. Chemical Society
Reviews, 41:5969–5985.
Bibliography 283
Malhotra, A. (2012). The Changing Burden of Valvular Heart Disease. British Cardio-
vascular Society, June.
Manji, R. a., Menkis, A. H., Ekser, B., and Cooper, D. K. C. (2012). Porcine bioprosthetic
heart valves: The next generation. American heart journal, 164(2):177–85.
Mark, J. E. (2007). Physical properties of polymers handbook. Springer Science+Business
Media, LLC, 2007 edition.
Marom, G., Peleg, M., Halevi, R., Rosenfeld, M., Raanani, E., Hamdan, A., and Haj-Ali,
R. (2013). Fluid-structure interaction model of aortic valve with porcine-specific colla-
gen fiber alignment in the cusps. Journal of biomechanical engineering, 135(10):101001–
6.
Mars, W. and Fatemi, A. (2002). A literature survery of fatigue analysis approaches for
rubber. International Journal of Fatigue, 24:949–961.
Mars, W. V. and Fatemi, A. (2003). Fatigue crack nucleation and growth in filled natural
rubber. Fatigue & Fracture of Engineering Materials & Structures, 26(9):779–789.
Martin, C. and Sun, W. (2015). Comparison of transcatheter aortic valve and surgical
bioprosthetic valve durability: a fatigue simulation study. Journal of Biomechanics,
page BMD1500350.
Masoumi, N., Annabi, N., Assmann, A., Larson, B. L., Hjortnaes, J., Alemdar, N.,
Kharaziha, M., Manning, K. B., Mayer, J. E., and Khademhosseini, A. (2014).
Tri-layered elastomeric scaffolds for engineering heart valve leaflets. Biomaterials,
35(27):7774–85.
Mathison, M., Furuse, a., and Asano, K. (1988). Doppler analysis of flow velocity profile
at the aortic root. Journal of the American College of Cardiology, 12(4):947–954.
Mavridis, H. (1988). The Effect of Fountain Flow on Molecular Orientation in Injection
Molding. Journal of Rheology, 32(6):639.
Mavrilas, D. (2004). Screening biomaterials with a new in vitro method for potential
calcification : Porcine aortic valves and bovine pericardium. Journal of Materials
Science: Materials in Medicine, 15:699–704.
Mavrilas, D., Apostolaki, A., Kapolos, J., Koutsoukos, P. G., Melachrinou, M., Zolota,
V., and Dougenis, D. (1999). Development of bioprosthetic heart valve calcification in
vitro and in animal models : morphology and composition. Journal of crystal growth,
205:554–562.
284 Bibliography
Maxwell, A., Broughton, W., Dean, G., and Sims, G. (2005). Review of accelerated ageing
methods and lifetime prediction techniques for polymeric materials. Technical report,
Engineering and Process Control Divison, National Physical Laboratory, Middlesex,
UK.
MDDI Admin (1999). Silicone rubber for medical device applications.
Medicore (2012). ECG Tutorial 1: Anatomy & Electrical System.
Mendelson, K. and Schoen, F. J. (2006). Heart valve tissue engineering: Concepts,
approaches, progress, and challenges. Annals of Biomedical Engineering, 34(12):1799–
1819.
Mercer, J., Benedicty, M., and Bahnson, H. (1973). The geometry and construction of
the aortic leaflet. Journal of Thoracic and Cardiovascular Surgery, 65(4):511–8.
Mohammadi, H., Boughner, D., Millon, L. E., and Wan, W. K. (2009). Design and
simulation of a poly(vinyl alcohol)–bacterial cellulose nanocomposite mechanical aortic
heart valve prosthesis. Proceedings of the Institution of Mechanical Engineers, Part H:
Journal of Engineering in Medicine, 223(6):697–711.
Mohammadi, H. and Mequanint, K. (2011). Prosthetic aortic heart valves: modeling and
design. Medical engineering and physics, 33(2):131–47.
Mohri, H., Hessel, E., Nelson, R., Anderson, H., Dillard, D., and Merendin, K. (1973).
Design and durability test of silastic trileaflet aortic valve prostheses. Journal of Tho-
racic and Cardiovascular Surgery, 65(4):576–582.
Morozov, A. N. and Fraaije, J. G. E. M. (2002). Orientations of the lamellar phase of
block copolymer melts under oscillatory shear flow. Physical Review E - Statistical,
Nonlinear, and Soft Matter Physics, 65(3):1–3.
Morrison, F. A. (2001). Understanding Rheology. Oxford University Press, Michigan Tech
University, 2001 edition.
Mullen, M., Yap, J., Tzamtzis, S., Seifalian, A., and Burriesci, G. (2013). The TRISKELE
Transcatheter Heart Valve. In PCR London Valves.
Napier, R. (2012). Hematology. World Technologies.
National Heart Lung and Blood Institute (2011). Heart valve disease. Technical report,
US Department of Health and Human Services.
Nawrat, Z. (2009). Handbook of Polymer Applications in Medicine and Medical Devices.
Elsevier Inc.
Bibliography 285
Neyt, M. and Van Brabandt, H. (2013). TAVI: reveal the truth, the whole truth, and
nothing but the truth. Journal of medical economics, 16(5):581–5.
Nistal, F., Garc´ıa-Mart´ınez, V., Arbe, E., Ferna´ndez, D., Artin˜ano, E., Mazorra, F., and
Gallo, I. (1990). In vivo experimental assessment of polytetrafluoroethylene trileaflet
heart valve prosthesis. Journal of Thoracic and Cardiovascular Surgery, 99(6):1074–81.
O’Brien, K. D. (2006). Pathogenesis of calcific aortic valve disease: a disease process
comes of age (and a good deal more). Arteriosclerosis, Thrombosis, and Vascular
Biology, 26(8):1721–8.
Otto, C. M. (2012). Practice of Clinical Echocardiography. Elsevier Health Sciences, 4
edition.
Papadopoulos, I. (2005). Predicting the fatigue life of elastomer components. PhD thesis,
Queen Mary, University of London.
Parfeev, V. M., Grushetskii, I. V., and Smurova, E. V. (1983). Mechanical properties
of elastomers for artificial leaflet heart valves. Mechanics of Composite Materials,
19(1):92–99.
Park, J.-C., Hwang, Y.-S., Han, D.-W., and Suh, H. (2000). A novel in vitro assessment of
tissue valve calcification by a continuous flow type method. Artificial organs, 24(2):158–
60.
Park, J. M. and Park, S. J. (2011). Modeling and Simulation of Fiber Orientation in
Injection Molding of Polymer Composites. Mathematical Problems in Engineering,
2011:1–14.
Park, K. D., Lee, W. K., Yun, J. Y., Han, D. K., Kim, S. H., Kim, Y. H., Kim, H. M., and
Kim, K. T. (1997). Novel anti-calcification treatment of biological tissues by grafting
of sulphonated poly(ethylene oxide). Biomaterials, 18(1):47–51.
Paul, A., Straub, A., Weber, N., Ziemer, G., and Wendel, H. P. (2009). CD41 Western
blotting: a new method to detect platelet adhesion to artificial surfaces used in extra-
corporeal circulation procedures. Journal of materials science. Materials in medicine,
20(1):373–8.
Peacock, J. A. (1990). An in vitro study of the onset of turbulence in the sinus of Valsalva.
Circulation research, 67(2):448–460.
PECA Labs (2015). PECA Labs.
Peskin, C. S. and McQueen, D. M. (1994). Mechanical equilibrium determines the fractal
286 Bibliography
fiber architecture of aortic heart valve leaflets. The American journal of physiology,
266(1 Pt 2):H319–H328.
Petrie, C. J. (2006). Extensional viscosity: A critical discussion. Journal of Non-
Newtonian Fluid Mechanics, 137(1-3):15–23.
Pettenazzo, E., Deiwick, M., Thiene, G., Molin, G., Glasmacher, B., Martignago, F.,
Bottio, T., Reul, H., and Valente, M. (2001). Dynamic in vitro calcification of biopros-
thetic porcine valves: evidence of apatite crystallization. The Journal of thoracic and
cardiovascular surgery, 121(3):500–9.
Pezzella, T. A. (2005). Global Expansion of Cardiothoracic Surgery - The African Chal-
lenge. African Annals of Thoracic and Cardiovascular Surgery, 1:9–11.
Phillip, W. a., O’neill, B., Rodwogin, M., Hillmyer, M. a., and Cussler, E. L. (2010).
Self-assembled block copolymer thin films as water filtration membranes. ACS Applied
Materials and Interfaces, 2(3):847–853.
Pibarot, P. and Dumesnil, J. G. (2006). Prosthesis-patient mismatch: definition, clinical
impact, and prevention. Heart (British Cardiac Society), 92(8):1022–9.
Pibarot, P. and Dumesnil, J. G. (2009). Prosthetic heart valves: selection of the optimal
prosthesis and long-term management. Circulation, 119(7):1034–48.
Pietsch, H., Kartheus, H., Holtzmann, H.-J., Sachau, G., and Reul, H. (1988). Heart
valve prosthesis and process for its production.
Pinchuk, L. (1994). A review of the biostability and carcinogenicity of polyurethanes in
medicine and the new generation of ‘biostable’ polyurethanes. Journal of Biomaterials
Science, Polymer Edition, 6(3):225–267.
Pinchuk, L., Wilson, G. J., Barry, J. J., Schoephoerster, R. T., Parel, J.-M., and Kennedy,
J. P. (2008). Medical applications of poly(styrene-block-isobutylene-block-styrene)
(“SIBS”). Biomaterials, 29(4):448–60.
Prawel, D. a., Dean, H., Forleo, M., Lewis, N., Gangwish, J., Popat, K. C., Dasi, L. P.,
and James, S. P. (2014). Hemocompatibility and Hemodynamics of Novel Hyaluronan-
Polyethylene Materials for Flexible Heart Valve Leaflets. Cardiovascular engineering
and technology, 5(1):70–81.
Proto Labs Ltd. (2010). Desiging for mouldability. Proto Labs, 2 edition.
Puskas, J. E., Dos Santos, L. M., Fischer, F., Go¨tz, C., El Fray, M., Altsta¨dt, V., and
Tomkins, M. (2009a). Fatigue testing of implantable specimens: Effect of sample size
Bibliography 287
and branching on the dynamic fatigue properties of polyisobutylene-based biomaterials.
Polymer, 50(2):591–597.
Puskas, J. E., El Fray, M., Tomkins, M., Dos Santos, L. M., Fischer, F., and Altsta¨dt, V.
(2009b). Dynamic stress relaxation of thermoplastic elastomeric biomaterials. Polymer,
50(1):245–249.
Radir, A. and Thomas, A. (1981). Tear behavior of rubbers over a wide range of rates.
Rubber Chemistry and Technology, 54(1):15–23.
Rahman, M. (2012). Degradation of Polyesters in Medical Applications. In Saleh, H.,
editor, Polyester, chapter 5. InTech.
Rahmani, B., Tzamtzis, S., Ghanbari, H., Burriesci, G., and Seifalian, A. M. (2012).
Manufacturing and hydrodynamic assessment of a novel aortic valve made of a new
nanocomposite polymer. Journal of biomechanics, 45(7):1205–11.
Rajamannan, N. M., Evans, F. J., Aikawa, E., Grande-Allen, K. J., Demer, L. L., Heis-
tad, D. D., Simmons, C. a., Masters, K. S., Mathieu, P., O’Brien, K. D., Schoen, F. J.,
Towler, D. a., Yoganathan, A. P., and Otto, C. M. (2011). Calcific aortic valve disease:
not simply a degenerative process: A review and agenda for research from the Na-
tional Heart and Lung and Blood Institute Aortic Stenosis Working Group. Executive
summary: Calcific aortic valve disease - 2011 update. Circulation, 124(16):1783–91.
Ramstack, J. M., Zuckerman, L., and Mockros, L. F. (1979). Shear-induced activation of
platelets. Journal of biomechanics, 12(2):113–125.
Reade, L. (2009). Nanocages for flexible heart valves. Materials World Magazine.
Remi, E., Khelil, N., Di Centa, I., Roques, C., Ba, M., Medjahed-Hamidi, F., Chaubet,
F., Letourneur, D., Lansac, E., and Meddahi-Pelle, A. (2011). Pericardial Processing:
Challenges, Outcomes and Future Prospects, Biomaterials Science and Engineering. In
Pignatello, R., editor, Biomaterials science and engineering, chapter 22, pages 437–456.
InTech.
Rice, J. R. (1968). A Path Independent Integral and the Approximate Analysis of Strain
Concentration by Notches and Cracks. Journal of Applied Mechanics, 35(2):379.
Rivlin, R. and Thomas, A. (1953). Rupture of rubber. I. Characteristic energy for tearing.
Journal of Polymer Science, 10(3):291–318.
Rizzoli, G., Bottio, T., Thiene, G., Toscano, G., and Casarotto, D. (2003). Long-term
durability of the Hancock II porcine bioprosthesis. Journal of Thoracic and Cardio-
vascular Surgery, 126(1):66–74.
288 Bibliography
Robicsek, F. (1991). Leonardo da Vinci and the sinuses of Valsalva. The Annals of
Thoracic Surgery, 52(2):328–335.
Roche, N. and Perier, L. (2013). Influence of Elastomers Formulation on Fatigue Crack
Growth Properties. Procedia Engineering, 66:705–712.
Rock, C. A., Han, L., and Doehring, T. C. (2014). Complex collagen fiber and membrane
morphologies of the whole porcine aortic valve. PloS one, 9(1):e86087.
Rode´s-Cabau, J. (2012). Transcatheter aortic valve implantation: current and future
approaches. Nature reviews. Cardiology, 9(1):15–29.
Roe, B. B. (1992). ‘Extinct’ Cardiac Valve Prosthesis. In Bodnar, E. and Frater, R.,
editors, Replacement Cardiac Valves, chapter 13, pages 307 – 332. McGraw-Hill, 1992
edition.
Roe, B. B., Kelly, P. B. J., Myers, J. L., and Moore, D. W. (1966). Tricuspid Leaflet
Aortic Valve Prosthesis. Circulation, 33(451):124–130.
Rosamond, W., Flegal, K., Furie, K., Go, A., Greenlund, K., Haase, N., Hailpern, S. M.,
Ho, M., Howard, V., Kissela, B., Kissela, B., Kittner, S., Lloyd-Jones, D., McDermott,
M., Meigs, J., Moy, C., Nichol, G., O’Donnell, C., Roger, V., Sorlie, P., Steinberger,
J., Thom, T., Wilson, M., and Hong, Y. (2008). Heart disease and stroke statistics–
2008 update: a report from the American Heart Association Statistics Committee and
Stroke Statistics Subcommittee. Circulation, 117(4):e25–146.
Rottenberg, D., Sondak, E., and Haimovich, D. (2000). Method for producing heart
valves.
Ruzette, A.-V. and Leibler, L. (2005). Block copolymers in tomorrow ’ s plastics. Nature
materials, 4:19–31.
Ruzicka, D. J., Hettich, I., Hutter, A., Bleiziffer, S., Badiu, C. C., Bauernschmitt, R.,
Lange, R., and Eichinger, W. B. (2009). The complete supraannular concept: in
vivo hemodynamics of bovine and porcine aortic bioprostheses. Circulation, 120(11
Suppl):S139–45.
Sacks, M., Smith, D., and Hiester, E. (1997). A small angle light scattering device for
planar connective tissue microstructural analysis. Annals of biomedical engineering,
25(4):678–89.
Sacristan, E., Corona, F., Suarez, B., Rodriguez, G., Duenas, B., Gorzelewski, a., and
Calderon, M. (2003). Development of a universal second generation pneumatic ven-
tricular assist device and drive unit. Proceedings of the 25th Annual International
Bibliography 289
Conference of the IEEE Engineering in Medicine and Biology Society (IEEE Cat.
No.03CH37439), 1:427–430.
Sakurai, S. (2008). Progress in control of microdomain orientation in block copolymers -
Efficiencies of various external fields. Polymer, 49:2781–2796.
Sauren, A. A. H. J. (1981). Mechanical behaviour of the aortic valve. Phd, Eindhoven
University of Technology.
Schoen, F. J. and Levy, R. J. (1992). Bioprosthetic heart valve calcification: membrane-
mediated events and alkaline phosphatase. Bone and Mineral, 17:129–133.
Schoen, F. J. and Levy, R. J. (2005). Calcification of tissue heart valve substitutes:
progress toward understanding and prevention. The Annals of Thoracic Surgery,
79(3):1072–80.
Schoen, F. J. and Levy, R. J. (2009). SnapShot : Calcification of Bioprosthetic Heart
Valves. Biomaterials, 30(26):4445–6.
Schoen, F. J., Tsao, J. W., and Levy, R. J. (1986). Calcification of bovine pericardium
used in cardiac valve bioprostheses. Implications for the mechanisms of bioprosthetic
tissue mineralization. The American journal of pathology, 123:134–145.
Sellaro, T. L. (2003). Effects of collagen orientation on the medium-term fatigue response
of heart valve biomaterials. PhD thesis, University of Pittsburgh.
Seyfert, U. T., Biehl, V., and Schenk, J. (2002). In vitro hemocompatibility testing of
biomaterials according to the ISO 10993-4. Biomolecular engineering, 19(2-6):91–6.
Shen, Z. L., Dodge, M. R., Kahn, H., Ballarini, R., and Eppell, S. J. (2008). Stress-strain
experiments on individual collagen fibrils. Biophysical journal, 95(8):3956–3963.
Sheriff, J., Bluestein, D., Girdhar, G., and Jesty, J. (2010). High-shear stress sensi-
tizes platelets to subsequent low-shear conditions. Annals of biomedical engineering,
38(4):1442–50.
Siddiqui, R. F., Abraham, J. R., and Butany, J. (2009). Bioprosthetic heart valves:
modes of failure. Histopathology, 55(2):135–44.
Sigler, M. and Jux, C. (2007). Biocompatibility of septal defect closure devices. Heart
(British Cardiac Society), 93(4):444–449.
Sih, G. C., Paris, P. C., and Irwin, G. R. (1965). On cracks in rectilinearly anisotropic
bodies. International Journal of Fracture Mechanics, 1(3):189–203.
290 Bibliography
Sim, E. K. W., Muskawad, S., Lim, C.-S., Yeo, J. H., Lim, K. H., Grignani, R. T.,
Durrani, a., Lau, G., and Duran, C. (2003). Comparison of human and porcine aortic
valves. Clinical anatomy (New York, N.Y.), 16(3):193–6.
Simmons, A., Hyvarinen, J., Odell, R. a., Martin, D. J., Gunatillake, P. a.,
Noble, K. R., and Poole-Warren, L. a. (2004). Long-term in vivo biosta-
bility of poly(dimethylsiloxane)/poly(hexamethylene oxide) mixed macrodiol-based
polyurethane elastomers. Biomaterials, 25(20):4887–4900.
Sinn, S., Scheuermann, T., Deichelbohrer, S., Ziemer, G., and Wendel, H. P. (2011). A
novel in vitro model for preclinical testing of the hemocompatibility of intravascular
stents according to ISO 10993-4. Journal of materials science. Materials in medicine,
22(6):1521–8.
Society for Cardiothoracic Surgery in GB and Ireland (2011). National Adult Cardiac
Surgery Audit. Technical report, UCL, London.
Sperling, C., Schweiss, R. B., Streller, U., and Werner, C. (2005). In vitro hemocompati-
bility of self-assembled monolayers displaying various functional groups. Biomaterials,
26(33):6547–6557.
Starcher, B. C. and Urry, D. W. (1973). Elastic coacervate as a matrix for calcification.
Biochemical and Biophysical Research Communications, 53(1):210–216.
Stasiak, J., Moggridge, G. D., Zaffora, A., Pandolfi, A., and Costantino, M. L. (2010). En-
gineering Orientation in Block Copolymers for Application To Prosthetic Heart Valves.
Functional Materials Letters, 03(04):249–252.
Stasiak, J., Nair, S., and Moggridge, G. D. (2014). Mechanical strength of sutured block
copolymers films for load bearing medical applications. Bio-Medical Materials and
Engineering, 24(1):563–569.
Stasiak, J., Squires, A. M., Castelletto, V., Hamley, I. W., and Moggridge, G. D. (2009).
Effect of Stretching on the Structure of Cylinder and Sphere Forming Styrene-Isoprene-
Styrene Block Copolymers. Macromolecules, 42(14):5256–5265.
Stasiak, J., Zaffora, A., Costantino, M. L., and Moggridge, G. D. (2011). A real time
SAXS study of oriented block copolymers during fast cyclical deformation, with po-
tential application for prosthetic heart valves. Soft Matter, 7:11475–11482.
Staudinger, U., Satapathy, B. K., and Weidisch, R. (2008). Influence of block composi-
tion on crack toughness behaviour of styrene-b-(styrene-random-butadiene)-b-styrene
triblock copolymers. European Polymer Journal, 44(6):1822–1833.
Bibliography 291
Streller, U., Sperling, C., Hu¨bner, J., Hanke, R., and Werner, C. (2003). Design and
evaluation of novel blood incubation systems for in vitro hemocompatibility assessment
of planar solid surfaces. Journal of Biomedical Materials Research. Part B, Applied
Biomaterials, 66(1):379–90.
Subhani, M., Kumar, R. K., and Balakrishnan, K. R. (2013). Normal aortic valves stay
open much longer in systole than porcine substitutes. Asian cardiovascular & thoracic
annals, 21(3):275–80.
Sumpio, B. E., Timothy Riley, J., and Dardik, A. (2002). Cells in focus: Endothelial cell.
International Journal of Biochemistry and Cell Biology, 34(12):1508–1512.
Sun, W., Abad, A., and Sacks, M. S. (2005). Simulated bioprosthetic heart valve deforma-
tion under quasi-static loading. Journal of biomechanical engineering, 127(6):905–914.
Swanson, W. and Clark, R. (1977). A simple cardiovascular system simulator: design
and performance. Journal of Bioengineering, 1(2):135–45.
Syedain, Z. H., Meier, L. A., Reimer, J., and Tranquillo, R. T. (2012). Tubular
heart valves from decellularized engineered tissue. Annals of biomedical engineering,
41(12):2645–54.
Talukder, N. and Reul, H. (1978). Fluid Mechanics of Natural Cardiac Valves. In Bauer,
R. and Busse, R., editors, The Arterial System, pages 269–74. Springer Berlin Heidel-
berg.
Tam, H., Zhang, W., Feaver, K. R., Parchment, N., Sacks, M. S., and Vyavahare, N.
(2015). A novel crosslinking method for improved tear resistance and biocompatibility
of tissue based biomaterials. Biomaterials, 66:83–91.
Tanzi, M. C., Mantovani, D., Petrini, P., Guidoin, R., and Laroche, G. (1997). Chemical
stability of polyether urethanes versus polycarbonate urethanes. Journal of Biomedical
Materials Research, 36(4):550–559.
Ten Berge, L. (1958). A flexible cardiac valve prosthesis; preliminary report on the devel-
opment of an experimental valvular prosthesis. Archivum Chirurgicum Neerlandicum,
10(1):26–33.
Teoh, S. (2000). Fatigue of biomaterials: a review. International Journal of Fatigue,
22(10):825–837.
Thomas, A. (1994). The development of fracture mechanics for elastomers. Rubber
chemistry and technology, 67(3):G50–G60.
292 Bibliography
Thomas, V. and Jayabalan, M. (2009). A new generation of high flex life polyurethane
urea for polymer heart valve–studies on in vivo biocompatibility and biodurability.
Journal of biomedical materials research. Part A, 89(1):192–205.
Thubrikar, M., Deck, J., Aouad, J., and Nolan, S. (1983). Role of mechanical stress in
calcification of aortic bioprosthetic valves. The Journal of Thoracic and Cardiovascular
Surgery, 86(1):115–25.
Thubrikar, M. and Klemchuk, P. (1990). The Aortic Valve. CRC Press, Boca Raton, FL.
Thubrikar, M., Piepgrass, W. C., Bosher, L. P., and Nolan, S. P. (1980). The elastic
modulus of canine aortic valve leaflets in vivo and in vitro. Circulation Research,
47(5):792–800.
Thyregod, H., Steinbru¨chel, D., Ihlemann, N., Nissen, H., Kjeldsen, B., Petursson, P.,
Chang, Y., Franzen, O., Engstrøm, T., Clemmensen, P., Hansen, P., Andersen, L.,
Olsen, P., and Søndergaard, L. (2015). Transcatheter versus surgical aortic valve re-
placement in patients with severe aortic valve stenosis: 1-year results from the all-
comers NOTION randomized clinical trial. Journal of the American College of Cardi-
ology, 65(20):2184–94.
Tien, W. H., Chen, H. Y., Berwick, Z. C., Krieger, J., Chambers, S., Dabiri, D., and
Kassab, G. S. (2014). Role of sinus in prosthetic venous valve. European Journal of
Vascular and Endovascular Surgery, 48(1):98–104.
Tokita, N. and White, J. L. (1966). Milling Behavior of Gum Elastomers: Experiment
and Theory. Journal of Applied Polymer Science, 10:1011–1026.
Touma, H., Sahin, I., Gaamangwe, T., Gorbet, M. B., and Peterson, S. D. (2014). Numer-
ical investigation of fluid flow in a chandler loop. Journal of biomechanical engineering,
136(7):1–8.
Trantina-Yates, A. E., Human, P., Bracher, M., and Zilla, P. (2001). Mitigation of
bioprosthetic heart valve degeneration through biocompatibility: In vitro versus spon-
taneous endothelialization. Biomaterials, 22(13):1837–1846.
Trebbin, M., Steinhauser, D., Perlich, J., Buffet, A., Roth, S. V., Zimmermann, W.,
Thiele, J., and Fo¨rster, S. (2013). Anisotropic particles align perpendicular to the flow
direction in narrow microchannels. Proceedings of the National Academy of Sciences
of the United States of America, 110(17):6706–11.
Vaesken, A., Heim, F., and Chakfe, N. (2014). Fiber heart valve prosthesis: Influence of
Bibliography 293
the fabric construction parameters on the valve fatigue performances. Journal of the
mechanical behavior of biomedical materials, 40C:69–74.
Vahanian, A., Alfieri, O., Andreotti, F., Antunes, M. J., Baro´n-Esquivias, G., Baumgart-
ner, H., Borger, M. A., Carrel, T. P., De Bonis, M., Evangelista, A., Falk, V., Iung, B.,
Lancellotti, P., Pierard, L., Price, S., Scha¨fers, H.-J., Schuler, G., Stepinska, J., Swed-
berg, K., Takkenberg, J., Von Oppell, U. O., Windecker, S., Zamorano, J. L., Zembala,
M., Bax, J. J., Ceconi, C., Dean, V., Deaton, C., Fagard, R., Funck-Brentano, C.,
Hasdai, D., Hoes, A., Kirchhof, P., Knuuti, J., Kolh, P., McDonagh, T., Moulin, C.,
Popescu, B. a., Reiner, Z., Sechtem, U., Sirnes, P. A., Tendera, M., Torbicki, A., Von
Segesser, L., Badano, L. P., Bunc, M., Claeys, M. J., Drinkovic, N., Filippatos, G.,
Habib, G., Kappetein, a. P., Kassab, R., Lip, G. Y. H., Moat, N., Nickenig, G., Otto,
C. M., Pepper, J., Piazza, N., Pieper, P. G., Rosenhek, R., Shuka, N., Schwammen-
thal, E., Schwitter, J., Mas, P. T., Trindade, P. T., and Walther, T. (2012). Guidelines
on the management of valvular heart disease (version 2012). European heart journal,
33(19):2451–96.
van Bilsen, P. H. J., Krenning, G., Billy, D., Duval, J.-L., Huurdeman-Vincent, J., and
van Luyn, M. J. a. (2008). Heparin coating of poly(ethylene terephthalate) decreases
hydrophobicity, monocyte/leukocyte interaction and tissue interaction. Colloids and
surfaces. B, Biointerfaces, 67(1):46–53.
Van Steenhoven, A. A. and Van Dongen, M. E. H. (1979). Model studies of the closing
behaviour of the aortic valve. Journal of Fluid Mechanics, 90(01):21.
Varadarajan, P., Patel, R., Turk, R., Kamath, a. R., Sampath, U., Khandhar, S., and Pai,
R. G. (2013). Etiology impacts survival in patients with severe aortic regurgitation:
results from a cohort of 756 patients. J Heart Valve Dis, 22(1):42–49.
Vesely, I. (1998). The role of elastin in aortic valve mechanics. Journal of biomechanics,
31(2):115–23.
Vesely, I. (2003). The evolution of bioprosthetic heart valve design and its impact on
durability. Cardiovascular Pathology, 12(5):277–286.
Vesely, I., Boughner, D., and Song, T. (1998). Tissue buckling as a mechanism of bio-
prosthetic valve failure. The Annals of Thoracic Surgery.
Vienken, J. (2008). Testing biomaterials for application in artificial organs: Impact of
procedures, donor and patient properties. Prilozi, 37(2):25–37.
Vlachopoulos, J. and Strutt, D. (2003). Polymer processing. Materials Science and
Technology, 19(9):1161–1169.
294 Bibliography
Vladkova, T. G. (2013). Surface Engineering of Blood Contacting Polymeric Biomaterials.
In Surface Engineering of Polymeric Biomaterials, chapter 4, pages 231–294. Smithers
Rapra.
Vondracek, P. and Dolezel, B. (1984). Biostability of medical elastomers : a review.
Biomaterials, 5:209–214.
Vongapatanisin, W., Hillis, D., and Ange, R. (1996). Characteristics of prosthetic valves.
The New England Journal of Medicine, August:407–416.
Vyavahare, N. R., Chen, W., Joshi, R. R., Lee, C.-h., Hirsch, D., Levy, J., Schoen, F. J.,
and Levy, R. J. (1997a). Current Progress in Anticalcification for Bioprosthetic and
Polymeric Heart Valves. Cardiovascular Pathology, 6(4):219–29.
Vyavahare, N. R., Hirsch, D., Lerner, E., Baskin, J. Z., Schoen, F. J., Bianco, R., Kruth,
H. S., Zand, R., and Levy, R. J. (1997b). Prevention of Bioprosthetic Heart Valve
Calcification by Ethanol Preincubation. Circulation, 95:479–488.
Wagner, I. M., Eichinger, W. B., Bleiziffer, S., Botzenhardt, F., Gebauer, I., Guenzinger,
R., Bauernschmitt, R., and Lange, R. (2007). Influence of completely supra-annular
placement of bioprostheses on exercise hemodynamics in patients with a small aortic
annulus. The Journal of thoracic and cardiovascular surgery, 133(5):1234–41.
Wang, Q., McGoron, A., and Bianco, R. (2010). In-vivo assessment of a novel polymer
(SIBS) trileaflet heart valve. Journal of Heart Valve Disease, 19(4):499–505.
Weber, N., Wendel, H. P., and Ziemer, G. (2002). Hemocompatibility of heparin-coated
surfaces and the role of selective plasma protein adsorption. Biomaterials, 23(2):429–
39.
Weinberg, S. and King, M. (2004). Medical fibers and biotextiles. In Ratner, B. D.,
editor, Biomaterials Science. An introduction to materials in medicine., chapter 2.4,
pages 86–99. Academic Press.
Weiss, A. S. (2010). The Science of Elastin. Elastagen Media, (February 2011):1–8.
Weiss, R. M., Miller, J. D., and Heistad, D. D. (2013). Fibrocalcific aortic valve dis-
ease: Opportunity to understand disease mechanisms using mouse models. Circulation
Research, 113(2):209–222.
Wendel, H. P. and Ziemer, G. (1999). Coating-techniques to improve the hemocompat-
ibility of artificial devices used for extracorporeal circulation. European Journal of
Cardio-Thoracic Surgery, 16:342–350.
Bibliography 295
Wheatley, D. J., Fisher, J., and Williams, D. (1998). Heart Valve Prosthesis.
White, J. L. and Tokita, N. (1968). Instability and Failure Phenomena in Polymer
Processing with Application to Elastomer Mill Behavior. Journal of Applied Polymer
Science, 12:1589–1600.
Williams, D. F. (2014). There is no such thing as a biocompatible material. Biomaterials,
35:10009–14.
Williams, D. F. (2015). Regulatory biocompatibility requirements for biomaterials
used in regenerative medicine. Journal of Materials Science: Materials in Medicine,
26(2):2014–2015.
Winter, H. H. (1975). Calculation and measurement of the pressure distribution in ra-
dial flow of molten polymers between parallel discs. Polymer Engineering & Science,
15(6):460–69.
World Heart Federation (2012). Rheumatic heart disease. Technical report, World Heart
Federation.
Wright, D. (1996). Environmental stress cracking of plastics. iSmithers Rapra Publishing.
Wurzinger, L., Opitz, R., Blasberg, P., and Schmid-Schonbein, J. (1985). Platelet and
coagulation parameters following millisecond exposure to laminar shear stress. Throm-
bosis and Haemostasis, 54(2):381–6.
Yacoub, M. H. and Takkenberg, J. J. M. (2005). Will heart valve tissue engineering
change the world? Nature clinical practice. Cardiovascular medicine, 2(2):60–1.
Yang, M., Zhang, Z., Hahn, C., King, M., and Guidoin, R. (1999). Assessing the resis-
tance to calcification of polyurethane membranes used in the manufacture of ventricles
for a totally implantable artificial heart. Journal of Biomedical Materials Research,
48(5):648–59.
Yap, C. H., Kim, H.-S., Balachandran, K., Weiler, M., Haj-Ali, R., and Yoganathan, A. P.
(2010). Dynamic deformation characteristics of porcine aortic valve leaflet under normal
and hypertensive conditions. American journal of physiology. Heart and circulatory
physiology, 298(2):H395–405.
Yeh, C. H., Hogg, K., and Weitz, J. I. (2015). Overview of the New Oral Anticoagulants:
Opportunities and Challenges. Arteriosclerosis, thrombosis, and vascular biology, pages
1056–1065.
296 Bibliography
Yoda, R. (1998). Elastomers for biomedical applications. Journal of Biomaterials Science,
Polymer Edition, 9(6):561–626.
Yoganathan, A. P., Chandran, K. B., and Sotiropoulos, F. (2005). Flow in prosthetic
heart valves: state-of-the-art and future directions. Annals of biomedical engineering,
33(12):1689–94.
Yoshida, M., Wearden, P. D., Dur, O., Pekkan, K., and Morell, V. O. (2011). Right
Ventricular Outflow Tract Reconstruction With Contegra Bovine Valved Conduit. The
Annals of Thoracic Surgery, 91(4):1235–8.
Zaffora, A. (2011). Computational method for the design of innovative materials for heart
valve prosthesis. PhD thesis, Politechnico di Milano.
Zarrin-Ghalami, T. and Fatemi, A. (2012). Material Deformation and Fatigue Behavior
Characterization for Elastomeric Component Life Predictions. Polymer Engineering
and Science, pages 1795–1805.
Zdrahala, R. J. and Zdrahala, I. J. (1999). Biomedical applications of polyurethanes: A
review of past promises, present realities and a vibrant future. Journal of Biomaterials
Applications, 14(1):67–90.
Zegdi, R., Bruneval, P., Blanchard, D., and Fabiani, J.-N. (2011). Evidence of leaflet
injury during percutaneous aortic valve deployment. European Journal of Cardio-
Thoracic Surgery, 40(1):257–259.
Zhang, M. and Wu, Q.-C. (2010). Intra-supra annular aortic valve and complete supra
annular aortic valve: a literature review and hemodynamic comparison. Scandina-
vian journal of surgery : SJS : official organ for the Finnish Surgical Society and the
Scandinavian Surgical Society, 99(1):28–31.
Zilla, P., Brink, J., Human, P., and Bezuidenhout, D. (2008). Prosthetic heart valves:
catering for the few. Biomaterials, 29(4):385–406.
Zoghbi, W. A., Chambers, J. B., Dumesnil, J. G., Foster, E., Gottdiener, J. S., Grayburn,
P. a., Khandheria, B. K., Levine, R. a., Marx, G. R., Miller, F. a., Nakatani, S.,
Quin˜ones, M. a., Rakowski, H., Rodriguez, L. L., Swaminathan, M., Waggoner, A. D.,
Weissman, N. J., and Zabalgoitia, M. (2009). Recommendations for Evaluation of
Prosthetic Valves With Echocardiography and Doppler Ultrasound. A Report From the
American Society of Echocardiography’s Guidelines and Standards Committee and the
Task Force on Prosthetic Valves, Developed in Conjunction. Journal of the American
Society of Echocardiography, 22(9):975–1014.