Internal snapping hip, also referred to as coxa saltans or “dancer’s hip,” is a relatively common and potentially debilitating disorder produced by the iliopsoas tendon slipping over soft tissue and bony prominences during hip movement, causing an audible clicking that is typically accompanied by pain. The condition was originally described by Nunziata and Blumenfeld (1951) and garnered growing interest in the last decade, coinciding with the increasing use of hip arthroscopy. The pioneering authors described the snapping as a consequence of the proximal portion of the iliopsoas tendon gliding over the iliopectineal eminence (Lyons and Peterson, 1984). Howse (1972), however, believed that a “clicking hip” in ballet dancers was caused by the iliofemoral ligaments sliding over the femoral head. These historical suppositions were mainly speculative, with little supporting scientific evidence.

Jacobson and Allen (1990) were among the first to perform anatomical and radiographic studies, associating the iliopsoas tendon snap with its movement across the femoral head from lateral to medial with the hip brought to extension. The exact pattern of psoas tendon movement, however, remains debated and unresolved to date, with most authors claiming snapping to occur either over the iliopectineal eminence or the anterior part of the caput femoris (Taher and Power, 2003; Verhelst et al., 2012; Lee et al., 2013; Philippon et al., 2014; Ilizaliturri et al., 2015; Yen et al., 2015). Furthermore, several anatomical variations and focal pathological conditions, e.g., labral cysts, have been reported to increase the likelihood of occasional snapping to become recurrent and therefore clinically symptomatic.

While the location of tendon movement remains debated, literature does agree on the hip positions which can provoke coxa saltans. In all descriptions of the condition it occurs predominantly when the hip is moved from combined flexion, abduction, and external rotation (FABER) back to neutral adduction and internal rotation (Winston et al., 2007; Yen et al., 2015). This explains why snapping hip can be provoked when climbing stairs or when one gets out of a car or gets up from a chair. This also explains its higher incidence in athletes who walk, dance, lift weight and play football (Lee et al., 2013). For competitive ballet dancers it is reported that symptoms mount up to 90% and bilateral involvement as high as 80% (Nolton and Ambegaonkar, 2018). Winston et al. (2007) reported snapping in ballet dancers to occur almost exclusively during abduction-external rotation activities: grande battement à la seconde (a high kick to the side with a straight leg) in 41.8%, grand plié (a deep knee bend) in 25.3%, and développé à la second (a slow extension of the leg from the FABER position) in 22.8%. In the general population the prevalence of audible snapping is estimated at 5–10%, with a higher frequency in females as compared with males (Taher and Power, 2003; Lee et al., 2013; Yen et al., 2015).

Despite the condition being relatively common, little is known about the exact mechanism or origin of psoas tendon snapping. Notwithstanding the numerous theories proposed on the topic, in depth analysis of psoas mechanics during movement has not been carried out previously. This is mainly due to its difficult dynamic visualization and the condition therefore largely remains unsubstantiated.

Studying correlations within the virtual cohort based on isolated discrete measures, e.g., femoral anteversion, would obviously brake any important covariations of shape and result in noisy data describing small effects, hence the advantage of using PCA. Nevertheless, discrete measures are valid anatomical descriptions and usually well known by clinical researchers. To be able to translate our findings to the clinical field, a series of commonly used radiographical measures were therefore studied additional to the PCA in the cohort. Findings are described in detail in Tables 1, 2.

PCA is a celebrated linear data compression technique. Although frequently performed, it is absolutely incorrect to interpret isolated PCs as if they were truly existing, independent anatomical variants. PCA can be used anatomically truthful, however, studying observed PC combinations instead of isolated PC findings, with the use of use multivariable statistics. To this end, we firstly evaluated the correlation between the individual shape components and the peak angular velocity of tendon movement. Several contributing PC’s were observed, conferring small to medium effects on peak tendon excursion. Because of the disproportionate size of the study cohort, p values were all small and even converged to zero for some of the variables. Based on the findings it was decided only the first 20 shape components were relevant to further include in the statistical analysis. Effect size and statistical significance of the correlation of these individual shape components are presented in a Manhattan-like plot in Figure 4.

In a second stage, the correlation between the composite of shape descriptors and peak tendon velocity was defined by CCA. The observed correlation was considerable in both males (r = 0.72, p < 1e-300) and females (r = 0.66, p < 1e-300). The consensus shape configurations relating to the variance in peak tendon velocity were defined using PLSR (Figures 5, 6) images below at ±3 SD. In both populations, abnormal tendon excursion correlated with changes predominantly in the femoral anatomy: increasing femoral head anteversion, valgus, torsion and medial bowing. In essence a rotational variant in femoral shape configuration presenting with a medially located minor trochanter in respect to the pelvis. Interestingly, such shape configuration translates as a decreasing ischiofemoral distance.

A similar pattern of tendon movement was observed in all cases. With ongoing abduction and femoral external rotation, the psoas tendon gradually follows a smooth excursion from the medial portion of the anterior femoral head to lateral, reaching a lateralized position at the valley of the femoral head neck junction (Figures 5, 6).

Upon return to the neutral position, however, the tendon tends to remain hooked in this valley – a stable local minimum in its shortest path position – before initiating its return back medially, causing a sudden increase in tendon velocity at the end of the movement when finding its new optimal medial position over the top of the femoral head. Differences in tendon velocity were observed nearly exclusively in the final stage of the movement, and it was decided to run further statistics against the peak velocity value at this point.

Females were slightly more prone to increased peak tendon movement as compared to males (p < 1e-60) The peak angular velocity was found to be slightly higher, 87.5 deg/s in females (range 27.5–247.5 deg/s) as compared to 82.5 deg/s in males (range 19.8–212.5 deg/s).

For the purpose of clinical translation of the findings, digitally reconstructed x-rays from the ±3 SD 3d shapes with low and high tendon velocity were generated and presented in Figure 7.

With larger datasets becoming available, in combinations with increased computational resources, statistical and probabilistic modeling presents an exciting and novel means for non-invasive testing and the evaluation of physiology and biomechanical variability across populations (Bischoff et al., 2014; Gosselin et al., 2014; Sims et al., 2018) In the context of musculoskeletal disease, pathology usually represents an extreme or outlier condition of function, shape, kinetics, or biomechanics at large. The unmet advantage of population numbers in virtual population models allows for the identification and study of these extremes, in cohort sizes beyond what is clinically achievable (Campbell and Petrella, 2016). Based on the current study on a virtual population of 40,000 cases, we were indeed able to describe variation in psoas movement and identify excessive psoas excursion in selected cases.

As mentioned earlier, opinions vary notably on the exact location of psoas tendon movement. Both the femoral head and iliopectineal eminence have been suggested at possible regions of involvement (Howse, 1972). In the present study, the tendon was found to move exclusively laterally and back to the tendon groove. The majority of tendon movement therefore occurred over the femoral head. This finding is in agreement with the sparse in vivo dynamic visualization data is available on the topic. Deslandes et al. (2008) observed a similar pattern of motion on dynamic ultrasonography and even claimed absolutely no involvement of the iliopectineal eminence at all. We can, however, based on the particular movement studied, not exclude snapping to occur over the ischiofemoral eminence. In particular during adduction a relative more medial positioning of the minor trochanter might result in such snapping mechanism. Further studies are definitely warranted.

Numerous associated factors, anatomical variants, and pathological conditions, have been reported to increase the risk of psoas movement to become symptomatic snapping. It can be anticipated, that any condition hampering a smooth medial return of the tendon during adduction will further trigger the snapping mechanics. Several clinical studies have indeed described a positive correlation between symptomatic snapping and the presence of labral cysts. The observation of such relationship has not led to any solid etiological proof to date, nevertheless (Blankenbaker et al., 2012). Further dynamic studies are required to elucidate whether snapping becomes symptomatic in the presence of labral pathology or whether labral pathology can occur in case of increased and recurrent psoas tendon movement. Other conditions and structures have also been associated including the iliopsoas bursa, undoubling of the psoas tendon, the ligamentum iliofemoralis, labrum lacerations, paralabral cysts, ligamentum teres lacerations, and instability of the joint (Verhelst et al., 2012; Lee et al., 2013; Philippon et al., 2014; Yen et al., 2015). Moreover, and likely ensuing a similar pathway, snapping has been reported following trauma and surgery of the hip joint (Taher and Power, 2003; Regev et al., 2011; Van Riet et al., 2011; Lee et al., 2013). Similarly, coxa saltans has been reported to occur after total hip prosthesis surgery due to the prosthesis protruding or incorrect positioning of the prosthetic components (Pattyn et al., 2011; Lee et al., 2013; Yen et al., 2015).

A particular interesting observation within this study is the increase in tendon excursion with a decreasing ischiofemoral distance. Ischiofemoral impingement is rather newly described condition with reports describing as well an audible snapping in a significant proportion of the patients reported (Stafford and Villar, 2011). The findings of the presents study seem to associate both snapping hip syndrome and ischiofemoral impingement with a similar underlying torsional femoral dysplasia. In the case of ischiofemoral impingement, excess femoral anteversion has indeed been documented in recent studies (Gomez-Hoyos et al., 2016). This observation might explain the disappointing outcome of psoas lengthening in anteverted femurs in the literature before ischiofemoral impingement was recognized as a distinct clinical condition (Fabricant et al., 2012). Further clinical studies are definitely warranted to elucidate this observed association as to define whether for distinct cases of snapping varisating derotational osteotomy is perhaps more in place than cutting the psoas tendon or removing the lesser trochanter.

The present work needs to be interpreted within is methodological limitations. Firstly, the study remains an in silico experiment which will require further in vivo confirmation and replication. Secondly, an important restriction of the presented work relates to the population model implemented, namely statistical model of shape obtained from Belgian people and the unknown extent of which findings can be extrapolated to other populations. The complex interaction between gens, environment, and culture, results in a population-based variation. Numerous studies have indeed demonstrated that the appropriate evaluation of this variation necessitates specific standards for each population (Rissech et al., 2013; San-Millan et al., 2017). Nonetheless, in general terms we expect our anatomical model results to be representative for a Western European population.

A second study limitation relates to the motion that was investigated. Snapping has been reported to occur during a broad spectrum of activities, ranging from simple walking to complex ballet exercises. In the present report a circumduction including abduction, flexion and external rotation back to neutral extension and adduction was investigated in order to cover a wide kinematic space. Even though an almost unphysiological 90 degrees of abduction was included in the analysis, increased tendon velocity was only observed in the final part of the motion sequence, close to the return to neutral position. This finding suggests researchers can restrict their analysis in future work in a smaller kinematic range as compared to what we have simulated. Further, this observation suggests more focused research is warranted to investigate common ADL activities in close range to a neutral hip position such as walking and stair climbing. While these activities have been reported to potentially evoke snapping, they mainly involve flexion-extension, with limited abduction and rotation. Considering we investigated a multiplanar hip motion, our findings in terms of the location and mechanism of the snapping can therefore not be generalized to ADL activities that are believed relevant but mainly involve pure flexion-extension. Clearly more work is needed.

In conclusion we found the psoas tendon to exclusively wrap around the anterior femoral head, not conflicting with the iliopectineal eminence. It was associated with female sex, femoral malrotation and a decreased ischiofemoral distance.

The datasets generated for this study are available on request to the corresponding author.

All authors contributed to the study design, data acquisition, and drafting.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.