The study was conducted within the Fatty Acids and Outcomes Research Consortium (FORCE) (http://force.nutrition.tufts.edu), a consortium of studies with circulating or adipose tissue fatty acid biomarker measurements and ascertained chronic disease events.2829303132 Studies were identified and study investigators were asked to participate based on previous FORCE projects, expert contacts, and literature searches. All prospective studies of adult participants with n-3 PUFA biomarkers measured in blood fractions or adipose tissue and with data on the primary outcome of incident CKD were eligible. A total of 50 potentially relevant studies were identified in May 2020, but 22 (44%) were not eligible because of lack of exposure or outcome data, and 5 (10%) studies were not included because of a limited number of patients with incident CKD (supplementary table S1). Of the remaining 23 cohorts, 19 (83%) participated in this project, while 4 (17%) declined. The study proposal and analytical plan were approved by FORCE, and all participating cohorts have ethics approval with appropriate data sharing agreements, which enabled the current project to be performed. Detailed cohort information is provided in the supplementary methods.

Biomarker levels of n-3 PUFAs (ALA, EPA, DPA, and DHA) were measured by each cohort in plasma phospholipids, erythrocyte phospholipids, total plasma, total serum, cholesterol esters, or several lipid compartments using gas chromatography (supplementary methods). Individual fatty acid levels were expressed as the proportion of total fatty acids.

All outcomes were determined using estimated glomerular filtration rate (eGFR) at baseline and follow-up visits, with the same definitions in all the 19 participating cohorts. The primary outcome was incident CKD using the same definition as used by the Chronic Kidney Disease Genetics (CKDGen) Consortium.33 Briefly, incident CKD was defined as an eGFR that decreased to <60 mL/min/1.73 m² during follow-up among participants with baseline eGFR≥60 mL/min/1.73 m². eGFR was estimated using the 2009 Chronic Kidney Disease Epidemiology Collaboration equation with calibrated serum or plasma creatinine.34 If creatinine levels were not calibrated and were obtained with a Jaffé assay before 2009, then data were corrected by multiplying creatinine levels by 0.95 before eGFR calculation, as described previously.35 A more stringent definition of incident CKD was used in a sensitivity analysis: both eGFR<60 mL/min/1.73 m² and <75% of baseline eGFR during follow-up among participants with baseline eGFR≥60 mL/min/1.73 m².36

We also assessed two secondary outcomes: annual absolute change in eGFR from baseline to the last follow-up as a continuous variable; and ≥40% decrease in eGFR from baseline to the last follow-up.37 Unlike the primary outcomes, analysis of the secondary outcomes comprised all participants, including those with baseline eGFR<60 mL/min/1.73 m².

A prespecified and harmonised protocol with the same participant inclusion and exclusion criteria, exposures, outcomes, covariates, and analytical methods were sent to all cohorts to perform de novo analyses of individual data. For the outcomes of incident CKD and ≥40% decrease in eGFR, Cox proportional hazards models were used to estimate hazard ratios, considered to approximate relative risk, for cohorts with eGFR available in two or more follow-up visits. In cohorts with data available for only one follow-up eGFR, Poisson regression models were used to estimate relative risks. For the outcome of annual absolute change in eGFR from baseline to the last follow-up, multivariable linear regression was used. All analyses estimated robust standard errors.

Because n-3 PUFA biomarker levels were measured in different lipid compartments in the participating cohorts, they were evaluated as continuous variables using study specific interquintile range (difference between midpoint of lowest fifth (10th centile) and highest fifth (90th centile)).29303132 To assess potential nonlinear associations, n-3 PUFA levels were also evaluated as variables (in fifths) using study specific cut-off points. In the primary multivariable model (model 1), analyses were adjusted for age, sex, race, clinical centre or field site, education, occupation, body mass index, smoking, alcohol intake, physical activity, prevalent coronary heart disease, and use of lipid lowering drugs at baseline, when applicable. In model 2, additional covariates that could confound or mediate the association of n-3 PUFAs with incident CKD were considered, including prevalent diabetes mellitus, urine albumin-creatinine ratio, systolic blood pressure, and use of antihypertensive drugs at baseline (see supplementary methods for definitions of all covariates). Participants with missing data on n-3 PUFAs were excluded from the analysis. Missing data on covariates were handled according to the usual practice of each cohort and study investigators (see supplementary methods for cohort specific details).

In each study, a statistical interaction was evaluated by fitting the n-3 PUFA biomarker (continuous), each potential effect modifier, and their cross product. The prespecified effect modifiers included age (≥60 v <60 years), race, hypertension, diabetes, coronary heart disease, and eGFR (60-89 v ≥90 mL/min/1.73 m²) at baseline. These modifiers were selected by considering their potential biological relevance.

Effect estimates from each participating cohort were provided to the lead author in standardised electronic forms and pooled using inverse variance weighted meta-analysis. Results from participating cohorts were pooled overall and within each specific type of lipid compartment. For studies with measures in multiple compartments, we prioritised the compartments in the overall pooled analysis on the basis of which best reflect long term intake, as prespecified in the following order: erythrocyte phospholipids, plasma phospholipids, cholesterol esters, and total plasma or serum.22 In sensitivity analyses, the overall pooled analyses were performed after excluding data from one cohort at a time or using alternative biomarker compartments. To assess potential nonlinear relations, exploratory analysis was conducted to model the association within each compartment using meta-regression with restricted cubic splines constructed from variables (in fifths) using study specific cut-off points.282930313238 Interaction terms were pooled using inverse variance weighted meta-regression with Bonferroni correction for multiple testing, and such analysis was considered exploratory. All meta-analyses were performed using Stata (version 16.0) or R (version 4.1.0), with two tailed P<0.05 considered statistically significant.

Patients and members of the public were not directly involved in this research owing to lack of funding, staff, and infrastructure to facilitate their involvement in this pooled analysis of data from 19 participating cohorts.

The primary outcome analysis included 25 570 participants without prevalent CKD from 19 cohorts in a total of 12 countries (table 1). Among the 19 cohorts, the mean age of participants ranged from 49 to 77 years and the mean body mass index from 23.2 to 28.3. The weighted median follow-up duration was 11.3 years. Sixteen cohorts recruited men and women, and most (n=15) recruited predominantly white participants (supplementary table S2). The mean baseline eGFR ranged from 76.1 to 99.8 mL/min/1.73 m² (supplementary table S3), and the mean annual change in eGFR ranged from −6.2 to 0.3 mL/min/1.73 m²/year. In total, 4944 participants (19.3%) developed the primary outcome of incident CKD during follow-up (supplementary table S4).

Most of the participating cohorts measured fatty acid levels in erythrocyte or plasma phospholipids (n=11), followed by total plasma or serum (n=7), and cholesterol esters (n=3; table 1). The proportion of total fatty acids for each n-3 PUFA varied between different lipid compartments (supplementary table S5; supplementary fig S1).

Higher levels of total seafood derived n-3 PUFAs were associated with 8% lower risk of incident CKD per interquintile range with low heterogeneity in the observed associations between cohorts (relative risk 0.92, 95% confidence interval 0.86 to 0.98; P=0.009, I²=9.9%; table 2 and fig 1). Findings were not appreciably altered by further adjustment for covariates that could be potential confounders or mediators (0.91, 0.85 to 0.97; P=0.006, I²=17.2%). Similar results were obtained when the n-3 PUFAs were assessed categorically—participants with total seafood n-3 PUFA level in the highest fifth had 13% lower risk of incident CKD than those in the lowest fifth (0.87, 0.80 to 0.96; P=0.005, I²=0.0%; table 2 and supplementary fig S2). Protective associations were found for the individual seafood n-3 PUFAs (table 2), and associations appeared consistent across lipid compartments (fig 1; supplementary table S6). Little evidence could be found of nonlinear associations between total seafood n-3 PUFAs and incident CKD in any of the biomarker compartments (supplementary fig S3).

When incident CKD was defined using a more stringent definition (new onset of eGFR<60 mL/min/1.73 m² and <75% of baseline eGFR) in a sensitivity analysis, the results were largely consistent with the primary outcome analysis. Higher levels of total seafood n-3 PUFAs were associated with a lower risk of incident CKD (0.91, 0.84 to 0.99 per interquintile range; P=0.009, I²=27.3%; supplementary table S7). In additional sensitivity analyses, the association of total seafood n-3 PUFAs with incident CKD remained significant after excluding data from one cohort at a time, or using alternative biomarker compartments in the overall pooled results (supplementary table S8).

Total seafood n-3 PUFA level was not associated with the outcome of ≥40% decrease in eGFR in continuous analyses, but was associated with a lower risk (about 15%) for those in the highest fifth versus those in the lowest fifth (relative risk 0.85, 95% confidence interval 0.74 to 0.98; P=0.03, I²=44.1%; table 3).

Higher levels of total seafood n-3 PUFAs, especially DHA, were associated with a slower annual decline in eGFR (table 4). For instance, the annual decline in eGFR was 0.07 mL/min/1.73 m² lower (95% confidence interval 0.02 to 0.13; P=0.007, I²=42.2%) for people with total seafood n-3 PUFA level in the highest fifth than those in the lowest fifth.

ALA was not associated with incident CKD, with low heterogeneity in the observed associations between cohorts (relative risk 1.00 per interquintile range, 95% confidence interval 0.94 to 1.06; P=0.94, I²=5.8%; table 2 and supplementary fig S4). No significant association was observed in analyses using a more stringent definition of incident CKD (supplementary table S7). No significant association was observed in the secondary outcomes of ≥40% decrease in eGFR (0.98, 0.89 to 1.07; P=0.08, I²=35.8%; table 3) and annual change in eGFR (mean difference −0.03 mL/min/1.73 m² per interquintile range, 95% confidence interval −0.06 to 0.00; P=0.16, I²=24.1%; table 4).

The association of n-3 PUFAs with the primary outcome of incident CKD was not modified by key demographic and clinical characteristics at baseline, including age, race, hypertension, diabetes, coronary heart disease, and eGFR (P for pooled interaction terms >0.05). No difference in association was observed based on study location, follow-up duration, and sample size (supplementary tables S9-11).

This de novo pooled analysis of 19 cohort studies gathered data from more than 25 000 patients as part of the FORCE consortium. We found that higher levels of seafood n-3 PUFA biomarkers were associated with a modestly lower risk of incident CKD and slower decline in renal function, whereas these associations were not found with higher levels of plant n-3 PUFAs (ALA). The results were consistent across a range of sensitivity analyses, highlighting the robustness of the findings. Similar trends were observed using secondary outcomes of >40% decrease in eGFR and annual changes in eGFR. These findings support a favourable role for increased intake of seafood n-3 PUFAs for the primary prevention of CKD.

Randomised controlled trials found increased intake of seafood n-3 PUFAs reduced blood pressure, a key risk factor for CKD development.9 Conversely, ALA generally exhibits weaker to no effect on metabolic risk factors such as lipid, glucose, and inflammatory profile in randomised controlled trials compared with EPA and DHA.39 Furthermore, findings from previous experimental animal studies provide strong biological plausibility similar to our results of the beneficial effects of seafood n-3 PUFAs on CKD.404142 For example, in a mouse model of nephrectomy induced chronic renal failure, supplementation with EPA and DHA for 12 weeks mitigated tubulointerstitial injury through reductions in fibrosis, inflammation, and oxidative stress.40 Increased endogenous production of DHA by a transgenic technique also protected against the development of kidney fibrosis and inflammation in mice with unilateral ureter obstruction induced nephropathy.41 Renoprotective mechanisms might be mediated through the endogenous conversion of n-3 PUFAs to specialised inflammation resolving mediators such as resolvins.4344 Most previous experiments focused on EPA and DHA, but less is known about DPA, although limited evidence also suggests DPA has favourable properties on cardiometabolic profile.4546

Although our findings do not prove a causal relation between seafood n-3 PUFAs and CKD risk, they are supportive and consistent with current clinical guidelines that recommend adequate intake of seafood as part of healthy dietary patterns, especially when seafood replaces the intake of less healthy foods.474849 Although ALA could be enzymatically transformed to the longer chain n-3 PUFAs like EPA and DHA, such endogenous conversion only occurs at a low rate27 and our findings suggest intake of this plant derived n-3 PUFA alone might not maintain renal health. However, although ALA is an essential fatty acid, circulating ALA generally has a weaker correlation with ALA intake compared with the seafood derived n-3 PUFAs,225051 potentially explained by a higher oxidation rate of ALA compared with other fatty acids.52 Therefore, if ALA intake were associated with incident CKD, we might not be able to detect such an association using an ALA biomarker. In our study, the levels of EPA and DHA probably mostly reflected variations in dietary intake from seafood, rather than n-3 PUFA supplements, because relatively small proportions of the general population used n-3 PUFA supplements, especially for studies that were conducted before the early 2000s when fish oil supplements were infrequently used (most of the included cohorts).53 Current dietary guidelines generally recommend at least two servings per week of oily fish to provide around 250 mg/day of long chain n-3 PUFAs for the general population.2749

Given the renoprotective effects of seafood n-3 PUFAs observed in experimental animal studies, human trials have also assessed the use of n-3 PUFA supplementation to prevent decline in kidney function among patients with existing CKD or kidney failure. A recent systematic review of randomised controlled trials found n-3 PUFA supplementation could improve lipid profile and reduce oxidative stress, but not blood pressure in patients with CKD.54 Another recent systematic review of randomised controlled trials found n-3 PUFA supplementation could prevent the progression to kidney failure in patients with CKD who were not receiving renal replacement therapy, although the certainty of evidence is low.55 Therefore, current clinical guidelines do not recommend the use of n-3 PUFA supplementation to prevent further decline in renal function for those with existing CKD, although using n-3 PUFA supplementation to treat hypertriglyceridaemia is recommended.56 Our findings highlight the need for large randomised controlled trials to assess increased intake of seafood n-3 PUFAs for the primary prevention of CKD and decline in renal function. One such randomised controlled trial of 1312 patients with type 2 diabetes found that fish oil supplementation (daily dose of 1 g, containing 465 mg EPA and 375 mg DHA) did not affect the change in eGFR over five years.57 Several other randomised controlled trials have also reported null or small beneficial effects of n-3 PUFA supplementation on kidney function, but these were secondary outcomes or post hoc analyses and were not powered to detect incident CKD as a primary outcome.5859 Further, randomised controlled trials of n-3 PUFA supplementation often included participants with sufficient seafood intake, in which additional n-3 PUFA supplementation might not confer additional cardiometabolic benefits.60 Therefore, future randomised controlled trials could target increased seafood intake, instead of n-3 PUFA supplementation, because consumption of seafood can replace the intake of less healthy foods. Finally, all these trials tested n-3 PUFA supplementation mainly in high risk populations, limiting the applicability of the study findings to the general population.

Our study has several strengths. Our literature search only identified two high quality prospective observational analyses of the association of n-3 PUFA biomarkers with incident CKD or decline in kidney. These two studies were included in our pooled analysis with updated patient numbers and harmonised analytical protocols.6162 Our findings expand upon these previous individual cohort based analyses with a sevenfold larger sample size and enhanced statistical power to more precisely quantify the association of n-3 PUFAs with CKD. The large sample size also enabled a detailed evaluation of potential effect modifiers of the association between n-3 PUFAs and CKD. By collaborating with participating cohorts using a standardised approach to define exposures, covariates, and outcome variables, we reduced potential heterogeneity that is inherent in a publication based meta-analysis. Our collaborative approach enabled the inclusion of most of the cohorts with available exposure and outcome data, which reduced the likelihood of publication bias.

The use of n-3 PUFA biomarkers allowed us to distinguish between individual n-3 PUFAs that could have differing biological effects,26 and avoid measurement errors related to self-reported dietary intake. Conducting analyses across cohorts from diverse demographic, dietary, and medical backgrounds with different incident CKD rates enhanced the generalisability of our findings, and the use of several related secondary outcomes and different sensitivity analyses enhanced the robustness of the findings. However, more studies are needed to further understand the generalisability of our findings, especially in countries experiencing high incidence of CKD such as China and India.

Our study also has some limitations. The n-3 PUFA biomarkers and covariates were only measured once at baseline and this might lead to increasing random misclassification over time, which would tend to bias the associations towards the null. However, n-3 PUFA biomarkers have good long term reproducibility as determined in a previous study with repeated within person measurements across 13 years.63 While we standardised our methods across cohorts as much as possible, fatty acid assays, determination of outcomes, and covariate measurements varied by cohort and might contribute to the heterogeneity in our findings. Among different prespecified covariates, many participating cohorts did not have data for urine albumin-creatinine ratio. However, in cohorts that measured urine albumin-creatinine ratio, association of n-3 PUFA levels with incident CKD was similar whether it was adjusted for or not (results not shown). Although, the possibility of residual confounding by other unmeasured or imprecisely measured covariates could not be excluded, the differential association of plant derived ALA and seafood n-3 PUFAs with incident CKD suggests residual confounding was unlikely to entirely account for our findings. In this analysis, data were not adjusted for n-6 PUFA biomarker levels because experimental evidence on their relation with kidney function was limited. Because n-6 PUFA levels are associated with a lower CVD risk,30 future studies assessing their association with incident CKD are warranted, pending more experimental work assessing their possible impact on CKD.

In this large de novo pooled analysis across 19 cohorts with more than 25 000 patients, higher seafood n-3 PUFA levels were associated with a lower incident CKD risk and a slower decline in renal function. Although the magnitude of these associations was modest, our findings suggest adequate consumption of seafood and oily fish should be part of healthy dietary patterns. Additionally, further randomised controlled trials are warranted to assess the potential beneficial role of seafood n-3 PUFAs in preventing and managing CKD.