Walker · The American journal of clinical nutrition 2019 · pooled individual participant data regression analysis · n=1422

Predicting the effects of supplemental EPA and DHA on the omega-3 index.

Cited 94 times in the scientific literature.

Level 1 - systematic review of randomized trials

Pooled individual participant data analysis from 14 intervention trials

PubMed 31396625 · doi:10.1093/ajcn/nqz161 · record verified 2026-08-26

What was done

Stepwise regression modeling using the Bayesian information criterion was conducted on pooled, deidentified individual participant data from 14 published omega-3 intervention trials (n = 1,422; 846 supplemented, 576 placebo controls) to predict post-supplementation erythrocyte EPA + DHA concentrations (omega-3 index [O3I]). All included trials used the same assay method to measure O3I. Potential predictors evaluated included daily dose, baseline O3I, chemical formulation (ethyl ester [EE] vs. triglyceride [TG]), treatment duration, sex, age, height, and weight.

What was found

Supplemented participants took a mean ± SD of 1,983 ± 1,297 mg/d of EPA + DHA for an average duration of 13.6 ± 6.0 weeks, whereas controls took none. O3I increased from 4.9% ± 1.7% to 8.1% ± 2.7% in supplemented individuals (P < 0.0001). The final predictive model retained only dose, baseline O3I, and chemical formulation, explaining 62% of the variance in response (P < 0.0001). For a baseline O3I of 4.9% and an intake of 850 mg/d EPA + DHA EE, the predicted final O3I was ~6.5% (95% CI: 6.3%, 6.7%). Gram for gram, TG-based formulations increased O3I by approximately 1 percentage point more than EE formulations. Sex, age, body weight, height, and trial duration were not significant predictors.

Why it matters

This study provides an empirical model to estimate erythrocyte omega-3 response based on dose, starting levels, and chemical form, confirming higher bioavailability of triglyceride over ethyl ester formulations in raising the omega-3 index.

Limits

The abstract does not describe external out-of-sample validation of the model. Thirty-eight percent of the variance in individual response remains unexplained. Clinical endpoints were not evaluated, and specific inclusion criteria or risk-of-bias characteristics of the 14 contributing trials are not detailed in the abstract.

Cited by