Aims Despite contemporary statin therapy, patients with atherosclerotic cardiovascular disease or high cardiovascular risk experience a residual risk of major adverse events. Eicosapentaenoic acid (EPA) has been evaluated for additional cardiovascular risk reduction, but the consistency and robustness of trial results remain uncertain. The aim of this study was to assess the robustness of randomized controlled trials (RCTs) of highly purified EPA for cardiovascular outcomes. Methods MEDLINE and Scopus were searched for phase 3/4 RCTs of EPA published through June 2025. Eligible studies were randomized and placebo-controlled, and reported dichotomous cardiovascular outcomes. Three trials met the inclusion criteria: REDUCE-IT, RESPECT-EPA, and JELIS. Primary and secondary cardiovascular endpoints were extracted. Conventional effect estimates (hazard ratios, relative risk reduction, number needed to treat) were calculated, and robustness was evaluated using the fragility index and fragility quotient. Results EPA significantly reduced the primary composite endpoint in REDUCE-IT (17.2 vs. 22.0%; hazard ratio 0.75; fragility index 123, fragility quotient 0.01), RESPECT-EPA (9.1 vs. 12.6%; hazard ratio 0.71; fragility index 49, fragility quotient 0.01), and JELIS (2.8 vs. 3.5%; hazard ratio 0.81; fragility index 15, fragility quotient 0.01). Secondary endpoints showed consistent but heterogeneous reductions in nonfatal myocardial infarction, stroke, and revascularization, with robustness highest in REDUCE-IT and more fragile results in RESPECT-EPA and JELIS. Differences in population phenotypes, background statin therapy, trial design, and endpoint definitions contributed to variability in the effect size and fragility. Conclusion EPA added to statins lowers cardiovascular events, most robustly in high-risk patients with elevated triglycerides. Trial design, endpoints, and patient characteristics drive heterogeneity, while fragility analyses complement conventional metrics in interpreting outcomes.
Zuin et al. (Wed,) studied this question.