Fish Oil Supplements for Prevention of Cardiovascular Disease: Response to Opponents

It was a pleasure to read the “pro” article by DiNicolantonio and O’Keefe,¹ and we commend them for their numerous contributions in the field of preventive cardiology. Nevertheless, we disagree with several of their comments and conclusions regarding the use of fish oil supplements to prevent cardiovascular diseases.

Foremost, we object to the heavy use of secondary endpoints to support their argument that “most ‘negative’ fish oil trials can actually be considered positive.” DiNicolantonio and O’Keefe rely heavily on secondary endpoints in making this statement because many clinical trials have failed to find a beneficial effect of EPA/DHA supplementation on primary endpoints, which are the most consequential output of any clinical trial. Secondary endpoints are important, as they provide key information about an intervention, such as its mechanism of action or associated toxicity. The utility of secondary endpoints, particularly if they are prespecified at the time of study design, is to help interpret the study’s primary results and guide future research.² However, secondary endpoints often lack the statistical power of primary endpoints. As such, they should not be used to interpret a clinical trial as “positive” if there was no statistically significant effect of the intervention on the study’s primary endpoint.³ Put another way, secondary endpoints are at best suggestive, but they are not conclusive, and they should not be used to override negative primary endpoint results and adopt an intervention into clinical practice.^{2, 4–9}

Similarly, an individual component (e.g., myocardial infarction, MI) of a study’s composite primary endpoint (e.g., MI, stroke, or cardiovascular death) should not be teased-out to interpret the study as positive when the primary composite endpoint was negative,⁷ though such data can be used to support hypotheses and inform future prospective studies.

Let’s discuss some specifics. Regarding the VITAL study,¹⁰ the conclusion by DiNicolantonio and O’Keefe that fish oil supplementation produced a “statistically significant 17% reduction in death from coronary heart disease” is incorrect, as Table 2 of this study shows that the difference in coronary deaths between treatment groups was statistically insignificant. Furthermore, this parameter was neither a primary nor secondary endpoint, which was specifically mentioned by VITAL’s authors. In the case of ASCEND,¹¹ vascular death was an individual component of the composite endpoint; as such, it does not provide rigorous evidence of benefit. We are intrigued that our Kansas City colleagues disapprove of including “soft events such as hospitalization” in composite primary endpoints, yet they do not criticize the JELIS¹² or REDUCE-IT¹³ trials, positive fish oil studies, for including hospitalization endpoints such as unstable angina in the composite endpoint. Similarly, we find it surprising that DiNicolantonio and O’Keefe refer to ASCEND as a “failure” because it found no significant effect of n-3 fatty acid (FA) supplementation on the primary composite endpoint of non-fatal MI, ischemic stroke, transient ischemic attack, or vascular death. Negative clinical trials are not failures, rather important findings that help physicians and patients avoid the cost and potential side-effects of treatments that do not provide rigorously determined benefits. Our opponents inaccurately state that “most deaths occur early on after a myocardial infarction.” While the risk of death is higher shortly after an MI than later, this does not mean that most deaths occur within this time window.

DiNicolantonio and O’Keefe cite the important role of meta-analyses in formulating clinical practice guidelines, with which we agree. In citing the meta-analysis by Bernasconi et al.,^{14, 15} our opponents refer to “the totality of this evidence regarding the cardiac effects of omega-3 supplementation.” However, the Bernasconi meta-analysis included trials that were unblinded and lacked a placebo control.^{12, 16} Furthermore, a closer look at the 40 included trials reveals that 9 of them (adding to 7,220 participants) studied populations and/or diseases lacking any cardiovascular connection.

For example, the FOSTAR trial studied the effects of fish oil on knee osteoarthritis without any assessment of cardiovascular outcomes.¹⁷ Also misleading is the statement by DiNicolantonio and O’Keefe that “fish oil supplementation has decades worth of evidence including a recent pooled analysis of 17 prospective studies in 42,466 individuals,” as these 17 studies were not randomized trials of n-3 FA supplementation! Rather, they were epidemiologic studies in which plasma n-3 FA levels (among many other parameters) were measured and correlated with risk of death.

In conclusion, we believe that currently available evidence is still not sufficiently convincing to endorse broad use of n-3 FAs to lower cardiovascular risk. We hold our position of restrictive use pending new data from refined clinical trials, as discussed in our “con” article. Until then, uncontested benefit may lie in the simplicity of recommending two meals of fish weekly to our patients. We appreciate the opportunity to participate in this scientific debate and commend our colleagues for their thoughtful argumentation and faithful disclosure of financial conflicts of interests.