NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Trikalinos TA, Lee J, Moorthy D, et al. Effects of Eicosapentanoic Acid and Docosahexanoic Acid on Mortality Across Diverse Settings: Systematic Review and Meta-Analysis of Randomized Trials and Prospective Cohorts: Nutritional Research Series, Vol. 4. Rockville (MD): Agency for Healthcare Research and Quality (US); 2012 Feb. (Technical Reviews, No. 17.4.)

Cover of Effects of Eicosapentanoic Acid and Docosahexanoic Acid on Mortality Across Diverse Settings: Systematic Review and Meta-Analysis of Randomized Trials and Prospective Cohorts

Effects of Eicosapentanoic Acid and Docosahexanoic Acid on Mortality Across Diverse Settings: Systematic Review and Meta-Analysis of Randomized Trials and Prospective Cohorts: Nutritional Research Series, Vol. 4.

Show details

Discussion

Principal Findings

In RCTs, supplementation with EPA and DHA reduced mortality, primarily in patients with a history of cardiovascular disease or with risk factors for the disease. Meta-regressions of data from prospective cohort studies suggested evidence of an association between higher EPA or DHA mean intake and lower risk of cardiac mortality phenotypes, up to a threshold of 0.20 grams per day. Higher intakes of EPA and DHA were not related with mortality. The conclusions drawn from the analysis of the available prospective cohort and randomized trial data were, for the most part, similar, despite differences in populations, settings, mean EPA and DHA intake, and duration of follow-up. The levels of EPA and DHA associated with lowest mortality risk, either cardiovascular or all-cause, are consistent with the current dietary guidance of the 2010 Dietary Guidelines for Americans58 and 2006 American Heart Association Diet and Lifestyle Recommendations to consume at least two fish meals per week.59.

Concordance Between Randomized Trials and Prospective Cohorts

The hypothesis that the effect of long chain n-3 fatty acids on cardiovascular clinical outcomes plateaus beyond a minimum average daily intake of approximately 0.30-0.50 grams is partly based on ecological observations in Japanese populations, where high average fish intakes are not associated with cardiovascular or overall mortality.4 At the same time, the risk of cardiovascular mortality is lower in Japan than Western countries such as the United States. Noteworthy, the mean daily EPA and DHA intake in Japan is approximately 0.80-0.90 grams per day27,30 versus 0.04-0.13 grams per day in the United States.23,29 These data suggest that supplemental EPA or DHA above a minimum threshold is not efficacious. If anything, there are concerns for environmental contaminants such as mercury60 at high fish intakes.

We identified agreement in the actual magnitude of the summary effects between RCTs and prospective cohorts using formal meta-analysis and meta-regression methods. The summary dose-response odds ratio for cardiac, cardiovascular or sudden cardiac death in prospective cohort studies was 0.64 to 0.88 per 0.20 grams per day of mean intake (using thresholds of 0.20 to 0.30 grams per day for the floor effect, respectively) and the relative risk from RCTs was 0.89 (for supplement doses that are well in excess of 0.20 grams per day). If the background EPA and DHA intake in RCTs is relatively small, the summary effect of EPA and DHA supplementation appears to correspond to that observed up to the threshold of 0.20 grams per day in prospective cohorts. Although no data were reported for background diets in RCTs, most are from western countries where the mean EPA or DHA intake is generally lower than in Japan or in Scandinavian countries.27,30 Further, the treatment effect in RCTs was not significantly associated with supplementation dose, perhaps because supplement doses in all RCTs were well above the plateau threshold.

A strength of this systematic review and meta-analysis is its detailed modelling of adjusted data from observational studies. Most prior meta-analyses of observational data are limited to a synthesis of unadjusted odds ratios across extreme intake categories (e.g., of highest versus lowest intake), thereby omitting useful information from the intermediate categories and including potentially confounded results. Other studies4 attempted to utilize all available data, but did not account for the correlations between the intake-group-specific odds ratios, or the uncertainty that accompanies these odds ratios. The importance of using adjusted data in dose-response meta-regressions is evident by comparing the results with meta-regressions based on unadjusted data: the former suggest that the risk of cardiac death decreases up to a mean EPA and DHA intake of 0.20 grams per day, and does not change in higher mean intakes, consistent with the aforementioned epidemiological observations. In contrast, the latter suggest an effect in the opposite direction, i.e. increased risk of mortality at higher mean intakes of EPA or DHA. A possible explanation is that unadjusted data are subject to confounding within each study.

Our approach has several limitations. Our analyses are based on systematic reviews of published evidence, and therefore may be susceptible to selection biases, publication bias61-63 and outcome reporting bias in particular.64-66 When such biases operate, positive results may be overrepresented among the analysable published data, resulting in summary effects that systematically deviate from the null. We focused on large prospective cohorts (with at least 1000 participants and at least 3 years of follow-up) assuming the results from larger studies are less susceptible to such biases – although we have no empirical data to support this assumption. At the same time, larger studies may not have as rigorous assessment of habitual intake levels as smaller studies, and their results may be subject to larger measurement error compared to smaller experimental studies. While random measurement error is expected to attenuate any associations, larger sample size increases statistical power. For these reasons, it is unclear whether limiting analyses to larger studies biases the summary estimates away from their “true” values.

There is substantial diversity in study characteristics, both among RCTs and among observational studies. However, these differences did not translate to systematic differences in the effect size across studies; if anything, we found suggestive evidence of concordance in the summary results of RCTs and prospective cohorts, despite their clinical and methodological diversity. Insufficient data were available to perform analyses by ethnic descent or gender, so it is unclear whether these factors are effect modifiers. Lastly, especially for analyses in prospective cohorts, we cannot rule of the possibility that higher fish intake displaces other foods, e.g., meat and dairy, that are major contributors of dietary saturated fat and that it is the lower intakes of saturated fat that mainly contribute to lower rates of cardiovascular mortality. Likewise, we cannot rule out the possibility that higher EPA and DHA intake is a marker for a healthier dietary pattern, e.g., more vegetables and fruits or whole grains.

EPA and DHA are examples of nutrients whose dietary reference intake values should be set taking into account the dose-response relationships with the risk of outcomes related to chronic diseases. Thus they are subject to different considerations compared with other nutrients whose reference intakes are set to prevent adverse health outcomes associated with a deficient status.44,63 Information on threshold effects for chronic disease is important in considering reference intake values. Our analyses are consistent with the hypothesis that the beneficial effect of EPA and DHA on mortality reaches a plateau after a mean intake threshold of approximately 0.20 to 0.30 grams per day. Ideally, one would refine this threshold by analysing individual participant data with suitable methodologies (e.g., using isotonic regression modelling,67 and would examine several other outcomes as well. We believe that a meta-analysis of individual participant data would be much more informative than yet another primary study. Notwithstanding ongoing (the VITAL trial68 on vitamin D and EPA and DHA supplementation) or recently completed study (the OMEGA trial69 that was completed in 2008 and whose results are still pending), it is at best unclear whether further RCTs and prospective cohort studies are necessary to assess the effects of EPA or DHA on mortality phenotypes.

PubReader format: click here to try

Views

  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (906K)

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...