Do omega-3 fats boost brain function in adults? Are we any closer to an answer?
Alan D Dangour Am J Clin Nutr May 2013 vol. 97 no. 5 909-910
DHA is particularly abundant in brain tissues, and mechanistic data suggest multiple potential roles for DHA in neuronal health (1). It is therefore not surprising that DHA, and a second long-chain omega-3 (n–3) fat, EPA, which are both commonly found in oily fish, are frequently cited as modifiers of cognitive function (2). Several, although by no means all, observational studies report positive associations of omega-3 fatty acid intake with cognition in both children and adults. To date, the majority of randomized controlled trials (RCTs) in adults on the potential benefits of EPA and DHA supplementation have been conducted in older people, in the hope of slowing age-related cognitive decline and preventing dementia. The available RCTs currently provide no evidence of benefit for cognitive function from EPA and DHA supplementation among cognitively healthy older people (3).
In this issue of the Journal, Stonehouse et al (4) report the findings of a small RCT in New Zealand in which 228 adults aged 18–45 y were randomly assigned to receive either 1.2 g DHA or placebo daily for 6 mo. The primary stated aim of the trial was to identify whether DHA supplementation would improve cognitive function. The trial included only adults who reported habitually consuming small amounts of oily fish, thereby making them a population who could potentially benefit from additional DHA in their diet. This was clearly not an easy trial to manage: slightly more than 20% of randomly assigned participants left the study over 6 mo, and these dropouts tended to be significantly younger than those participants who completed the study. Furthermore, the supplements appear to have been rather hard to stomach: nearly 50% of the intervention arm reported mild side effects (compared with 20% in the placebo arm). There were, however, several strengths to the study including the impressively standardized manner in which the computer-based cognitive testing was conducted.
This is the first robustly designed study to test the hypothesis that provision of supplemental DHA in young adults with habitually low dietary DHA intake will boost cognitive performance, and the findings are potentially important. Six months after randomization, serum DHA status was significantly enhanced in the intervention but not in the placebo arm, and the intervention arm scored better than the placebo arm in 2 measures of cognitive function, namely the reaction times of working and episodic memory. Unfortunately, the confidence that can be placed in the study conclusions is severely limited by 2 important issues, one to do with the conduct and reporting of trials and the other to do with assessment methods in trials on cognitive function.
First, and probably most important in the modern era of RCTs, the way in which the trial has been reported leaves considerable uncertainty for the reader. We do not know what primary endpoint was defined a priori by the study investigators: this critical information is not provided in the article nor on the trial register (we were unable to find a published trial protocol). There are now very clear guidelines on the preparation and publication of high-quality study protocols to facilitate the conduct and review of trials (4). These guidelines highlight the critical importance of clearly specifying trial outcomes at the protocol stage so that any reporting biases can be identified and deterred (4). Similarly, there are very clear guidelines on transparent reporting of trials (5). The Consolidated Standards of Reporting Trials (CONSORT) guidelines require “completely defined pre-specified primary and secondary outcome measures” (5).
The lack of an apparently prespecified primary outcome has resulted in the investigators reporting statistical tests on multiple cognitive function outcomes. CONSORT states that “authors should exercise special care when evaluating the results of trials with multiple comparisons [....] In such circumstances, some statistically significant findings are likely to result from chance alone” (5). Indeed, Table 4 in Stonehouse et al's article (4) reports the results of 25 statistical tests of main effects (some but not all sex disaggregated) and of treatment × sex and treatment × APOE interactions, and with this number of tests we would expect to find 1 to 2 significant findings purely by chance. The omission of a prespecified primary outcome in the trial by Stonehouse et al (4) is extremely unfortunate; it significantly reduces confidence in the headline findings and clearly also means that the trial has not been reported according to CONSORT guidelines.
Our second major concern relates to a collective need for investigators working in the area of cognitive function to think more clearly about the tests used to assess cognitive function in adults. Stonehouse et al (4) reported improvements in reaction times for episodic and working memory as a result of supplementation. A rapid review of studies published in the Journal in 2011 and 2012 that investigated cognitive outcomes in adults (n = 10) identified that 29 different cognitive tests were reported and that, of the studies that reported a primary outcome, no 2 studies reported the same primary outcome. Is this a sensible way to proceed? Which of the reported primary outcomes is of the greatest consequence for cognitive health and which is the most clinically relevant? In 2008 an expert group attempted to define outcomes for primary and secondary prevention trials in Alzheimer disease (6) but was unable to select an outcome to be used universally when assessing cognitive decline in primary prevention trials because of a lack of a sufficiently developed knowledge base.
We would suggest that there is an urgent need for experts in cognitive function testing to transparently and objectively propose a small set of outcomes to use in trials on cognitive outcomes in adults. This set of tests could then be used in future trials to at last allow some cross-study comparisons and significantly advance the field. However, until such time as we have greater clarity on the best cognitive tests to use, we would urge that all trials should publish their protocols according to the new Standard Protocol Items: Recommendations for Interventional Trials guidelines and that all trials should be reported according to the internationally recognized CONSORT guidelines.
