Comment on “Limitations of Observational Evidence: Implications for Evidence-Based Dietary Recommendations”^,

Dear Editor,

In their recent article, Maki et al. (1) wrote about the use of observational studies and randomized controlled trials (RCTs) when recommendations for nutrition should be evidence based. Although the authors state that they do not want to minimize the importance of observational data, especially in areas where RCTs cannot be conducted for practical or ethical reasons, they list a range of possible limitations of the observational design and emphasize how the design of the RCT is more valid. Indeed, it is a great challenge to conduct research on the basis of observational data. However, with the philosophy of the hierarchy of evidence, and now also with the article by Maki et al., the design of the RCT is conferred too much validity and the observational design too little. Examples of conflicting results from observational studies and RCTs are often used, as by Maki et al., as an argument for the limited validity of the observational study. However, when systematically summing up the evidence of reviews that look at comparable observational studies and RCTs across different medical topics, the results are similar (2,3). This has also been shown by the Cochrane Collaboration when comparing randomized and nonrandomized trials (4). This suggests that when the studies are similar, except for randomization, observational studies and RCTs reach similar results and the potential risk for confounding is not as widespread as often thought. Supporting this, it has been shown that the observational studies and RCTs investigating the controversy of hormone replacement therapy and coronary heart disease, mentioned by Maki et al. as an example of diverged evidence, investigated very different scientific questions; and when observational data were analyzed in a way comparable to an RCT, the result was similar (5). A crucial notion here is that observational studies and RCTs can vary in many ways, making it very difficult to compare the studies and their results. This also applies to nutritional research. While observational studies often investigate the long-term effects of habitual diet in a broad population, RCTs often investigate the short-term effect of an intervention diet among a selected subgroup of the general population. It can be argued that neither the observational study nor the RCT shows causal inferences (6) and that RCTs are of less value than observational studies since they include the problematic of a selected study population and the need for intention-to-treat analyses (7,8). However, overall, it is counterproductive to argue that one method is superior to another, as also stated by others (9), and relevant for nutritional research (10). Instead, observational studies and RCTs should be used as complementary knowledge, the quality of each study should be evaluated according to the merits of the design, and no single study should be the base for any recommendation.