Concerns about the Burden of Proof studies

We read with interest the Burden of Proof (BoP) studies [1][2][3][4][5] in which the authors conducted meta-analyses of epidemiological studies to provide an overall conservative quantitative assessment for several important public health questions.For ease of interpretation, they transformed the overall assessment into a star rating (1-5 stars).Examples include five stars for smoking and lung cancer, two stars for low vegetable intake and ischemic heart disease (IHD) and two stars for unprocessed red meat and type 2 diabetes (T2D), colorectal cancer and IHD 6 .They used this same method to assign just three stars to smoking in relation to IHD and one or two stars to other well-established relationships [1][2][3][4] .However, we believe there are serious methodological issues with their meta-analyses 5 ; the star rating of evidence strength is overly simplistic and could cast doubt on existing recommendations and policies intended to prevent chronic disease and treat illnesses.In the BoP analyses, the estimated uncertainty intervals (UIs) were several times wider than the 95% confidence intervals (CIs) generated by existing methods that account for heterogeneity among methods 7 .As an example of the BoP analyses 1 , for unprocessed red meat and T2D 15 studies were included in the meta-analysis, many of which were large and without evidence of small study bias.With what the authors described as a conventional analysis, there is a highly statistically significant and approximately linear positive association and the relative risk for 100 g per day of red meat intake versus no consumption was approximately 1.24.However, due to an