How much is too much? A methodological investigation of the literature on alcohol consumption and health

Until a few years ago, moderate alcohol consumption was thought to have (mild) beneficial effects on health. However, some recent studies have suggested that “there is no safe level” of alcohol intake. Consequently, public health institutions have responded by advising against any level of alcohol use and suggesting governments a number of policies to reduce overall alcohol consumption. Nonetheless, medical studies suffer from a variety of intrinsic limitations that could undermine the reliability of their findings, especially when focusing on low-intake levels. On the one hand, we show that the literature on alcohol consumption may suffer from publication bias; such a problem is known to be present in the scientific literature in general. On the other hand, we discuss other potential sources of bias, which are inevitable due to the infeasibility of randomized controlled trials. We assess a sample of articles for the presence of omitted variable bias, miscalculation of alcohol intake, use of linear in place of non-linear models, lack of validation of Mendelian randomization assumptions, and other possible weaknesses. We conclude that the claim that “there is no safe level” of alcohol intake is not sufficiently supported based on our current scientific knowledge.