• Public Health
• Diet
• Policy

Studies reporting ‘new’ associations of food ingredients with diseases are common, and sensational headlines appear almost daily in the news media. Thus, in a recent provocative paper, Schoenfeld and Ioannidis,1 randomly selected 50 common ingredients from a cookbook, and reported that 40 were apparently associated with increased cancer risk in peer reviewed studies. Unsurprisingly, most of these associations disappeared in subsequent meta-analyses.1 The net result: increases in media profits, public anxiety and a number of confused politicians. Indeed, the current landscape in nutritional epidemiology research is blighted by an oversaturation of contradictory evidence which risks confusing policy makers, journalists and public about what aspects of the Western diet deserve attention and then intervention. Randomised controlled trails and meta-analyses offer an evidence ‘gold standard’ relatively free of biases. However, trials are simply not feasible, affordable or ethical for many of the most important dietary questions. We therefore fall back on analyses of long-term cohorts, at which point considerable cautions then need to be sounded. Thus, the populations under study may be highly selected (eg, US doctors or nurses), and the results may not be directly generalisable to the …