Randomly selected 50 ingredients from a cookbook and searched the literature for studies linking each to cancer risk; 40/50 ingredients had at least one such study, and among those, 39% reported increased risk, 33% decreased risk, 5% borderline, and only 23% no association. Illustrates how routinely nutritional-epidemiology study designs, multiple testing, and publication practices generate “X causes/prevents cancer” findings for nearly any food, with meta-analyses only partially correcting the exaggeration.
relevance_note: Empirical demonstration of the base-rate/multiple-testing problem behind any single egg-and-disease association isolated from the wider literature.