John Ioannidis critiques nutritional epidemiology for generating highly confounded, often biologically implausible associations that are frequently misinterpreted as causal. He highlights meta-analyses in which nearly every food appears to meaningfully raise or lower mortality risk—sometimes to absurd degrees—showing how residual confounding, selective reporting, and correlated dietary patterns undermine the credibility of observational findings. Ioannidis stresses that even biomarker-based studies and high-profile trials have struggled to produce reliable, reproducible results.

He calls for sweeping reform: full data transparency, standardized analyses, independent re-evaluation of large cohorts, and far greater caution in translating observational findings into public policy. He warns that the current model confuses the public, distorts dietary guidelines, and diverts attention from more clear-cut health threats. Ultimately, Ioannidis argues that nutritional epidemiology should rebuild itself on more rigorous methods, better-designed trials, and openly shared data to regain scientific credibility.

https://statmodeling.stat.columbia.edu/wp-content/uploads/2018/08/jama_Ioannidis_2018_vp_180095.pdf