factual

Nutritional Epidemiology Cannot Infer Causation

The ability to infer cause from nutritional epidemiology is 'somewhere between zero and epsilon' because its hazard ratios are tiny (1.1-1.12), inconsistent (results reverse week to week), show no dose effect or biologic plausibility, and cannot be reproduced in randomized controlled trials—though epidemiology has had wins where the signal is strong (cholera, smoking).

factualpending

Speaker

Peter Attia

Evidence Quote

“with nutritional epidemiology, there are a lot of problems statistically. The hazard ratios are tiny. They're like 1.1, 1.12. They never go in the same direction twice. One day red meat causes cancer, the next week it doesn't... There's no dose effect... There's no biologic plausibility... you can't reproduce these results in randomized control trials.”

Source

Peter Attia on Lifespan, Healthspan, and Outlive— EconTalk
Created: 6/13/2026, 12:18:42 AM

Your notes

Loading notes...