Smoke, Diet, and the Confounding Problem: What's Actually Doing the Damage?

Health research keeps measuring kinds of people and reporting kinds of exposures. This paper examines the confounding problem on both sides of the ledger: how much of the disease burden attributed to smoking is the smoke itself, versus the cluster of co-occurring habits — poorer diet, heavier drinking, less exercise, lower socioeconomic status — that travel with smoking in observational studies; and symmetrically, how much of the benefit attributed to healthy diets is the food, versus the healthy-user bias of the people who eat that way. Drawing on multi-covariate cohort studies, smoking-discordant twin pairs, Mendelian randomization, direct mechanistic evidence, and the catalog of reversed findings (beta-carotene, PREDIMED, Wansink, hormone replacement therapy), it concludes that smoking's effect is real and primarily the smoke — the largest single behavioral factor in the cohorts reviewed — while diet's benefits are real but inflated by bias. It closes by identifying the missing study: a direct comparison of healthy-lifestyle smokers against unhealthy-lifestyle non-smokers, which the literature has never run.