Reverse causation

Reverse causation is when the outcome under study causes the exposure, not the other way around. Observational cohorts record what people do and what happens to them later. They cannot see why a habit started. A concrete case: someone receives an early diabetes diagnosis, or is told that fasting glucose is drifting upward, and switches from sugared soft drinks to diet versions. Years later the cohort reports that sweetener use predicts diabetes. The causal arrow runs backwards. The disease process came first and drove the exposure. Statistical adjustment helps less than people expect. Adjusting for body mass index or physical activity captures a person's state on the day of the survey. It does not capture the reason behind the switch, and that reason is what carries the bias. This is not the same as confounding, where a third factor drives exposure and outcome independently. Here the causal path itself is inverted. The usual defences are excluding the first years of follow-up in a lag analysis, restricting to people healthy at baseline, checking whether an association strengthens close to diagnosis, and Mendelian randomization, which uses genetic variants fixed at conception. In longevity and nutrition research it is a leading explanation for associations that trials later fail to reproduce.

Last reviewed: August 25, 2026

This definition is educational and is not medical advice, a diagnosis, or treatment. Talk to a doctor about any health decisions. Read our full medical disclaimer