Outcome-wide Epidemiology

Abstract

The author proposes that epidemiologic studies should more often assess the associations of a single exposure with multiple outcomes simultaneously. Such "outcome-wide epidemiology" will be especially important for exposures that may be beneficial for some outcomes but harmful for others. Outcome-wide epidemiology may also be helpful in prioritizing public health recommendations. Methodologically, the conduct of outcome-wide epidemiology will generally be more straightforward than recent proposals for exposure-wide epidemiologic studies, in which the associations between a single outcome and many exposures are assessed simultaneously. Such exposure-wide studies are likely to be subject to numerous biases because of the inability to make simultaneous confounding control and because exposures are likely to affect, and mediate the effects of, other exposures. These problems simplify considerably in an outcome-wide approach when a single exposure is being considered. Moreover, outcome-wide approaches will generally be more useful than exposure-wide approaches in shaping public health recommendations.

FIGURE.

Causal diagram illustrating the confounding challenges that can arise in trying to assess the effects of multiple exposures (E₁–E₅) on a single outcome (Y) in the presence of measured (C) and unmeasured (U) confounding variables.