Placebo Adherence as a Negative Control Exposure.

Abstract

In placebo-controlled randomized clinical trials, adherence to the placebo is often supposed to have no effect on the primary outcome of interest: when unbiased methods are used, investigators expect to estimate a null effect. Estimating the 'effect' of adherence to placebo in these settings has thus been proposed and popularized as a strategy for detecting bias, for example, from unmeasured confounding. This strategy can be viewed as an application of a type of negative control exposure, which we term a placebo-adherence negative control exposure. Here, we formally state its defining assumptions and discuss the unique advantages of single-world intervention graphs for reasoning about them.