Abstract
Background. The Food and Drug Administration recently added a new clinical endpoint for evaluating the efficacy of alcohol use disorder (AUD) treatment that is more inclusive of treatment goals besides abstinence: no heavy drinking days (NHDD). However, numerous critiques have been noted for such binary models of treatment outcome. Further, there is mounting evidence that participants inaccurately estimate the quantities of alcohol they consume during drinking episodes (i.e., drink size misestimation), which may be particularly problematic when using a binary criterion (NHDD) compared to a similar, continuous alternative outcome variable: percent heavy drinking days (PHDD). Yet, the impact of drinking misestimation on binary (e.g., NHDD) versus continuous outcome variables (e.g., PHDD) has not been studied. Objectives. Using simulation methods, the present study examined the potential impact of drink size misestimation on NHDD and PHDD. Methods. Data simulations were based on previously published findings of the amount of error in how much alcohol is actually poured when estimating standard drinks. We started with self-reported daily drinking data from COMBINE study participants with complete data (N = 888; 68.1% male), then simulated inaccuracy in those estimations based on literature on standard drink size misestimation. Results. Clinical trial effect sizes were consistently lower for NHDD than for PHDD. Drink size misestimation further lowered effect sizes for NHDD and PHDD. Conclusions. Drink size misestimation may lead to inaccurate conclusions about drinking outcomes and the comparative effectiveness of AUD treatments, including inflated type-II error rates, particularly when treatment "success" is defined by binary outcomes such as NHDD.