A1231
Title: Exploring the nature of individualized treatment effects using a large crossover trial
Authors: Bryan Blette - Vanderbilt University Medical Center (United States) [presenting]
Abstract: Methods for assessing individualized treatment effects (namely, conditional average treatment effects estimated for each individual in a study sample) have become increasingly popular for characterizing treatment effect heterogeneity in clinical trials. The interpretation of these effects is non-standard, and they are frequently conflated with more narrowly defined individual treatment effects. While individual treatment effects are never known, within-individual differences in a gold-standard crossover trial can provide a close analogue to individual effects. Individualized treatment effects are estimated using a suite of machine learning methods and data from the first stage of a relatively large crossover trial studying the effect of high- vs low-sodium diet on blood pressure. The second stage of the crossover trial is then revealed, and these estimates are compared to the within-individual differences observed in the trial. Through this illustration and under certain assumptions, subtle differences are highlighted between individualized and individual treatment effects, quantifying the extent to which the estimated individualized treatment effects would or would not accurately characterize heterogeneity of effect of the study treatment.
