|
Abstract:
|
In clinical trials, patients respond either favorably or unfavorably to the same treatment because of patient heterogeneity and chance variation. In this work, we focus on one type of heterogeneity, specifically the existence of distinct subgroups of patients who respond favorably only to a subset of study treatments. We propose using a mixture of mixed logistic regression models to estimate these subgroup proportions and the probabilities of a favorable response. An algorithm for parameter estimation as well as consideration of the identifiability conditions are provided. With this information, an optimal dynamic treatment regime can be determined. Simulation studies are performed to demonstrate the effectiveness of the proposed method and determine adequate sample size. We further apply the proposed model and method to a prostate cancer trial data and compare different dynamic treatment regimes.
|
