Shujie Ma

Main Content

Exploration of heterogeneous treatment effects via concave fusion
16 February 2017 from 4:00 PM to 5:00 PM
201 Thomas Building
Contact Name
Lorey Burghard
Add event to calendar

Understanding treatment heterogeneity (HTE) is essential to the development of precision medicine, which seeks to tailor medical treatments to individual patients. Traditionally, possible HTE is assessed in subgroup analysis which determines whether individuals respond differently to a treatment based on one or more measured characteristics. For example, regression analysis can be performed by relating the outcome to treatment and a collection of baseline covariates. Such a regression model can incorporate HTE as interactions between treatment and baseline covariates. In practice, the collection of observed baseline covariates is often limited, and thus may be insufficient for characterizing the true HTE across individual patients. Some important moderators can be unknown or unobserved.

In this talk, I will introduce a general latent class model to explore the true HTE and a machine learning method to estimate the model. Our approach provides a way to identify subgroups without having a priori knowledge of the grouping information of patients with respect to treatment. Specifically, in our model we assume that the coefficients for treatment variables are subject-dependent and belong to different subgroups with unknown grouping information. We develop a concave fusion penalized  method for automatically estimating the grouping structure and the subgroup-specific treatment effects, and derive an alternating direction method of multipliers algorithm for its implementation. We also derive the theoretical properties for statistical inference about the subgroup-specific treatment effects. The performance of the proposed method is evaluated by simulation studies and illustrated by analyzing the data from the AIDS Clinical Trials Group Study.



Filed under: ,

Navigation for this Section