Marginal Treatment Effect Estimation Using Pattern Mixture Model

• Activity Number: 308
• Type: Contributed
• Date/Time: Tuesday, August 6, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Zhenzhen Xu*+
• Companies: FDA
• Keywords: Pattern Mixture Model ; MNAR ; Binary Responses ; Missing Data ; Longitudinal Clinical Trial

Abstract

Missing data often occur in clinical trials where the primary objective is to estimate the overall treatment effect. When the missingness depends on unobserved responses, pattern mixture model is frequently used. This model stratifies the data according to dropout patterns and estimates the pattern-specific parameters. To obtain the marginal treatment effect estimator, one can calculate a weighted average of pattern-specific treatment effects, assuming that the treatment assignment is equally distributed across missing patterns. However, in practice, the treatment assignment may vary among patterns and consequently the existing marginal estimator based on a weighed average is subject to bias.

We propose an approach to estimate the marginal treatment effect in longitudinal studies with continuous responses, relaxing the assumption on homogeneous treatment assignment distribution across missing patterns. A simulation study is conducted to evaluate the performance of the proposed estimator and compare it with other approaches under different missing mechanisms. We further consider a generalization of the proposed method to clinical trials with repeated binary responses.