|
Activity Number:
|
529
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 5, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #304728 |
|
Title:
|
Estimation of Treatment Efficacy in the Presence of Noncompliance and Competing Risks in Randomized Controlled Trials
|
|
Author(s):
|
Lily Altstein*+
|
|
Companies:
|
University of California, Los Angeles
|
|
Address:
|
, Los Angeles, CA, 90024,
|
|
Keywords:
|
subgroup analysis ; censored data ; EM algorithm ; causal inference ; accelerated failure time models ; mixture model
|
|
Abstract:
|
Unobserved subgroups arise in clinical trials when we are unable to assess membership to particular cohorts in which we wish to estimate a treatment effect. This occurs, for example, when patients fail to comply with the randomized treatment assignment or when a diagnostic test or biomarker is available only on a subset of patients. We develop a fully parametric mixture model that estimates biological treatment efficacy and other covariate effects in a subgroup of patients from a two-arm randomized trial when membership is unobserved in the control arm. We extend this methodology to accommodate competing risks, allowing estimation of biological efficacy and other effects separately for each failure type. The EM-algorithm facilitates parameter estimation in the mixture model.
|
