Testing for Cluster-Level Random Effects in Joint Modeling of Survival Time and Marker Responses in Clinical Trials

• Activity Number: 195
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 10:30 AM to 12:20 PM
• Sponsor: SSC
• Author(s): Wenyu Jiang* and Xin Yao and Bingshu Chen
• Companies: Queen's University and Queen's University and Queens University
• Keywords: binary outcomes ; bootstrap ; clustered survival data ; joint modeling ; random effects ; score test

Abstract

A modern clinical trial is often designed to include multiple institutions, to ensure an adequate number of participants in the study and that the inference applies to a broad patient population. Failure to account for heterogeneity among clinical centers may lead to bias in point estimation and underestimation of variability. For a study with a single response variable, score test is often used to test the existence of clustered effects. In this paper, we extend the score test method to joint modeling of survival time and binary marker responses. We propose methods for testing whether the variance of the cluster-level random effects is greater than zero, by deriving the test statistic and approximating its distribution through bootstrap and asymptotic methods. Simulation study shows that the bootstrap method has higher power compared to the asymptotic method. We then apply both methods to a Hodgkin Lymphoma trial and find that neither the survival time outcome nor the marker response shows any significant cluster-level random effects.