Analysis of Correlated Data and Time-to-Event Using the Markov Chain Imbedding Approach

• Activity Number: 610
• Type: Topic Contributed
• Date/Time: Thursday, August 5, 2010 : 8:30 AM to 10:20 AM
• Sponsor: WNAR
• Author(s): Wendy Lou*+
• Companies: University of Toronto
• Keywords: Repeated measures ; Time-to-event ; Finite Markov chain imbedding ; Transition probabilities

Abstract

In many follow-up studies, outcomes of interest include changes within clusters over time, such as individual rate of health decline or time to a pre-specified clinical threshold. Joint modeling of two such outcomes together, including two-stage modeling, has received much recent attention in the literature. In this talk, an approach based on the finite Markov chain imbedding technique will be introduced. The approach provides an intuitive formulation for joint modeling, and requires few assumptions. The methodology relies on the construction of transition probability matrices, which can be flexibly defined over time. An overview of the proposed method will be presented, and detailed demonstrations of the approach will be given via two examples from a clinical cancer trial and a cystic fibrosis study. Some comparisons to other methods and possible extensions will be discussed.