Application of the Pattern Mixture Model to Mini Mental Status Exam (MMSE) Scores in a Radiation Therapy Oncology Group (RTOG) Trial for Brain Tumors

• Activity Number: 627
• Type: Contributed
• Date/Time: Thursday, August 5, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Kyounghwa Bae*+ and Seunghee Baek and James Dignam
• Companies: Radiation Therapy Oncology Group and University of Pennsylvania and Radiation Therapy Oncology Group
• Keywords: A pattern-mixture model ; Overall survival ; repeated measurment ; Bayesian inference ; Mini Mental Status Exam

Abstract

Joint modeling of longitudinal assessments and survival is necessary to comprehensively evaluate the effects of aggressive cancer treatment. A pattern-mixture model (Li and Heitjan,2006) was used to analyze RTOG 9402, a phase III trial comparing radiation therapy (RT)+chemotherapy (PCV) to RT alone for brain tumors. The model assumes a multinomial distribution for the overall survival (OS) and a multivariate normal distribution for the MMSE conditional on OS, and uses Bayesian methods to estimate parameters and impute missing information. In an analysis of MMSE scores using observed data only and ignoring OS, mean MMSE increased for RT+PCV and decreased for RT . In contrast, the joint model showed increasing MMSE in both arms after 2 years, favoring RT+PCV. Additional findings and guidance for the use of pattern mixture models in oncology studies will be presented.