A0254
Title: Time-varying proportional odds models for mega-analysis of clustered event times
Authors: Tanya Garcia - Texas A\&M University (United States) [presenting]
Karen Marder - Columbia University (United States)
Yuanjia Wang - Department of Biostatistics (United States)
Abstract: Mega-analysis, or the meta-analysis of individual patient data, is the gold standard for synthesizing data from multiple studies when individual level data are available. It borrows information across studies to attain reliable and more precise estimation and to analyze multiple outcomes. An important aspect in the mega-analysis of time-to-event data is estimating the distribution function while accounting for the data hierarchy and potential right-censoring. An often encountered challenge is when the time-to-event data are clustered as in biomedical studies where multiple, life-impacting events on each subject are often measured (e.g., first loss of cognitive ability, first lost of motor-function). The multiple events measured on the same subject lead to clusters of correlated event times, and the intraclass dependency needs to be properly modeled to ensure correct inference on regression parameters.
