Dynamic Prediction of Time to Disease Progression Using Longitudinal Biomarker Data
As time goes by, more and more data are observed for each patient. Dynamic prediction is to keep making updated predictions of disease prognosis using all the available information. This proposal is motivated by the need of real-time monitoring of the disease progress of chronic myeloid leukemia patients using their BCR-ABL gene expression levels measured during their follow-up visits. We provide real-time dynamic prediction for future prognosis using a series of marginal Cox proportional hazards models over continuous time with constraints. Comparing with separate landmark analyses on different discrete time points after treatment, our approach can achieve more smooth and robust predictions. Comparing with approaches of joint modeling of longitudinal biomarkers and survival, our approach does not need to specify a model for the changes of the monitoring biomarkers, and thus avoids the need of any kind of imputing of the biomarker values on time points they are not available. This helps eliminate the potential bias introduced by mis-specified models for longitudinal biomarkers.

Bio: Xuelin Huang is Professor of Biostatistics in the University of Texas MD Anderson Cancer Center. He obtianed his Ph.D in Biostatistics from the University of Michigan in 2002, and joined MD Anderson in the same year. His research areas include survival analysis and longitudinal studies, clinical trial design and statistical methods for bioassay.