摘要： 

Biomedical studies often collect longitudinal data on both clinical endpoints (i.e., the onset of chronic diseases) and a variety of disease markers, the latter of which may provide crucial information on the antecedent progression prior to the clinical onset of many disease processes. Such information has the potential to better identify individuals at higher risk of developing the diseases. We propose a joint model of longitudinal disease marker data and time to clinical onset data which allows a possible antecedent acceleration on the rate of changes for disease markers prior to the onset. We rely on the standard general linear mixed models for longitudinal data and the standard Cox proportional hazards model for time to event data and link them with a random changepoint model on the rate of change for disease markers. We provide estimates to the regression parameters in these models as well as to the parameters associated with the changepoints through an EM algorithm. The proposed model is demonstrated by using a real world study of Alzheimer’s disease (AD) that seeks to understand the antecedent cognitive changes prior to the onset of AD and their implications on the risk of developing the disease.