ABSTRACT
In longitudinal studies in the biomedical and health sciences, disease progress is often monitored by a marker over time, e.g., CD4 count in AIDS studies, hemoglobin level in end-stage renal disease (ESRD) patients. There could exist an event of interest, e.g., death or diagnosis of a specific disease. Patients may also drop out of the study due to side effects, ineffective treatment, or poor health status (too sick to continue the study). These events may be correlated with the longitudinal markers. The relation between the longitudinal data and time to event data is often of interest. For example, if the survival outcome takes a long time to occur, we might be interested in using a biomarker as a surrogate to the survival endpoint. On the other hand, if our primary interest is the longitudinal outcome, we would like to account for the possible informative dropout to reduce potential bias.
