Joint modeling of longitudinal data and survival outcomes with multiple missing sources: a simulation study

Background Dynamic information is crucial for monitoring and predicting patients’ health status. The Time-Dependent Cox Model (TDCM) is widely used to analyze the association between biomarkers and time-to-event data. However, when dealing with longitudinal endogenous variables, such as biomarkers, the terminal event truncates the observation of their full trajectory, leading to Missing Not At Random (MNAR) data on the longitudinal variable. Joint Models (JMs) accommodate this type of MNAR data by explicitly modeling the missing process due to the terminating event. They integrate a longitudinal sub-model for the biomarker trajectory, through a mixed-effects framework, with a time-to-event sub-model for the terminating event. However, challenges remain when biomarker measurements are subject to other MNAR sources. Methods This study uses simulations to assess JMs’ robustness under different missing data mechanisms, comparing them to TDCM in estimating the association between the biomarker value and the risk of a terminal event, in terms of hazard. The JM is also applied in an Intensive Care setting with multiple missing sources. Results Simulation results indicate that JM remains robust despite truncated or intermittently missing markers, but the presence of elevated measurement error on the biomarker value increases uncertainty and can cause moderate-to-large bias. The TDCM remains effective with minimal measurement error, but joint modeling is preferable when error is abundant, especially under MNAR missingness in longitudinal data. Conclusions Correct modeling of the trajectory in JM is essential for conducting a robust analysis.