A two-stage joint model approach to handle incomplete time dependent markers in survival data through inverse...

Joint models for longitudinal and survival data are essential in biomedical research, enabling the simultaneous analysis of biomarker progression and clinical events. These models account for the interdependence between longitudinal and survival outcomes, improving insights into disease progression. However, missing data in longitudinal studies pose challenges, particularly when time dependent markers contain missing values, leading to biased estimates. This paper proposes a two-stage joint modeling framework integrating multiple imputation and inverse probability weighting. First, a linear mixed-effects model estimates biomarker trajectories, handling missing data using multiple imputation. Second, predicted biomarker values are incorporated into a Cox model, where inverse probability weight corrects for selection bias in survival estimation. A detailed simulation study has been conducted to study the performance of the proposed method compared to other common approaches. Results demonstrate the framework's effectiveness in handling incomplete time dependent covariates while providing precise estimates of the relationship between biomarker progression and survival outcomes.