OBJECTIVES: Joint modeling provides a powerful framework for integrating longitudinal and survival data to assess the prognostic influence of time-dependent biomarkers on survival outcomes. However, applications to rare diseases, where small sample sizes often represent a major challenge, are currently scarce. This study aims to validate previous findings of Rizopolous (2011) in the context of rare diseases using simulated data, and to explore the performance of joint survival models with varying sample size.

METHODS: Simulated datasets were generated based on data for non-Cushing’s Disease Cushing’s Syndrome, including key baseline characteristics, cortisol levels as longitudinal biomarkers, and disease progression outcomes. Joint modeling was applied following the methodology outlined by Rizopoulos (2011). To examine the impact of sample size on model performance, joint models were also fitted to subsamples of the simulated dataset. Model performance was evaluated based on predictive accuracy, bias, and concordance, as well as uncertainty in key model parameters, e.g., association parameters for longitudinal biomarkers and hazard ratios for baseline covariates.

RESULTS: The analyses of the complete simulated dataset demonstrated high predictive accuracy of joint models and led to robust conclusions on the relationship between the longitudinal biomarkers and survival outcomes. However, analysis of credible intervals revealed that reduced sample sizes substantially increased uncertainty, indicating potential limitations of standard formulations of joint modeling for studies with very small cohorts.

CONCLUSIONS: This study supports the applicability of joint modeling in rare disease, but also indicates that small sample sizes could be a limit practical usage of the method in some rare diseases. In scenarios with especially small patient populations, leveraging external data through informed Bayesian frameworks is appealing and could potentially recover tractable models. Future exploration of such methodological enhancements could further improve the utility of this promising method in rare diseases, ultimately supporting better clinical and policy decision-making.