Dynamic predictions and prospective accuracy in joint models for longitudinal and time-to-event data

Area of Science:

• Biostatistics
• Longitudinal Data Analysis
• Survival Analysis

Background:

• Longitudinal studies frequently examine associations between repeatedly measured markers and time-to-event outcomes.
• Joint modeling of longitudinal and time-to-event data is a growing biostatistical research area.
• Assessing the predictive power of longitudinal markers for event outcomes is crucial in clinical research.

Purpose of the Study:

• To present a joint modeling framework for longitudinal and time-to-event data.
• To focus on assessing the predictive ability of longitudinal markers for time-to-event outcomes.
• To illustrate survival probability estimation and accuracy measures for future subjects.

Main Methods:

• Utilizing a joint modeling framework to integrate longitudinal marker data with time-to-event data.
• Deriving accuracy measures to evaluate the predictive performance of the longitudinal marker.
• Estimating survival probabilities for future individuals based on fitted joint models and available longitudinal measurements.

Main Results:

• Demonstrated how to estimate future survival probabilities using fitted joint models and longitudinal data.
• Developed and applied accuracy measures to assess the marker's discriminative ability.
• Illustrated the methodology with a real-world dataset of HIV patients predicting time-to-death using CD4 cell counts.

Conclusions:

• Joint modeling provides a robust framework for predicting time-to-event outcomes using longitudinal markers.
• The proposed accuracy measures effectively evaluate the predictive and discriminative capabilities of longitudinal markers.
• The application to HIV data highlights the potential of this approach for clinical risk prediction.