Jointly modelling multiple transplant outcomes by a competing risk model via functional principal component analysis

Area of Science:

• Biostatistics
• Clinical Epidemiology
• Medical Data Science

Background:

• Longitudinal biomarkers like glomerular filtration rate (GFR) are crucial for monitoring disease progression and patient survival in clinical studies.
• Joint modeling of longitudinal and survival data enhances covariate effect estimation, but often overlooks detailed analysis of longitudinal trajectory variations.
• Kidney transplant outcomes involve multiple time-to-event data, such as graft failure and mortality, necessitating advanced statistical approaches.

Purpose of the Study:

• To develop an advanced joint model that integrates functional principal component analysis (FPCA) of longitudinal data with competing risks survival analysis.
• To investigate how decomposed features of longitudinal trajectories, specifically GFR, influence multiple time-to-event outcomes in kidney transplant recipients.
• To provide a more accurate and comprehensive framework for analyzing complex clinical data with both continuous and time-to-event endpoints.

Main Methods:

• Utilized functional principal component analysis (FPCA) to extract key functional features from longitudinal biomarker trajectories (e.g., GFR).
• Employed a competing risk survival model to simultaneously analyze multiple time-to-event outcomes (e.g., transplant failure, death).
• Linked longitudinal features and survival outcomes through shared functional components within a unified joint modeling framework.

Main Results:

• The application to kidney transplant data demonstrated the statistical significance of the extracted functional features in predicting survival outcomes.
• The joint model provided a more nuanced understanding of the relationship between kidney function trajectories and patient survival.
• Simulation studies confirmed the accuracy and reliability of the proposed estimation method.

Conclusions:

• The developed joint model effectively leverages FPCA to characterize longitudinal data, offering significant improvements in predicting multiple survival outcomes.
• The identified functional features are critical for understanding disease progression and patient prognosis in contexts like kidney transplantation.
• This approach enhances the analysis of complex biomedical data, paving the way for more personalized patient management and treatment strategies.