Survival analysis with functional covariates for partial follow-up studies

Area of Science:

• Biostatistics
• Longitudinal Data Analysis
• Survival Analysis

Background:

• Personalized medicine increasingly relies on predictive and prognostic analyses.
• Traditional time-dependent covariate models require continuous covariate monitoring throughout follow-up.
• Partial follow-up study designs, common in clinical settings, pose challenges for existing models.

Purpose of the Study:

• To develop a novel functional survival model for situations with partial covariate follow-up.
• To address the analysis of sparsely and irregularly measured longitudinal data.
• To identify patient subsets benefiting most from treatments using predictive biomarkers.

Main Methods:

• Proposed a new class of functional survival models.
• Combined functional principal components analysis (FPCA) with survival analysis.
• Incorporated covariate selection and handling of sparse, irregular measurements and errors.

Main Results:

• The new method effectively handles partial follow-up designs.
• Functional principal components analysis revealed patterns in covariate trajectories.
• Disease-free survival time is influenced by the patterns of absolute lymphocyte count trajectories, not just individual measurements.

Conclusions:

• The proposed functional survival model offers a viable solution for partial follow-up data.
• Biomarker trajectory patterns, rather than single measurements, are crucial for prognostic predictions.
• This approach enhances predictive modeling in personalized medicine with longitudinal data.