Mean residual life regression with functional principal component analysis on longitudinal data for dynamic...

Area of Science:

• Biostatistics
• Medical Informatics
• Oncology

Background:

• Accurate patient life expectancy prediction is crucial for clinical treatment decisions.
• Dynamic prediction requires modeling longitudinal, irregularly measured biomarker data.
• Patient biomarker trajectories exhibit significant heterogeneity over time.

Purpose of the Study:

• To develop a flexible and powerful approach for dynamic life expectancy prediction.
• To incorporate longitudinal biomarker information into survival prediction models.
• To improve the accuracy of remaining life expectancy predictions for patients undergoing treatment.

Main Methods:

• Utilized functional principal component analysis (FPCA) to extract dominant features from individual biomarker trajectories.
• Employed these features as time-dependent covariates in a transformed mean residual life (MRL) regression model.
• Validated the model's performance through simulation studies and application to chronic myeloid leukemia (CML) patient data.

Main Results:

• The transformed MRL model incorporating longitudinal biomarker information demonstrated improved prediction performance.
• FPCA effectively captured individual patient biomarker trajectory patterns.
• The method successfully predicted remaining life expectancy until disease progression in CML patients using BCR-ABL transcript levels.

Conclusions:

• The proposed dynamic prediction model offers a robust method for integrating longitudinal biomarker data into life expectancy estimations.
• This approach enhances clinical decision-making by providing updated predictions based on real-time patient data.
• The method shows promise for application in various diseases requiring dynamic prognostic modeling.