A functional proportional hazard cure rate model for interval-censored data.

Area of Science:

• Biostatistics
• Neuroscience
• Medical Data Science

Background:

• Current survival models often use the Cox proportional hazards structure and assume right censoring.
• Predicting Alzheimer's disease (AD) conversion from sparse biomarker trajectories in mild cognitive impairment (MCI) is challenging.
• Existing methods may not adequately handle interval censoring and sparsely sampled functional data.

Purpose of the Study:

• To propose a novel functional mixture cure rate model for predicting time to Alzheimer's disease conversion in MCI patients.
• To incorporate both functional (biomarker trajectories) and scalar covariates.
• To address challenges of interval censoring and sparsely sampled functional data.

Main Methods:

• Utilized functional principal component analysis (FPCA) to extract features from sparse, irregular biomarker trajectories.
• Developed a functional mixture cure rate model.
• Applied the expectation-maximization (EM) algorithm with Poisson data augmentation for parameter estimation.
• Handled interval censoring and sparsely sampled functional data.

Main Results:

• The proposed model accurately estimates the nonparametric coefficient function, reflecting trajectory shape effects on survival and cure.
• Simulation studies confirmed the estimation accuracy of the developed method.
• The model was successfully applied to the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset.

Conclusions:

• The functional mixture cure rate model provides a robust framework for predicting Alzheimer's disease conversion from sparse longitudinal biomarker data.
• The method effectively handles interval censoring and sparsely sampled functional data, offering improved prediction accuracy.
• This approach has significant implications for understanding disease progression and identifying individuals at risk in clinical settings.