Constrained parametric model for simultaneous inference of two cumulative incidence functions

Area of Science:

• Biostatistics
• Survival Analysis
• Epidemiology

Background:

• Competing risks data analysis is crucial in medical research.
• Existing regression models may overlook the constraint that cumulative incidence functions (CIFs) must sum to one.
• Accurate modeling of covariate effects is essential for understanding disease progression and treatment efficacy.

Purpose of the Study:

• To propose a novel parametric regression model for cumulative incidence functions (CIFs) in competing risks data.
• To ensure the model explicitly accounts for the constraint that the sum of CIFs equals one.
• To provide a robust method for analyzing treatment effects and prognostic factors in the presence of competing events.

Main Methods:

• A parametric regression model using a modified logistic function for baseline CIFs.
• A generalized odds-rate model to incorporate covariate effects.
• Modeling competing cause effects based on primary cause effects and baseline distribution asymptotes.
• Inference via standard maximum likelihood theory.

Main Results:

• The proposed model satisfies the additivity constraint for CIFs.
• Simulation studies demonstrate favorable finite-sample performance compared to existing methods.
• The model effectively analyzes breast cancer recurrence data, assessing tamoxifen's treatment effect.

Conclusions:

• The developed parametric regression model offers a reliable approach for analyzing competing risks data.
• It accurately captures covariate effects while respecting the fundamental CIF additivity constraint.
• The model has practical utility in clinical research, exemplified by its application to breast cancer data analysis.