A semiparametric marginal mixture cure model for clustered survival data

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Clustered failure time data with a cure fraction presents analytical challenges in regression.
• Existing marginal methods may lack efficiency when analyzing such complex data structures.
• Accurate modeling is crucial for understanding treatment effects in diseases with potential cures.

Purpose of the Study:

• To develop and evaluate a novel marginal model for regression analysis of clustered failure time data with a cure fraction.
• To estimate regression parameters within a semiparametric proportional hazards mixture cure model using generalized estimating equations.
• To assess the efficiency of the proposed method compared to existing marginal approaches.

Main Methods:

• Utilized novel generalized estimating equations within an expectation-maximization algorithm.
• Modeled intra-cluster dependence in cure statuses and survival times using working correlation matrices.
• Employed a bootstrap method for variance estimation of regression parameters.

Main Results:

• The proposed method demonstrated substantial efficiency gains over existing marginal methods in simulation studies.
• The model effectively handles dependence structures within clusters for both cure status and survival times.
• Successful application to a real-world dataset from a multi-institutional study of tonsil cancer patients.

Conclusions:

• The novel generalized estimating equations approach provides an efficient and robust method for analyzing clustered failure time data with a cure fraction.
• This method offers improved statistical power for identifying significant predictors in complex survival data.
• The findings have implications for analyzing similar data in clinical research, particularly in oncology.