Partly linear single-index cure models with a nonparametric incidence link function

Area of Science:

• Statistics
• Biostatistics
• Survival Analysis

Background:

• Cancer studies often involve cured patients, complicating survival analysis.
• Standard mixture cure models have limitations in modeling covariate effects on cure and latency.
• Covariates can influence both cancer recurrence (incidence) and time to event (latency).

Purpose of the Study:

• To develop flexible semiparametric mixture cure models for cancer survival data.
• To overcome limitations of traditional models regarding covariate effect structures.
• To provide a more accurate framework for analyzing cancer patient outcomes.

Main Methods:

• Proposed a class of semiparametric mixture cure models with single-index functions.
• Employed a hybrid nonparametric maximum likelihood estimation (NPMLE) approach.
• Utilized Bernstein polynomials for estimating regression components and an expectation-maximization algorithm for parameter estimation.

Main Results:

• The proposed method offers enhanced flexibility in modeling covariate effects.
• Simulation studies demonstrated the good finite-sample performance of the estimator.
• The methodology was successfully applied to real-world cancer datasets.

Conclusions:

• The developed semiparametric mixture cure models provide a flexible and effective tool for cancer survival analysis.
• The proposed estimation method is robust and performs well in practice.
• This approach can lead to improved understanding and prediction of cancer patient outcomes.