A quantile cure model with partially functional covariate effects.

Area of Science:

• Biostatistics
• Survival Analysis
• Econometrics

Background:

• Quantile regression is valuable for analyzing continuous outcomes, handling heteroscedasticity, and modeling covariate effects across quantiles.
• Existing methods for mixture cure models often rely on restrictive assumptions, limiting their application and interpretability.
• Previous analyses of lung cancer data were constrained by the limitations of traditional survival models.

Purpose of the Study:

• To propose novel estimating equation approaches for mixture cure models using quantile regression for latency time.
• To develop methods that offer double robustness and relax restrictive global log-linear assumptions.
• To provide a more flexible and robust framework for survival data analysis.

Main Methods:

• Developed estimating equation approaches for a mixture cure model incorporating quantile regression for latency.
• Ensured double robustness, where misspecification in one component does not impact the other.
• Relaxed global log-linear assumptions, allowing for quantile-varying and invariant effects.

Main Results:

• Established asymptotic properties of the proposed estimators.
• Simulation studies confirmed double robustness and efficiency gains compared to existing methods.
• Application to lung cancer data identified significant survival differences between uncured patient groups.

Conclusions:

• The proposed quantile regression-based mixture cure model offers a robust and flexible alternative for survival data analysis.
• The method provides valuable insights, as demonstrated by its application to lung cancer data, revealing previously unreported survival differences.
• This approach enhances the ability to model complex survival data, particularly in the presence of cure fractions and heteroscedasticity.