Large-scale survival analysis with a cure fraction.

Area of Science:

• Biostatistics
• Epidemiology
• Statistical Modeling

Background:

• Analyzing large-scale survival data with cure fractions presents significant regression challenges.
• Existing methods struggle with the computational demands of massive datasets and identifying risk factor impacts.

Purpose of the Study:

• To propose a novel probability-weighted method for semiparametric cure regression models.
• To develop efficient estimation and inference techniques for large-scale survival data analysis.

Main Methods:

• Developed a flexible mixture cure model combining model-free incidence and semiparametric proportional hazards latency.
• Introduced a weighted estimating equation method using susceptible probability as a weight.
• Proposed a recursive probability-weighted estimation for computational and memory efficiency in large-scale/online settings.

Main Results:

• Established asymptotic properties for the proposed estimators.
• Demonstrated robust nonparametric estimation of weights for stable regression parameter estimation.
• Achieved computational and memory efficiency suitable for massive or online data.

Conclusions:

• The proposed probability-weighted method effectively handles large-scale survival data with cure fractions.
• The method provides stable and efficient estimation of risk factor effects in population studies.
• Simulation studies and real-data application confirm the method's empirical performance.