Covariate hypothesis tests for the cure rate in mixture cure models based on martingale difference correlation

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Cure models analyze time-to-event data where some individuals never experience the event.
• These models estimate cure rates and covariate effects, but testing these effects is challenging.
• Existing methods for testing covariate influence on cure rates are limited.

Purpose of the Study:

• To propose novel nonparametric hypothesis tests for assessing covariate effects on cure probability.
• To extend these methods for evaluating the impact of multiple covariates.
• To provide tools for better understanding cure rates in medical research.

Main Methods:

• Utilized martingale difference correlation for nonparametric hypothesis testing.
• Employed permutation and chi-square tests to approximate null distributions.
• Extended methodology to include partial martingale difference correlation for multiple covariates.

Main Results:

• The proposed tests effectively evaluate the influence of covariates on cure probability.
• Simulation studies demonstrated the performance of the tests under various conditions.
• The methodology was successfully applied to a real-world dataset of rheumatoid arthritis patients.

Conclusions:

• The developed nonparametric tests offer a valuable addition to the analysis of cure models.
• These methods enhance the ability to test covariate effects on cure rates.
• The approach provides insights into factors influencing long-term outcomes in patient populations.