Testing for change-point in the covariate effects based on the Cox regression model

Area of Science:

• Biostatistics
• Survival Analysis
• Cancer Research

Background:

• Change-point models are vital in analyzing time-to-event data, particularly in cancer research.
• Cox models with change-points allow flexible specification of covariate effects over time.

Purpose of the Study:

• To propose and evaluate statistical tests for detecting change-point effects in covariate patterns within Cox models.
• To assess the finite sample performance of these tests under various scenarios.

Main Methods:

• Development of maximal score, maximal normalized score, and maximal Wald tests.
• Establishment of asymptotic properties for the test statistics.
• Utilizing Monte Carlo simulations to determine critical values and assess performance.

Main Results:

• The maximal Wald test showed generally satisfactory performance, even with moderate sample sizes.
• The maximal score test's performance could be sensitive to the true change-point location.
• All proposed methods offered natural estimates for change-point locations.

Conclusions:

• The proposed statistical tests are effective for detecting change-point effects in Cox models.
• The maximal Wald test is recommended for its reliable performance in survival data analysis.
• The methods were successfully applied to real-world medical datasets, including primary biliary cirrhosis and breast cancer.