A regression survival model for testing the proportional hazards hypothesis

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• The proportional hazards (PH) model is a cornerstone of survival analysis.
• A key assumption of the PH model is that hazard functions do not cross over time.
• Violations of the PH assumption can lead to biased results in survival data analysis.

Purpose of the Study:

• To define a semi-parametric generalization of the proportional hazards regression model.
• To allow for crossing hazard functions based on covariate values.
• To propose and evaluate a global test for the proportional hazards assumption against these generalized alternatives.

Main Methods:

• Developed a semi-parametric generalization of the proportional hazards model.
• Formulated a global test for the proportional hazards assumption.
• Conducted simulation experiments to compare the power of the proposed test against existing methods in a two-sample setting.
• Utilized survival data from breast carcinoma patients for illustration.

Main Results:

• The proposed semi-parametric model accommodates crossing hazard functions, extending the standard proportional hazards model.
• The global test demonstrates power against the defined alternatives, particularly in the two-sample comparison.
• Simulation results indicate favorable performance of the new test compared to previously described tests.
• The model and test are illustrated using real-world survival data.

Conclusions:

• The generalized proportional hazards model provides a flexible framework for survival data where hazard functions may cross.
• The proposed global test is a valuable tool for assessing the proportional hazards assumption.
• The method offers improved power for detecting departures from the proportional hazards assumption in survival analysis.
• The application to breast carcinoma data highlights the practical utility of the approach.