Understanding Landmarking and Its Relation with Time-Dependent Cox Regression.

Area of Science:

• Biostatistics
• Survival Analysis
• Epidemiology

Background:

• Time-to-event data analysis frequently involves time-dependent covariates.
• Time-dependent Cox regression and landmark analysis are primary methods for handling these covariates.
• Landmarking estimates covariate effects based on their value at a specific time point.

Purpose of the Study:

• To derive mathematical expressions relating the regression coefficient of a time-dependent covariate in landmark analysis to time-dependent Cox regression.
• To provide a theoretical link between these two common survival analysis techniques.

Main Methods:

• Derivation of regression coefficient expressions for landmark analysis.
• Utilizing the framework of time-dependent Cox regression.
• Validation through simulation studies and real-world data analysis (Stanford heart transplant data).

Main Results:

• Formulas were derived for the time-varying regression coefficient in landmark analysis.
• These formulas express the landmark coefficient in terms of the time-dependent Cox regression coefficient and baseline hazard.
• The derived relationships were confirmed via simulations and empirical data.

Conclusions:

• The study establishes a clear mathematical connection between time-dependent Cox regression and landmark analysis.
• This provides a deeper understanding of how landmarking estimates relate to the full time-dependent Cox model.
• The findings facilitate more accurate interpretation and application of landmark analyses in survival data.