A Cautionary Note on Extended Kaplan-Meier Curves for Time-varying Covariates

Area of Science:

• Biostatistics
• Epidemiology
• Survival Analysis

Background:

• The Kaplan-Meier curve is a standard tool for survival analysis in cohort studies with time-fixed covariates.
• An extended Kaplan-Meier curve has been proposed for time-varying covariates, allowing subjects to transition between risk sets.

Purpose of the Study:

• To investigate the causal interpretation of the extended Kaplan-Meier curve for time-varying covariates.
• To evaluate the validity of the independence assumption required for causal inference.

Main Methods:

• Theoretical analysis of the extended Kaplan-Meier curve under specific assumptions.
• Examination of the implications of the independence assumption for causal interpretation.

Main Results:

• The extended Kaplan-Meier curve does possess a causal interpretation under the stated independence assumption, in the absence of confounding.
• The independence assumption, which posits that covariate changes are unrelated to future risk, is often unrealistic in practice.

Conclusions:

• While mathematically sound under specific conditions, the causal interpretation of the extended Kaplan-Meier curve is typically unwarranted due to the unrealistic nature of its underlying assumptions.
• Researchers should exercise caution when interpreting extended Kaplan-Meier curves as causal effects in the presence of time-varying covariates.