Inference in a survival cure model with mismeasured covariates using a simulation-extrapolation approach

Area of Science:

• Biostatistics
• Survival Analysis
• Epidemiology

Background:

• Survival analysis often encounters situations where a fraction of subjects are 'cured' and will never experience the event of interest.
• Promotion time cure models account for this cure fraction.
• Explanatory variables in these models can be subject to measurement error, leading to biased estimators.

Purpose of the Study:

• To extend the simulation-extrapolation (SIMEX) algorithm to the promotion time cure model.
• To address and reduce bias caused by measurement error in explanatory variables within cure models.
• To provide a statistically sound method for analyzing data with both cure fractions and measurement error.

Main Methods:

• The study adapts the simulation-extrapolation (SIMEX) algorithm for use with promotion time cure models.
• The SIMEX approach involves simulations to estimate and correct for the bias introduced by measurement error.
• Theoretical properties of the proposed estimator, including consistency and asymptotic normality, are investigated.

Main Results:

• The proposed estimator, extended from the SIMEX algorithm, is shown to be approximately consistent and asymptotically normally distributed.
• The method demonstrates good performance in finite sample simulations.
• The analysis of a cardiology database, including the ejection fraction (a variable known to have measurement error), illustrates the practical application.

Conclusions:

• The extended SIMEX algorithm effectively handles measurement error in explanatory variables within promotion time cure models.
• The developed method provides reliable and less biased estimates in the presence of measurement error.
• This approach offers a valuable tool for survival data analysis in fields like cardiology where measurement errors are common.