Semiparametric Regression Estimation for Recurrent Event Data with Errors in Covariates under Informative Censoring

Area of Science:

• Biostatistics
• Epidemiology
• Longitudinal Data Analysis

Background:

• Recurrent event data are common in longitudinal studies, necessitating analysis of covariate effects on event rates.
• Existing methods often assume independent censoring and ignore covariate measurement errors, limiting their applicability.
• Informative drop-out, where censoring is related to the event process, further complicates analysis.

Purpose of the Study:

• To develop robust semiparametric regression methods for recurrent event data.
• To address challenges posed by informative drop-out and covariate measurement errors simultaneously.
• To provide accurate estimation of covariate effects on the occurrence rate function.

Main Methods:

• Utilized semiparametric regression models for recurrent event rates.
• Incorporated an unspecified frailty distribution to model recurrent events.
• Employed a classical measurement error model for covariates.
• Proposed two corrected statistical approaches to handle informative censoring and measurement error.

Main Results:

• Developed two statistically identical corrected methods for parameter estimation.
• Established asymptotic properties of the proposed estimators.
• Demonstrated the finite sample performance through simulation studies.
• Applied the methods to the Nutritional Prevention of Cancer trial data.

Conclusions:

• The proposed methods effectively handle informative drop-out and covariate measurement errors in recurrent event data analysis.
• Accurate estimation of covariate effects is crucial for understanding disease recurrence and treatment efficacy.
• The methods offer a valuable tool for analyzing complex longitudinal data in clinical and epidemiological research.