Correction of confidence intervals in excess relative risk models using Monte Carlo dosimetry systems with shared...

Area of Science:

• Epidemiology
• Biostatistics
• Survival Analysis

Background:

• Epidemiological studies often face exposure measurement uncertainties, particularly complex correlated errors.
• Existing methods for independent errors are well-developed, but correlated error correction is less explored.
• Monte Carlo dosimetry systems (MCDS) are used to model complex error structures in exposure data.

Purpose of the Study:

• To develop and evaluate a novel method for calculating corrected confidence intervals in linear excess relative risk (ERR) models.
• To address the challenges posed by correlated measurement errors in epidemiological dose-response assessments.
• To improve parameter estimation accuracy in radiation epidemiology, especially for the dose-effect parameter.

Main Methods:

• Implementation of a new method using an approximate asymptotic distribution for parameter estimates in linear ERR models.
• Application of a mixture distribution (normal and lognormal components) for the dose-effect parameter's asymptotic approximation.
• Evaluation through a simulation study using both simulated MCDS and real dose histories from the Mayak Worker Cohort.

Main Results:

• The proposed method significantly improves coverage probabilities for the dose-effect parameter.
• Noticeable improvements in coverage were also observed for other model parameters.
• The mixture distribution ensures bounded confidence intervals, unlike a purely normal approximation.

Conclusions:

• The novel method effectively corrects confidence intervals for correlated exposure measurement errors in ERR models.
• This approach enhances the reliability of parameter estimates in radiation epidemiology.
• The method offers a valuable tool for analyzing complex exposure data in epidemiological research.