A toolkit for measurement error correction, with a focus on nutritional epidemiology

Area of Science:

• Epidemiology
• Biostatistics
• Nutritional Epidemiology

Background:

• Exposure measurement error biases epidemiological study results.
• Validation samples are ideal but often infeasible.
• Repeated exposure measurements offer an alternative data source.

Purpose of the Study:

• To provide a toolkit for measurement error correction using repeated measurements.
• To address classical and non-classical error types (systematic, heteroscedastic, differential).
• To facilitate practical application in various fields, including nutritional epidemiology.

Main Methods:

• Regression calibration for classical error.
• Moment reconstruction and multiple imputation for differential error.
• Adaptations for continuous and categorized exposures.

Main Results:

• The toolkit integrates multiple methods for robust measurement error correction.
• Demonstrates practical application using fibre intake and colorectal cancer data.
• Provides methods applicable to both continuous and categorized exposure variables.

Conclusions:

• Repeated measurements are valuable for correcting exposure measurement error.
• The presented toolkit enhances the accuracy of exposure-disease association estimates.
• Methods are applicable across diverse epidemiological research, particularly in nutritional studies.