Improving transportability of regression calibration under the main/external validation study design

Area of Science:

• Epidemiology
• Biostatistics

Background:

• Measurement error (ME) is a common challenge in epidemiological studies.
• Regression calibration using validation studies is a standard method to correct for ME.
• External validation studies (EVS) can introduce bias if their parameters are not transportable to the main study (MS).

Purpose of the Study:

• To improve the regression calibration method for linear regression models when using an external validation study.
• To develop a method that ensures the transportability of the regression calibration model to the main study.
• To reduce bias introduced by non-transportable parameters from EVS.

Main Methods:

• Proposed an improved regression calibration method for linear regression models.
• Estimated parameters of the ME generating process using EVS.
• Obtained remaining regression calibration model parameters directly from the MS to ensure transportability.
• Derived theoretical properties of the proposed method.

Main Results:

• The proposed method effectively reduces bias in parameter estimation.
• Maintained nominal confidence interval coverage in simulation studies.
• Demonstrated the method's applicability using data from the Health Professionals Follow-Up Study and the Men's Lifestyle Validation Study.

Conclusions:

• The improved regression calibration method ensures model transportability, leading to less biased results.
• This approach enhances the reliability of epidemiological studies with ME.
• The method is effective for assessing exposure-outcome relationships, such as dietary intake and body weight.