Regression calibration in studies with correlated variables measured with error

Area of Science:

• Epidemiology
• Biostatistics

Background:

• Measurement error in exposure variables introduces bias into regression analyses.
• This bias can lead to a significant loss of statistical power in multivariate models.
• Accurate measurement of exposure variables is crucial for reliable study outcomes.

Purpose of the Study:

• To evaluate the effectiveness of regression calibration in correcting biases caused by measurement error.
• To assess the impact of correlated exposure variables on statistical power.
• To determine the influence of calibration study size on statistical power.

Main Methods:

• Utilized regression calibration logistic analyses on simulated study populations.
• Employed calibration data from California Seventh-day Adventists to create new calibration studies.
• Estimated statistical power for pairs of nutritional variables.

Main Results:

• Regression calibration effectively corrects large biases in estimated effects when measurement error is present.
• Strong correlations between variables measured with error lead to substantial power loss.
• Type I error probabilities can be non-nominal, causing false significance in crude analyses, especially in large studies.

Conclusions:

• Regression calibration is a vital technique for mitigating bias and power loss due to measurement error.
• Collinearity between variables and the correlation between crude and true variables are key determinants of power.
• Increasing calibration study size up to 1,000 subjects yields significant power gains when collinearity is present.