Statistical methods for analysis of combined biomarker data from multiple nested case–control studies

By combining data across multiple studies, researchers increase sample size, statistical power, and precision for pooled analyses of biomarker–disease associations. However, researchers must adjust for between-study variability in biomarker measurements. Previous research often treats the biomarker measurements from a reference laboratory as a gold standard, even though those measurements are certainly not equal to their true values. This paper addresses measurement error and bias arising from both the reference and study-specific laboratories. We develop two calibration methods, the exact calibration method and approximate calibration method, for pooling biomarker data drawn from nested or matched case–control studies, where the calibration subset is obtained by randomly selecting controls from each contributing study. Simulation studies are conducted to evaluate the empirical performance of the proposed methods. We apply the proposed methods to a pooling project of nested case–control studies to evaluate the association between circulating 25-hydroxyvitamin D (25(OH)D) and colorectal cancer risk.