Abstract:
|
Pooled analyses that aggregate data from multiple studies are becoming increasingly common in collaborative epidemiologic research with the advantage to provide enhanced sample size. However, study-specific calibration processes must be incorporated in the statistical analyses to address between-study/laboratory variability in the biomarker measurements. We propose statistical methods for evaluating the biomarker-disease relationship, for categorical biomarkers, while accounting for the calibration processes in the settings of matched/nested case-control studies. A joint estimating equation method is introduced to obtain a valid asymptotic variance. We consider different sampling mechanisms of the calibration study, including sampling from controls only, sampling from both cases and controls, and sampling externally. Extensive simulation studies with varying degrees of sample sizes and biomarker-disease associations are used to evaluate finite sample performance of the proposed methods, and the methods are illustrated using a Vitamin D pooling project of colorectal cancer.
|