Analysis of Matched Case-Control Study with a Misclassified Exposure

• Activity Number: 123 - Binary and Ordinal Outcome Regression
• Type: Contributed
• Date/Time: Monday, July 30, 2018 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Samiran Sinha* and Christopher Manuel and Suojin Wang
• Companies: Texas A & M University and Texas A&M University and Texas A&M University
• Keywords: conditional logistic; control group; instumental variable; identifiability; misclassification; odds ratio

Abstract

Matched case-control studies are used for finding the association after controlling the effect of important confounding variables. It is a known fact that the estimators of the regression parameters are biased when the exposure is misclassified, and case-control and matched case-control study is of no exception. Any bias correction method relies on a validation data that contain the true exposure and the misclassified exposure value. In contrast, here we propose a consistent method of correcting such bias in the analysis of a matched case-control data when measurements on an instrumental variable are available for each subject in the study. The significance of this approach is that it works without any validation data that often is not available when the measurement of the true exposure is impossible or too costly to obtain. The asymptotic property of the method is theoretically justified. The finite sample performance of the proposed method is assessed via simulation studies. The method is illustrated by analyzing a real dataset on the US birth cohort.