In matched case-control studies with incomplete exposure data, two analysis methods are commonly used: a matched analysis (conditional logistic regression) of the complete pairs and an unmatched analysis (logistic regression) of all subjects with exposure variable by ignoring the matching. The missing-indicator method, which has the advantage of making use of all the data while still preserving the matching, was recommended as an alternative method (American Journal of Epidemiology, 150(12), 1340-5, 1999). Since the performance of the missing-indicator method remains unknown, Monte Carlo simulation was conducted to evaluate three methods with incomplete data. Data were generated from a 1:1 matched case-control design. The completed data were first analyzed using conditional logistic regression and yielded a standard error (SE), which was used as a reference for SE comparison. Given different target values and SE reference, when confounding effects exist and missing percentage of exposure is the same in both case and control group, conditional logistic regression yields unbiased estimated log odds ratio but biggest standard errors, while logistic regression yields the most biased log odds ratios but appropriate SEs; the missing-indicator method produces much less biased estimated logOR and bigger SEs. Therefore, the missing-indicator is a compromise between the conditional logistic regression of the complete pairs and the logistic regression of subjects with exposure variable, and could be an alternative approach in matched case-control studies with missing exposure.
Learning Objectives: At the conclusion of the session, the participant (learner) in this session will be able to: 1) identify the advantages and disadvantages of logistic regression, conditional logistic regression, and missing-indicator method in analyzing 1:1 matched case-control data with missing exposure. 2) apply the missing-indicator method in 1:1 matched case-control studies with missing exposure
Keywords: Methodology, Simulation
Presenting author's disclosure statement:
Organization/institution whose products or services will be discussed: None
I do not have any significant financial interest/arrangement or affiliation with any organization/institution whose products or services are being discussed in this session.