A likelihood approach to proper analysis of secondary outcomes in matched case-control studies

TL;DR

This paper introduces a likelihood-based statistical approach for analyzing secondary outcomes in matched case-control studies, addressing biases from traditional methods and improving estimation accuracy.

Contribution

It proposes novel likelihood methods that properly account for study design and sampling, enhancing the validity of secondary outcome analysis in matched case-control studies.

Findings

Methods provide consistent estimates and accurate confidence intervals.

Simulation studies show improved performance over naive methods.

Application to diabetes data demonstrates practical utility.

Abstract

Matched case-control studies are commonly employed in epidemiological research for their convenience and efficiency. Analysis of secondary outcomes can yield valuable insights into biological pathways and help identify genetic variants of importance. Naive analysis using standard statistical methods, such as least-squares regression for quantitative traits, can be misleading because they fail to account for unequal sampling induced by the case-control design and matching. In this paper, we propose novel statistical methods that appropriately reflect the study design and sampling scheme in the analysis of secondary outcome data. The new methods provide consistent estimation and accurate coverage probabilities for the confidence interval estimators. We demonstrate the advantages of the new methods through simulation studies and a real application with diabetes patients. R code implementing the proposed methods is publicly available.