Estimating the receiver operating characteristic curve in matched case control studies.

Area of Science:

• Epidemiology
• Biostatistics
• Medical Informatics

Background:

• Matched case-control designs offer efficiency for complex disorder studies, particularly with biomarker analysis.
• Risk prediction in matched case-control studies presents unique challenges.
• Existing methods may not fully leverage combined cohort and case-control data for prediction.

Purpose of the Study:

• To propose an inverse-probability weighting approach for estimating predictive ability in matched case-control studies.
• To develop an algorithm for calculating the area under the curve (AUC) for risk prediction models.
• To provide methods for estimating population parameters and AUC when combining parent cohort and matched case-control data.

Main Methods:

• Developed an inverse-probability weighting algorithm for risk prediction.
• Proposed methods to estimate population parameters and AUC by integrating cohort and matched case-control data.
• Evaluated method bias through simulations across various parameter settings.

Main Results:

• The proposed inverse-probability weighting method allows for estimation of predictive ability in matched case-control studies.
• The algorithm successfully estimates the area under the curve (AUC) for covariate-based prediction of binary outcomes.
• Simulations indicated acceptable bias under diverse parameter settings, validating the approach.

Conclusions:

• The inverse-probability weighting approach provides a robust method for risk prediction in matched case-control settings.
• This method enhances the utility of matched case-control studies for biomarker research and disease risk assessment.
• The approach is applicable to various complex diseases, as demonstrated in cardiovascular disease and breast cancer studies.