Combining and comparing area estimates across studies or strata.

Area of Science:

• Medical Statistics
• Diagnostic Test Evaluation
• Biomarker Research

Background:

• Comparing diagnostic test accuracy across studies is crucial for evidence-based medicine.
• Existing methods may lack simplicity or flexibility in combining results.
• Receiver Operating Characteristic (ROC) curve analysis is a standard tool for assessing test performance.

Purpose of the Study:

• To present a straightforward method for combining and comparing medical test performance metrics.
• To utilize the area under the ROC curve (AUC) as the primary parameter for comparison.
• To introduce a statistical test for evaluating the equality of AUCs across different studies or strata.

Main Methods:

• Calculating a weighted average of the area under the ROC curve (AUC) from individual studies or strata.
• Employing a chi-square test to assess the statistical significance of differences in AUCs.
• Exploring the statistical power of the proposed comparison test.

Main Results:

• The proposed method provides a simple approach to meta-analysis of diagnostic test accuracy.
• The weighted average AUC offers a combined measure of test performance.
• The chi-square test allows for formal comparison of test performance across different settings.

Conclusions:

• The presented method is practical and requires only basic statistical inputs (AUC estimates and standard errors).
• Both parametric and nonparametric AUC estimates can be incorporated.
• This approach facilitates more robust comparisons of medical tests in diverse populations or study designs.