Comparison of screening tests without a gold standard-A pragmatic approach with virtual reference testing

Area of Science:

• Medical Diagnostics
• Epidemiology
• Biostatistics

Background:

• Diagnostic gold standards are often unavailable for rare diseases, including certain tropical infections.
• This limitation hinders the reliable evaluation and comparison of diagnostic tests.

Purpose of the Study:

• To establish a modeling approach for comparing diagnostic tests in the absence of a gold standard.
• To address situations where diagnostic tests may not meet the conditional independence requirement.

Main Methods:

• The study builds upon previous work by Gart and Buck, Rogan and Gladen, and Hui and Walter.
• A novel modeling approach is introduced, utilizing two populations with varying prevalence rates.
• Virtual reference tests are defined based on the results from these two populations.

Main Results:

• The proposed model allows for the comparison of diagnostic tests without requiring a gold standard or reliably characterized reference samples.
• It is particularly applicable to scenarios where conditional independence between tests is not met.

Conclusions:

• This modeling approach provides a valuable tool for evaluating diagnostic tests for rare tropical diseases where gold standards are typically absent.
• It offers a method to compare test performance even under challenging diagnostic circumstances.