Comparing disease screening tests when true disease status is ascertained only for screen positives.

Area of Science:

• Medical screening
• Biostatistics
• Health services research

Background:

• Evaluating screening test accuracy typically requires comparison with definitive diagnostic tests.
• Definitive tests are often invasive and ethically limited to individuals with positive screening results, preventing absolute accuracy assessment.
• This limitation is common in cancer screening, where biopsies are only performed on screen-positive cases.

Purpose of the Study:

• To develop a statistical framework for comparing the relative accuracy of screening tests when definitive diagnostic data are only available for screen-positive individuals.
• To enable the evaluation of factors influencing the relative accuracies of different screening modalities.
• To provide methods for determining when one screening test may be preferable to another.

Main Methods:

• Development of a marginal regression analysis framework to infer relative accuracy.
• Utilizing relative true positive rates and relative false positive rates as measures of comparative accuracy.
• Application of the framework to analyze factors affecting screening test performance in real-world studies.

Main Results:

• Demonstrated that relative accuracy can be inferred even when absolute accuracy cannot be determined.
• Established a method analogous to case-control studies for relative risk estimation.
• Identified factors influencing relative screening test accuracies in prostate and cervical cancer screening examples.

Conclusions:

• The proposed marginal regression framework allows for robust inference on relative screening test accuracy.
• This method is valuable for comparing screening modalities when full diagnostic follow-up is not possible.
• Understanding factors influencing relative accuracy is crucial for optimizing healthcare screening programs.