Real-World Performance of COVID-19 Antigen Tests: Predictive Modeling and Laboratory-Based Validation.

Area of Science:

• Diagnostic Test Evaluation
• Regulatory Science
• Infectious Disease Diagnostics

Background:

• Real-world performance assessment of antigen tests lacks quantitative, lab-based approaches.
• Current regulatory science rarely employs laboratory and quantitative methods for antigen test appraisal.
• Standardized and accelerated early-phase appraisal is crucial for pandemic preparedness.

Purpose of the Study:

• To present a quantitative, laboratory-anchored framework for antigen test evaluation.
• To link image-based test line intensities and limit of detection (LoD) to probabilistic positive percent agreement (PPA).
• To predict PPA as a function of viral-load-related variables, such as quantitative real-time polymerase chain reaction (qRT-PCR) cycle thresholds (Cts).

Main Methods:

• Combined quantitative evaluation of test signal response with target concentrations.
• Statistically characterized the LoD using observer visual acuity.
• Calibrated a gold-standard method (qRT-PCR) against virus concentration.
• Developed a Bayesian predictive model for real-world antigen test performance.

Main Results:

• Applied the methodology to characterize COVID-19 antigen tests and estimate real-world agreement with qRT-PCR.
• Modeled PPA as a continuous function of qRT-PCR Ct using logistic regression.
• Standardized performance comparisons across different testing sites were enabled.

Conclusions:

• Modeling real-world performance requires coupling laboratory evaluation with user perception of test signals.
• The framework generates PPA-Ct curves by integrating signal-to-concentration models, LoD distribution, and Ct-to-viral-load calibration.
• Inferences are context-specific; external validity depends on the test, user population, and calibration.