Evaluation of the accuracy, ease of use and limit of detection of novel, rapid, antigen-detecting point-of-care diagnostics for SARS-CoV-2 (preprint)

ABSTRACT

Abstract Background: Reliable point-of-care (POC) diagnostics not requiring laboratory infrastructure could be a game changer in the COVID-19 pandemic, particularly in the Global South. We assessed performance, limit of detection (LOD) and ease-of-use of three antigen-detecting, rapid POC diagnostics (Ag-RDT) for SARS-CoV-2. Methods: This prospective, multi-centre diagnostic accuracy study, recruited participants suspected to have SARS-CoV2 in Germany and UK. Paired nasopharyngeal swabs (NP) or NP and/or oropharyngeal swabs (OP) were collected from participants (one for clinical real-time reverse transcription polymerase chain reaction (RT-PCR) and one for Ag-RDT testing). Performance of each of three Ag-RDTs was compared to RT-PCR overall, and according to predefined subcategories e.g. cycle threshold (CT)-value, days from symptom onset, etc. In addition, limited verification of analytical limit-of-detection (LOD) was determined. To understand the usability of each Ag-RDT a System Usability Scale (SUS) questionnaire and ease-of-use assessment were performed. Results: Between April 17th and August 25th, 2020, 2417 participants were enrolled, with 70 (3.0%) testing positive by RT-PCR. The best-performing test (SD Biosensor, Inc. STANDARD Q) was 76.6% [95% Confidence Interval (CI) 62.8-86.4] sensitive and 99.3% [CI 98.6-99.6] specific. A sub-analysis showed all samples with RT-PCR CT-values <25 were detectable by STANDARD Q. The test was considered easy-to-use (SUS 86/100) and suitable for POC. Bioeasy and Coris showed specificity of 93.1% [CI 91.0%-94.8%] and 95.8% [CI 93.4%-97.4%], respectively, not meeting the predefined target of [≥]98%. Conclusion: There is large variability in performance of Ag-RDT tests with one test showing promise. Given the usability at POC, these tests are likely to have impact despite imperfect sensitivity; however further research and modelling are needed.