Evaluation and Modelling of the Performance of an Automated SARS-CoV-2 Antigen Assay According to Sample Type, Target Population and Epidemic Trends

Abstract

The Lumipulse^® G SARS-CoV-2 Ag assay performance was evaluated on prospectively collected saliva and nasopharyngeal swabs (NPS) of recently ill in- and outpatients and according to the estimated viral load. Performances were calculated using RT-PCR positive NPS from patients with symptoms ≤ 7 days and RT-PCR negative NPS as gold standard. In addition, non-selected positive NPS were analyzed to assess the performances on various viral loads. This assay yielded a sensitivity of 93.1% on NPS and 71.4% on saliva for recently ill patients. For NPS with a viral load > 10³ RNA copies/mL, sensitivity was 96.4%. A model established on our daily routine showed fluctuations of the performances depending on the epidemic trends but an overall good negative predictive value. Lumipulse^® G SARS-CoV-2 assay yielded good performance for an automated antigen detection assay on NPS. Using it for the detection of recently ill patients or to screen high-risk patients could be an interesting alternative to the more expensive RT-PCR.

Keywords: COVID-19; SARS-CoV-2; assay; diagnostic; model; test.

Figure 1

Receiver Operating Curve (ROC) analysis of antigen quantification (Ag) on nasopharyngeal swabs (NPS) vs. RT-PCR on NPS as gold standard.

Figure 2

Receiver Operating Curve (ROC) analysis of antigen quantification (Ag) on saliva vs. RT-PCR on nasopharyngeal swabs (NPS) as gold standard.

Figure 3

Modelling SARS-CoV-2 automated antigen detection performance on epidemic trends. Data computed from 1 May 2020 to 30 October 2021 using a backward sliding window of 14 days (14-day Se: 14-day sensitivity, 14-day PPV: 14-day Positive Predictive Value, 14-day NPV: 14-day Negative Predictive Value).