Abstract
This study was performed as a head-to-head comparison of the performance characteristics of (1) two SARS-CoV-2-specific rapid antigen assays with real-time PCR as gold standard as well as (2) a fully automated high-throughput transcription-mediated amplification (TMA) assay and real-time PCR in a latent class analysis-based test comparison without a gold standard with several hundred samples in a low prevalence "real world" setting. Recorded sensitivity and specificity of the NADAL and the LumiraDx antigen assays and the Hologic Aptima SARS-CoV-2 TMA assay were 0.1429 (0.0194, 0.5835), 0.7644 (0.7016, 0.8174), and 0.7157 (0, 1) as well as 0.4545 (0.2022, 0.7326), 0.9954 (0.9817, 0.9988), and 0.9997 (not estimable), respectively. Agreement kappa between the positive results of the two antigen-based assays was 0.060 (0.002, 0.167) and 0.659 (0.492, 0.825) for TMA and real-time PCR. Samples with low viral load as indicated by cycle threshold (Ct) values > 30 were generally missed by both antigen assays, while 1:10 pooling suggested higher sensitivity of TMA compared to real-time PCR. In conclusion, both sensitivity and specificity speak in favor of the use of the LumiraDx rather than the NADAL antigen assay, while TMA results are comparably as accurate as PCR, when applied in a low prevalence setting.