Letter to the editor -
We read with great interest the article by You et al 2020  in which excellent agreement between two #SARSCoV2 #coronavirus molecular diagnostic assays of different formats was reported. The cobas 6800 SARS-CoV-2 test (Roche Molecular Systems, Rotkreuz, Switzerland) is a sample-to-result platform of high throughput. Conversely, the Taiwan CDC assay is a scalable manual test with flexibility. In the setting of a sudden and severe pandemic, the adoption of multiple diagnostic platforms for clinical testing to circumvent routine service disruption due to potential supply shortage of reagents and consumables is prudent. The Taiwan CDC assay utilises a stand-alone nucleic acid extraction process which enables polymerase chain reaction (PCR) testing of a wider spectrum of specimen types with previously published primers and probes . A laboratory-developed test like this allows the detection of SARS-CoV-2 in non-respiratory swab specimens such as saliva, lower respiratory tract fluids, stool and blood while testing these specimens on the cobas 6800 SARS-CoV-2 is an off-label use of the Emergency Use Authorized test. With 3 discordant results, the positive agreement between these two studied assays was determined to be 80% (overall N = 15). However, the positive agreement was in fact 100% when comparing the results of the E gene target detection only. The three discordant samples were very likely to carry low amount of SARS-CoV-2 since the follow-up specimens were subsequently tested positive by both assays. We would like to suggest that when there is late amplification in one of the two viral targets, either in an assay with multiple singleplex components or in a multiplex PCR assay, the PCR test should be repeated, preferably in multiple replicates. In samples with low viral load close to the assay’s limit of detection (LOD), stochastic or random sampling effects contribute to the observation of single target amplification. An argument against repeating the test on the same assay in multiple replicates is that fluctuation in results between replicates will unlikely resolve the uncertainty, other than just increasing the cost and turn-around-time of reporting. In such scenario, repeating the sample with an alternate test with similar, if not better sensitivity will help to justify the final result. PCR should be repeated using nucleic acid re-extracted from the remaining primary specimens if contamination is suspected. For samples with reproducible late amplification of the E gene without co-amplification of the other SARS-CoV-2-specific target after repeat, we recommend the following statement as such “Sarbecovirus RNA was detected with low copy numbers close to the assay’s LOD. This result indicates a presumptive positive for SARS CoV-2 in patients with appropriate clinical and epidemiological findings”. As false negative results are more consequential, we will rather report an “inconclusive result” and recommend repeat sampling for retesting to safeguard from issuing a false-negative report for samples with non-reproducible late amplification of E gene. In this study You et al tested samples from symptomatic patients. However in a setting of low prevalence, false positives are more likely than false negatives and vice versa.