Comparison of Commercially Available and Laboratory-Developed Assays for In Vitro Detection of SARS-CoV-2 in Clinical Laboratories

Multiple laboratory-developed tests (LDTs) and commercially available assays have emerged to meet diagnostic needs related to the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic. To date, there is limited comparison data for these different testing platforms. We compared the analytical performance of a LDT developed in our clinical laboratory based on CDC primer sets and four commercially available, FDA emergency use authorized assays for SARS-CoV-2 (Cepheid, DiaSorin, Hologic Panther, and Roche Cobas) on a total of 169 nasopharyngeal swabs.

S ince the first infection with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) was detected in the United States in January 2020 (1), there has been an exponential growth in cases and deaths (2). At the time of this writing, the U.S. case count exceeds 600,000 with more than 30,000 deaths and considerable geographic heterogeneity (2,3). Despite social distancing policies, the outbreak of coronavirus disease 2019 (COVID-19), the disease caused by SARS-CoV-2, continues to grow and threatens to overwhelm hospital systems in multiple states (2).
The explosion of COVID-19 cases in the United States has highlighted the critical role diagnostic testing plays in medical and public health decision-making in containing and mitigating the SARS-CoV-2 pandemic. Reliable test results enable appropriate utilization of scarce hospital resources, including personal protective equipment (PPE) and negative-pressure isolation rooms, as well as public health resources for contact tracing or isolation decision-making (4). In rapid succession in March 2020, multiple assays have become available, including both FDA emergency use authorization (EUA) test platforms and laboratory-developed tests (LDTs) for use in high-complexity clinical laboratories. To solve supply chain difficulties, clinical laboratories have had to implement multiple assays using scarce reagent resources, rendering thorough comparisons challenging. A clear understanding of the analytical parameters of these options is important to help guide assay selection by clinical laboratories when supply chain considerations subside (4).
Reverse transcription-PCR (RT-PCR) is the mainstay of SARS-CoV-2 detection in vitro (5). FDA EUA authorized assays for SARS-CoV-2 have mostly targeted two loci of the positive-sense, single-stranded RNA virus by real-time RT-PCR and are reported qualitatively. Our laboratory has recently reported that the CDC N2 and WHO E-gene primer/probe sets are among the most sensitive and have detected no false-positive results in our LDTs (6). FDA EUA authorized platforms use a variety of different primer/probe sets, resulting in the potential for differing analytical sensitivities. In addition to differing analytical sensitivities, commercially available platforms have important operational differences, including integrated sample extraction, run time, random access, and acceptable sample types.
Here, the performances of one LDT-EUA assay developed in our clinical laboratory and four FDA-EUA cleared assays were evaluated for detection of SARS-CoV-2. The FDA-EUA cleared assays included were Hologic Panther Fusion (both research use only [RUO] and EUA versions; EUA version performed at two study sites), DiaSorin Simplexa COVID-19 Direct (EUA), Cepheid Xpert Xpress SARS-CoV-2 (EUA), and Roche Cobas 6800 (EUA). The test performance characteristics of each RT-PCR were determined compared to those of our reference LDT assay. Cobas 6800, n ϭ 40. Finally, 28 specimens were used to compare the SARS-CoV-2 assay on the UW Panther Fusion with the DiaSorin Simplexa assay. All same-sample comparisons were performed on specimens stored at 4°C for less than 72 h with no freeze-thaws. Inconclusive results (one of two targets detected) were considered positive due to the high specificity of all assays and limited cross-reactivity seen for SARS-CoV-2 primer sets. This work was approved under a consent waiver from the University of Washington Institutional Review Board.

MATERIALS AND METHODS
Sample processing. For the UW CDC LDT, total nucleic acid (NA) was extracted from 200 l of viral transport medium (VTM) on the Roche MP96 and eluted in 50 l of elution buffer. Real-time RT-PCR was set up on 5 l of eluate using the CDC N1, N2, and RP (or Exo internal control) primers and run on ABI 7500 real-time PCR instruments as reported previously (6). For the Hologic Panther Fusion, 500 l of VTM was transferred to lysis buffer in manufacturer-provided tubes and loaded directly on the instrument. For the DiaSorin Simplexa and Cepheid Xpert Xpress, 50 l or 300 l of VTM sample, respectively, was loaded directly into the reaction cartridge with integrated sample process. For the Roche Cobas 6800, 600 l of specimen VTM was added to a barcoded secondary tube (12 by 75 mm) and loaded directly on the instrument.

RESULTS
Panther Fusion SARS-CoV-2. The Panther Fusion SARS-CoV-2 assay was tested first as research use only (RUO) reagents (n ϭ 36) and tested again following FDA authorization (n ϭ 20). Both Panther Fusion RUO and EUA assays were slightly less sensitive than the CDC-based LDT, missing one positive/inconclusive sample in each sample set (Tables 1 and 2). One additional specimen was initially negative with the RUO reagents but was detected upon repeat with the Panther Fusion EUA assay. Discordant specimens were either inconclusive (one target of two detected) or had high average threshold cycles (C T ) (Ͼ37) by the CDC LDT test. All 29 negative specimens generated "Not detected" results by the Hologic Panther Fusion SARS-CoV-2 assay.
Roche Cobas SARS-CoV-2. We next compared the Roche Cobas SARS-CoV-2 assay to our CDC LDT. All 20 negative specimens demonstrated complete concordance between the two platforms ( Table 5). One of the 20 positive specimens was not detected by the Roche assay. This specimen had C T s of 38.0 (N1) and 37.4 (N2) in the LDT. Across the 20 positive specimens, C T s were only slightly higher on the Roche Cobas assay compared to the CDC-based LDT, with an average C T difference of 0.6 (IQR, Ϫ0.1 to 1.5. Five-way same-sample comparison, including Cepheid Xpert Xpress SARS-CoV-2 assay. After performing the above pairwise comparisons, we next compared 26 specimens (13 positive, 13 negative) from another high-complexity hospital laboratory (LabCorp Seattle). All 26 specimens were also tested on the Cepheid Xpert Xpress   (Table 6). All specimens with C T s of Ͻ35 on the CDC-based LDT were detected by all platforms, and all specimens not detected by the Cepheid Xpert were not detected by two other platforms examined (CDC LDT and Roche Cobas). One of 13 positive specimens was a presumptive positive on the Cepheid assay (E-gene C T of 42.6, N2 gene negative); upon repeat per package insert, the N2 gene was detected at a C T of 42.7 but the E gene was not detected, yielding a positive result. The CDC LDT demonstrated 100% concordance with the Cepheid Xpert Xpress, also detecting the extremely low viral load specimen above as an inconclusive (N1 C T of 37.4, N2 not detected). No other assay detected SARS-CoV-2 RNA in this specimen. In addition, the DiaSorin Simplexa failed to detect a positive specimen that on repeat was detected only by the ORF1ab primer set.

DISCUSSION
This analysis compared the performance characteristics of several in vitro diagnostic real-time RT-PCR assays to detect SARS-CoV-2 in high-complexity clinical laboratories in one of the early U.S. epicenters of the COVID-19 pandemic. The results demonstrated excellent performance of a CDC-based LDT and the Cepheid  Xpert Xpress, concurring with a previous evaluation that demonstrated high sensitivity of the E-gene and N2 primer sets used by the Cepheid assay (6). The Panther Fusion was somewhat less sensitive than either the LDT or DiaSorin; however, the Panther Fusion detected SARS-CoV-2 RNA in one specimen that was inconclusive (one of two targets detected, thus presumed positive) by the UW CDC LDT. The Roche assay performed on the Cobas 6800 platform detected 28/30 positive samples; both of these discordant specimens had low viral titers (UW CDC LDT C T of Ͼ37), and one was the inconclusive specimen. Therefore, we conclude that all the tested assays show good sensitivity for the detection of SARS-CoV-2, with the UW CDC LDT and Cepheid Xpert Xpress SARS-CoV-2 assays having the best and similar sensitivity, followed by the Roche Cobas 6800, DiaSorin Simplexa, and Panther Fusion SARS-CoV-2 assays. Our results are chiefly limited by the small sample sets used to compare these different assays as well as asynchronous comparisons that allowed only for pairwise comparisons early in the pandemic. For instance, these asynchronous panels most greatly affected our CDC LDT versus Hologic Panther Fusion comparison, which had a greater proportion of high-C T positive specimens that resulted in a lower measured sensitivity for the Panther Fusion. In clinical practice, the minor differences in sensitivity are likely to have little effect on Hologic Panther Fusion SARS-CoV-2 assay performance on VTM specimens, given the C T ranges we have observed in our clinical populations.
Despite their limitations, these data provide a basis for differences in analytical sensitivity at different C T s that may be seen between platforms. For instance, recent reports have demonstrated a slightly higher analytical sensitivity of the Cepheid Xpert Xpress SARS-CoV-2 assay compared to the Roche Cobas SARS-CoV-2 test, and a slightly lower sensitivity of the DiaSorin Simplexa SARS-CoV-2 assay compared to a modified CDC assay, both of which are concordant with our data (7,8). We also note that, while analytical sensitivity is of critical importance, many other considerations factor into assay platform selection, including assay availability, cost, turnaround time, and throughput.
Our results provide an early assessment of performance characteristics of five separate assays for the detection of SARS-CoV-2. During March 2020, reagent availabil- ity for SARS-COV-2 RT-PCR assays was heavily constrained, necessitating more-limited assay comparisons. All platforms examined here had acceptable performance criteria for testing during the early part of this pandemic. As the supply chain for SARS-CoV-2 RT-PCR attempts to catch up with testing demand, we look forward to additional assay comparison data. a Thirteen negative (Neg) and thirteen positive (Pos) samples were tested. The asterisk indicates that the sample was from a known positive patient in the process of clearing virus. b NDET, not detected; nd, not done.