Evaluation of a Rapid Antigen Test To Detect SARS-CoV-2 Infection and Identify Potentially Infectious Individuals

ABSTRACT The identification and isolation of highly infectious SARS-CoV-2-infected individuals is an important public health strategy. Rapid antigen detection tests (RADT) are promising tools for large-scale screenings due to timely results and feasibility for on-site testing. Nonetheless, the diagnostic performance of RADT in detecting infectious individuals is not yet fully determined. In this study, RT-qPCR and virus culture of RT-qPCR-positive samples were used to evaluate and compare the performance of the Standard Q COVID-19 Ag test in detecting SARS-CoV-2-infected and possibly infectious individuals. To this end, two combined oro- and nasopharyngeal swabs were collected at a routine SARS-CoV-2 diagnostic center. A total of 2,028 samples were tested, and 118 virus cultures were inoculated. SARS-CoV-2 infection was detected in 210 samples by RT-qPCR, representing a positive rate of 10.36%. The Standard Q COVID-19 Ag test yielded a positive result in 92 (4.54%) samples resulting in an overall sensitivity and specificity of 42.86 and 99.89%, respectively. For adjusted CT values of <20 (n = 14), <25 (n = 57), and <30 (n = 88), the RADT reached sensitivities of 100, 98.25, and 88.64%, respectively. All 29 culture-positive samples were detected by the RADT. Although the overall sensitivity was low, the Standard Q COVID-19 Ag test reliably detected patients with high RNA loads. In addition, negative RADT results fully corresponded with the lack of viral cultivability in Vero E6 cells. These results indicate that RADT can be a valuable tool for the detection of individuals with high RNA loads that are likely to transmit SARS-CoV-2.

As high RNA loads are typically associated with a higher probability of infectiousness, the diagnostic performance of RADT in the context of infectivity models is yet to be determined (27)(28)(29). Therefore, there is a need for large-scale field studies with a focus on virus cultivability to be able to appropriately interpret RADT results. Here, we examined the performance of RADT in detecting both infected and infectious individuals in a routine diagnostic center using RT-qPCR and virus culturing as reference methods.

MATERIALS AND METHODS
Study design. Between 26 October 2020 and 11 January 2021, all individuals were tested for SARS-CoV-2 infection by RT-qPCR at the University Hospital Cologne, Department I of Internal Medicine, Cologne, Germany. SARS-CoV-2 testing was available for individuals from the general population with COVID-19 symptoms, suspected disease, or SARS-CoV-2 exposure as part of routine diagnostics, as well as for hospital staff members as part of screening measures. For quality control, an RADT was simultaneously performed after verbal consent. Two swab specimens, each collected from oro-and nasopharynges, were obtained by the same trained personnel. The first swab was transferred into virus transport and preservation medium (Biocomma, Shenzhen, China) or BD ESwab (Becton Dickinson, Franklin Lakes, NJ, USA) for RT-qPCR testing, and the second swab was used for on-site RADT testing. All samples for RT-qPCR were routinely processed within 12 h after collection. Upon approval by the Institutional Review Board of the University of Cologne, results were retrospectively analyzed, including clinical data retrieved from a symptoms diary webtool that all individuals registering for a SARS-CoV-2 test are asked to complete. Due to several implausible self-reported entries, only webtool entries not older than a week from the time of testing were included into symptom analysis. Patients were categorized as symptomatic if the reported symptom duration at the time of testing was #14 days and one of the following symptoms was found: fever, cough, rhinorrhea, nausea, diarrhea, shortness of breath, and/or a new olfactory or taste disorder.
Rapid antigen detection test. The Standard Q COVID-19 Ag test (SD Biosensor, Inc., Suwon-si, Republic of Korea/Hoffmann La Roche AG, Basel, Switzerland) is a rapid chromatographic immunoassay for the qualitative detection of SARS-CoV-2 nucleocapsid protein. The test was performed according to the manufacturer's instructions using the enclosed dry swab for sample collection with one modification. Instead of a nasopharyngeal swab only, a combined oro-and nasopharyngeal swab was performed to ensure comparability with RT-qPCR sample collection. The operating instructions, in brief, were as follows. The collected swab was mixed in the provided tube of collection medium, and three drops were applied through a nozzle cap onto the test strip. Results were read out visually after 15 to 20 min by medically trained and instructed personnel. In accordance with the manufacturer's reference guide, faint lines were considered positive if the control line was also present.
Real-time reverse transcription-PCR. RT-qPCR was performed using different SARS-CoV-2 RNA detection protocols that were normalized according to the same standard. The following SARS-CoV-2 detection protocols were utilized. (i) Nucleic acid extraction was done for 935 (46.10%) samples using the MagNA Pure 96 system DNA and viral NA large volume kit (Roche Diagnostics, Mannheim, Germany) according to the manufacturer's instructions. After RNA purification from 500 ml of viral transport medium and elution into 100 ml of elution buffer, multiplex RT-qPCR was performed using an in-house Ngene primer set on a LightCycler 480 II system (Roche Diagnostics). The LightMix SarbecoV E-gene plus equine arteritis virus (EAV) control kit (TIB Molbiol, Berlin, Germany) was included in every RT-qPCR run. (ii) A cobas SARS-CoV-2 test kit running on the cobas 6800 (Roche Diagnostics) was used for 407 (20.07%) samples targeting the viral E-gen and ORF1a/b regions according to the manufacturer's instructions. (iii) A SARS-CoV-2 AMP kit running on an Alinity m (Abbott, Chicago, IL, USA) was used for 63 (3.11%) specimens targeting the viral N and RdRp genes according to the manufacturer's instructions. (iv) A multiplex RT-qPCR with a LightMix SarbecoV E-gene (TIB Microbiol), an in-house N-gene primer/ probe set, and a human b-globin primer set as an internal control, running on the Panther Fusion (Hologic, Wiesbaden, Germany), was used for 36 (1.78%) samples. (v) For 579 (28.55%) specimens, samples from up to 10 asymptomatic employees were pooled and tested for SARS-CoV-2 using the methods described in steps i, ii, and iii. Positive pools were resolved, and samples were tested separately as described previously in steps i to iv. (vi) An Xpert Xpress SARS-CoV-2 (Cepheid, Sunnyvale, CA, USA) test kit was used for 8 (0.39%) samples according to the manufacturer's instructions.
To enable comparison of cycle threshold (C T ) values obtained by the different RT-qPCR methods, C T values were translated into copies/ml and then converted to a cobas 6800 adjusted C T value. For this purpose, seven serial dilutions from a high titer SARS-CoV-2 sample were tested by all five RT-qPCR methods described above. Standard curves for each amplification method were generated by a regression model. For the following conversion of device-specific C T values into copies/ml, two SARS-CoV-2 samples with a quantified RNA load from INSTAND (Society for the Promotion of Quality Assurance in Medical Laboratories, e.V., Düsseldorf, Germany; in cooperation with the Robert Koch-Institute and the Institute of Virology, Charité, Berlin) were tested on every device and subsequently used for C T -based absolute RNA quantification.
SARS-CoV-2 culture. Vero E6 cells (ATCC CRL-1586) were cultured in complete medium (CM) consisting of Dulbecco modified Eagle medium (Thermo Fisher Scientific/Gibco, Waltham, MA, USA) supplemented with 10% fetal calf serum (FCS; GE Healthcare, Chicago, IL, USA), 200 U/ml penicillin, 200 mg/ml streptomycin, 0.25 mg/ml amphotericin B, 2 mM L-glutamine, and 1 mM sodium pyruvate (all from Thermo Fisher Scientific/Gibco) at 37°C in an incubator with 5% CO 2 . One day prior to infection, 3 Â 10 5 cells were seeded onto T25 flasks in 5 ml of CM. Retained swab samples were stored for up to 8 h at 4°C in transport medium until RT-qPCR results became available. For virus cultures, 250 ml of SARS-CoV-2 RT-qPCR-positive samples were diluted 1:5 in infection medium (IM) consisting of complete medium with FCS reduced to 2%. After removal of the cell culture supernatant, the diluted samples were added to Vero E6 cells and incubated for 1 h at 37°C and 5% CO 2 . After washing with 5 ml of phosphate-buffered saline (Thermo Fisher Scientific/Gibco), 5 ml of IM was added, and the cells were cultured for 7 days at 37°C and 5% CO 2 . The cells were checked for the presence of cytopathic effects (CPE) on days 4 and 7. On both days, 1 ml of culture supernatant was harvested and stored at 280°C for conformation of positive cultures through RT-qPCR as described above. An increase in RNA load and unchanged RNA loads of .9 log 10 copies/ml between day 4 and day 7 were considered successful virus cultivation. All virus isolation experiments were performed under BSL-3 conditions. Dilution series. To validate our SARS-CoV-2 culturing approach, four dilution series with sample CT values ranging from 15.5 to 17.6 were analyzed for virus recovery and RADT result. Two hundred fifty microliters of retained samples was thawed and serially diluted in IM (1:5). Virus cultivation and RT-qPCR were performed as described above. RADT were performed with 250 ml of original samples and each dilution as described above. Virus cultivation and RADT testing showed that all SARS-CoV-2-positive cultures were previously detected by RADT (see Fig. S1 in the supplemental material).
Statistical analysis. Sensitivity (positive percent agreement) and specificity (negative percent agreement), as well as positive and negative predictive values, were calculated using RT-qPCR as a reference. Culture and RADT results were evaluated by a contingency table, and a P value was calculated with Fisher exact test. Confidence intervals (CI) were calculated using the Wilson/Brown method. Mann-Whitney U-test (MWU) was used to compare differences between medians. P values of ,0.05 were considered significant. Probit regression was carried out using a generalized linear model (R-function glm) with the probit link function. To correct for repeated measurements from the same individual on different days, the basic analyses were modified as follows. (i) Confusion matrices for the calculation of all performance measures were calculated in a weighted manner so that the contribution of a single test is inversely proportional to the number of tests taken from the corresponding individual. (ii) To compare two populations of data points, a generalized estimating equation (R-function gee::gee) was fitted using each patient as its own cluster and an exchangeable correlation structure. (iii) Probit regression was carried out by fitting a generalized linear mixed model with probit link function and random intercepts for each individual (R-function GLMMadaptive::mixed_model). Marginal means and confidence bands were calculated with R-function ggeffects::ggpredict. Data analysis was performed using Microsoft Excel 16.44 (Microsoft), Prism 9 (GraphPad Software, Inc.), Python 3.8.3, and R 3.6.3.
Ethics. The Institutional Review Board of the University of Cologne acknowledged and approved the study under application 21-1039.
Data availability. All data used in the analysis will be made available upon request.

RESULTS
RT-qPCR and RADT testing in a large cohort under real-life conditions. To validate RADT performance, two swabs, each obtained from oro-and nasopharynges, were collected and tested using both RT-qPCR and RADT. RT-qPCR-positive samples were additionally cultivated in Vero E6 cells to determine the ability of the RADT to detect replication-competent virus in individual specimens. A total of 2,032 samples were tested, 4 of which were excluded due to 3 RADT results not recorded and 1 incorrect execution of the RADT leaving 2,028 (99.80%) samples from 1,849 individuals eligible for analysis. SARS-CoV-2 was detected in 210 samples by RT-qPCR representing a study prevalence of 10.36% (Fig. 1A). Data on symptoms were obtained and analyzed for 1,676 (82.64%) of 2,028 samples. At the time of sampling 866 (42.70%) swabs were taken from symptomatic individuals, while 810 (39.94%) specimens were collected from asymptomatic individuals. For 352 (17.35%) samples, the symptom status of the respective patients was unknown at the time of analysis. For 599 specimens (69.17%), individuals reported up to three symptoms, for 247 (28.52%) more than three symptoms, and for 20 samples (2.31%) the number of symptoms was not reported by the respective subject. In our cohort 320 (15.78%) samples were obtained from 141 (7.62%) individuals who were tested twice or more (Fig. 1B) (Fig. 1C). The median adjusted C T value determined by RT-qPCR was 31.49 (IQR = 24.19 to 34.16) (Fig. 1D). Of 210 RT-qPCR-positive and 2 RT-qPCR-inconclusive samples, 126 (60%) were cultivated in Vero E6 cells on the same day of sample collection (Fig. 1A). Of these, 8 (4.80%) were excluded due to culture contaminations (n = 6) or negative RT-qPCR results upon retesting of inconclusive samples (n = 2) (Fig. 1B). A detailed cohort description can be found in Table S1 in the supplemental material.
Decreased RADT sensitivity over the course of symptom duration. We determined that 130 (15.01%) and 73 (8.43%) of 866 samples from symptomatic subjects tested positive in RT-qPCR and RADT, respectively. Symptom duration at the time of sampling was reported for 860 (99.31%) samples with a median duration of 2 days since symptom onset for both RT-qPCR-positive (IQR = 1 to 6) and -negative (IQR = 1 to 4) samples (Fig. 3A). Of samples tested positive by RT-qPCR, RADT detected 56 (68.29%) samples within 4 days since symptom onset. When reported, median symptom duration for RT-qPCR-positive samples with either a RADT-positive (n = 69) or -negative (n = 55) result was 2 days (IQR = 1 to 3;   3B). For samples from symptomatic patients, the median adjusted C T and median copies/ml were 28.90 (IQR = 23.17 to 32.80) and 4.83 log 10 (IQR = 3.53 log 10 to 6.74 log 10 ), respectively. In individuals not reporting any symptoms, detected RNA load was significantly lower (median adjusted C T = 33.45 [IQR = 31.26 to 35.31]; median copies/ml = 3.31 log 10 [IQR = 2.69 log 10 to 4.05 log 10 ; P , 0.0001]) (Fig. 3C). Samples with RNA concentrations of .6 log 10 were only observed up to 6 days after symptom onset. Thus, we conclude that the sensitivity of the RADT decreased with symptom duration as RNA loads declined (Fig. 3D).
Negative RADT result corresponds with lack of viral cultivability. For 118 inoculated cultures, CPE was observed in 29 (24.58%). To confirm virus replication, RT-qPCR of culture supernatant taken on days 4 and 7 was performed as described previously. The observed CPE and positive RT-qPCR results matched in 116 (98.30%) cases. Since CPE is operator dependent, RT-qPCR results were used for further analysis ( Table 2). Initial C T values and copies/ml of positive cultures ranged from 15.64 to 24.97 and 6.14 log 10 to 9.25 log 10 , whereas negative cultures ranged from 21.5 to 38.27 or 1.71 log 10 to 7.30 log 10 , respectively   Table 2). Therefore, the calculated PPV of RADT for viral cultivability in Vero E6 cells was 53.70% (95% CI = 40.61 to 66.31) and the NPV was 100% (95% CI = 94.34 to 100.00; Table 2). For RADT-and culture-positive results, the median adjusted C T was 20.8 and the median RNA load in copies/ml was 7.53 log 10 compared to 33.39 and 3.34 log 10 for RADTand culture-negative results. The median adjusted C T and copies/ml for RADT-positive but culture-negative results were 25.43 (IQR = 23.39 to 27.72) and 5.99 log 10 (IQR = 5.23 log 10 to 6.67 log 10 ), respectively (Fig. 4B). With viral RNA declining over the course of disease, the ability to isolate virus decreased, with no positive culture after 6 days since symptom onset (Fig. 4C). Probit regression of RADT and cell culture analyses show greater probabilities of positive result (PPR) for low adjusted C T values and high RNA loads, respectively. The virus culture assay shows 90 and 50% PPR for an adjusted C T value around 21.45 or 7.31 log 10 copies/ml and 23 or 6.8 log 10 copies/ml, respectively. RADT show a PPR of 90 and 50% at an adjusted C T value of 24.7 or 6.24 log 10 copies/ml and 29.0 or 4.78 log 10 copies/ml, respectively (Fig. 4D). In summary, these data show that a negative RADT result can reliably predict noninfectiousness in Vero E6 cells.
Adjustments for repeated testing reveal no statistically significant difference. To evaluate whether repeated testing of several individuals affects our conclusions, additional statistical analyses were carried out (see Materials and Methods). Correction for repeated RADT measurements was based on weighted counts (see Fig. S2A) rounded to whole numbers, yielding very similar results compared to the unweighted variant (see Fig. S2). In addition, the performance of the RADT was analyzed for the first measurement of each subject only (simplified analysis; Fig. S2A). Calculated sensitivities showed almost identical values as overall sensitivities described above (see Fig. S2B to D). For all studies that applied an MWU test (Fig. 2B, 3C, and 4A and B), the P values remain P , 0.0001 according a generalized estimating equation (GEE).

DISCUSSION
RADT are inexpensive and fast diagnostic tools that can be immediately performed at the point of care. Here, we present comprehensive data on the use of Standard Q COVID-19 RADT for high-throughput testing of a large cohort tested under real-life conditions in a SARS-CoV-2 outpatient diagnostic center.
Implementation and execution of the RADT testing was completed without any difficulties. Upon following the manufacturer's instructions, there were only two cases in which the result could not be read out clearly. As a consequence, we were able to conduct 2,028 paired RT-qPCR and RADT tests directly on site and cultivate virus on the same day without prior sample freezing. While the sensitivities of other RADT vary between 24 and 93% in different studies (33)(34)(35), the reported Standard Q RADT sensitivities are mostly in the range from 68 to 90% (14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24). The overall diagnostic sensitivity observed in our study was 42.86%. However, the investigated cohort of nonhospitalized patients was to a large extent comprised of individuals  with adjusted C T values over 30 (122/210). Stratification by RNA load revealed that the Standard Q RADT performed reliably for patients with RNA loads over 5.4 log 10 or below C T 27, which is in accordance to recent studies (14,23,25,36). Since an RNA concentration of 6 log 10 copies/ml is commonly suspected to be the threshold for contagiousness, we aimed to investigate the correlation between RADT result and SARS-CoV-2 in vitro infectivity (28,29). In contrast to highly sensitive nucleic acid-based detection methods that do not specifically test for intact viral particles required for transmission, viral culture is a commonly used, albeit laborious, method outside of routine diagnostics to determine the presence of infectious virus in samples (7,(30)(31)(32)37). Although the detection of viable virus in cell culture models is strong evidence of infectiousness, a negative result does not eliminate the possibility of human transmission (11,38). Moreover, the validity of viral culture as a surrogate for infectivity may depend on the susceptibility of the cell line used (31,32,37). However, loss of infectious titer in classical Vero E6 cells has been associated with a lack of transmission despite detection of viral RNA in preclinical models (39).
To investigate whether the Standard Q RADT might be able to reliably detect culture-positive samples in Vero E6 cells, we attempted virus isolation from samples positive for SARS-CoV-2 RNA. All samples from which virus could be recovered had previously tested positive in RADT. Moreover, none of the samples tested negative by RADT contained infectious virus determined by cell culture. Furthermore, when taking symptom duration into account, we detected no positive culture 6 days after symptom onset, indicating a decreased probability of recovering viable virus as symptom duration increases (29,30,(40)(41)(42); at the same time, the RADT identified positive samples for up to 9 days. While some groups have described virus isolation from samples above C T 30 (31,43), our study only observed positive cultures from samples with higher RNA loads, which is in accordance with previous observations (10,36,41,44). Taking the time of suspected exposure and duration of symptoms into account (31,38), our results suggest that RADT testing is of potential use for estimating infectivity at the time of sampling.
This study, however, is subject to some limitations. Although the examined singlecenter study population was large, our cohort might not be considered representative of the general population due to young age and disproportionate gender distribution. The data on symptoms and their duration are only reliable to a limited extent, since they were retrospectively analyzed from mostly self-reported symptoms entered into a web tool. Furthermore, instead of a nasal swab, we used a combined oro-and nasopharyngeal swab to investigate RADT performance, which impedes the feasibility for the general public. Lastly, although we have corrected for using different SARS-CoV-2 detection assays by reporting data in RNA loads, the diverse set of amplification methods remains a limitation.
In combating overdispersed SARS-CoV-2 transmission, rapid detection and isolation of highly infectious individuals is a primary goal (8,9,(45)(46)(47)(48)(49). In our investigation the Standard Q RADT was able to reliably detect high RNA loads, as well as all culture-positive samples. Therefore, this test could be used as a fast surrogate marker for viral cultivation to identify and prevent SARS-CoV-2 transmissions by highly infectious individuals. Although less sensitive than RT-qPCR and therefore less suited for an inpatient setting, RADT could compensate for this disadvantage through easy and feasible mass screenings in a community setting (8,50). Furthermore, one might suspect that RT-qPCR positive, but RADT-negative individuals do not pose a high risk of transmissions, since all samples remained culture negative in our experimental setup. However, individual results must be interpreted with caution, as SARS-CoV-2 infection could remain undetected in early stages. In summary, our results suggest that SARS-CoV-2 transmission could be reduced by systematic RADT, despite the fact that some infected individuals will not be detected by the test (51). Simple to perform and applicable anywhere, RADT enable mass testing as a complementary method to RT-qPCR to more effectively combat the SARS-CoV-2 pandemic.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 0.7 MB.