key: cord-0897483-ir1ni5mb
authors: Hirotsu, Yosuke; Maejima, Makoto; Shibusawa, Masahiro; Amemiya, Kenji; Nagakubo, Yuki; Hosaka, Kazuhiro; Sueki, Hitomi; Hayakawa, Miyoko; Mochizuki, Hitoshi; Tsutsui, Toshiharu; Kakizaki, Yumiko; Miyashita, Yoshihiro; Omata, Masao
title: Prospective Study of 1,308 Nasopharyngeal Swabs from 1,033 Patients using the LUMIPULSE SARS-CoV-2 Antigen Test: Comparison with RT-qPCR
date: 2021-02-05
journal: Int J Infect Dis
DOI: 10.1016/j.ijid.2021.02.005
sha: 2dff760c3aeb1ce4664596f8efa4d21087250daa
doc_id: 897483
cord_uid: ir1ni5mb

Background Reverse-transcription PCR (RT-PCR) is the gold standard for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) detection. Previously, we demonstrated the accuracy of the quantitative LUMIPULSE SARS-CoV-2 antigen test using samples collected retrospectively. Here, the antigen test was clinically validated using prospective samples. Methods A total of 1,033 nasopharyngeal swab samples were collected 1,033 individuals and an additional 275 follow-up samples were collected from 43 patients who later tested positive for COVID-19. All 1,308 samples were subjected to quantitative RT-PCR (RT-qPCR) and the LUMIPULSE antigen test. Antibody response was investigated for patients with discordant results to clarify whether seroconversion had occurred. Results RT-qPCR identified 990 samples as negative and 43 as positive, while the antigen test identified 992 as negative, 37 as positive, and 4 as inconclusive. The overall concordance rate was 99.7% (1,026/1,029). The sensitivity, specificity, positive predictive value and negative predictive value of the antigen test were 92.5% (37/40), 100% (989/989), 100% (37/37) and of 99.7% (989/992), respectively, after excluding the four inconclusive results. The kappa coefficient was 0.960 (95% confidence interval, 0.892‐0.960), suggesting excellent agreement between the two tests. The seropositive in five out of the seven patients with discordant results suggested that the discrepancy was caused by samples collected during the late phase of infection. Using the follow-up samples, we observed a correlation between the antigen level and the viral load or threshold cycle (Ct) value. The concordance rate between these test results tended to be high among samples collected up to 9 days after symptom onset but this gradually decreased thereafter. Conclusions This prospective study demonstrated that the LUMIPULSE antigen test is a highly accurate diagnostic for SARS-CoV-2.

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has spread rapidly around the world. Approximately 80% of infected patients with coronavirus disease 2019 (COVID- 19) develop mild symptoms and recover without specific treatment [1] . However, 20% of patients deteriorate rapidly within 7 to 10 days of symptom onset and 25% of these patients will face mechanical ventilation and high mortality rates [1] .

The reverse-transcription polymerase chain reaction (RT-PCR) is a sensitive and specific assay that is considered the gold standard for SARS-CoV-2 testing [2, 3] . However, RT-PCR requires specialized equipment and skilled technicians, and it is time-consuming and expensive.

As an alternative, the antigen test has been approved for and is currently in clinical use for the diagnosis of COVID-19 patients [4] .

Most antigen tests are developed for use as a rapid diagnostic at the point of care. These tests are based on paper assays or lateral flow immunochromatography and provide qualitative detection of SARS-CoV-2. The simple-to-use format of the rapid antigen test does not require special equipment or operator skills. In general, these rapid kits have high specificity but low sensitivity, which can yield false-negative results for samples containing low viral loads [5] [6] [7] [8] [9] . For instance, the rapid antigen test sensitivity is 30.2% (32/106) for the COVID-19 Ag Respi-Strip (Coris BioConcept) [5] , 45.7% (16/ J o u r n a l P r e -p r o o f (44/60 and 43/55) for the Panbio COVID-19 Ag Rapid Test (Abbott) [8, 9] .

The World Health Organization (WHO) recommended that antigen rapid diagnostic tests are used for symptomatic individuals within the first 5-7 days following the onset of symptoms, but should not be used for individuals without any symptoms [10] . There is a growing demand for the clinical utility and high accuracy of the quantitative antigen test to be demonstrated. However, few prospective validation studies have been reported for large cohorts.

We previously evaluated the accuracy of the LUMIPULSE SARS-CoV-2 antigen test (Fujirebio, Tokyo, Japan), which is a fully automated system based on the chemiluminescent enzyme immunoassay (CLEIA) principle [11] . This antigen test can quantitatively measure the antigen level of SARS-CoV-2 nucleocapsid protein. The antigen test is approved in Japan and widely used by hospitals, clinical laboratory centers, and airport quarantines. Fujirebio Europe acquired CE marking in August 2020 and have begun to provide the antigen test globally (e.g., at German airports) [12, 13].

Here, we performed a prospective validation study of the LUMIPULSE antigen test using a total of 1,308 nasopharyngeal swab samples. Of these, 1,033 were initial samples collected prospectively from 1,033 individuals. In addition, 275 follow-up samples were longitudinally collected from 42 COVID-19-confirmed patients identified in this cohort. The antigen test and quantitative RT-PCR (RT-qPCR) were conducted for each sample. We also examined the antibody response in seven patients with discordant results to clarify their seroconversion status.

J o u r n a l P r e -p r o o f

A total of 1,308 nasopharyngeal swab samples were collected from 1,033 individuals.

First, we prospectively analyzed 1,033 samples from 1,033 individuals including symptomatic individuals (i.e., those with a fever, cough, sore throat, fatigue, and/or headache), asymptomatic individuals who had contact with an infected patient, and returnees from abroad. Among these 1,033 individuals, we confirmed 43 PCR-positive individuals (hereafter termed "COVID-19

patients") including 36 symptomatic and 7 asymptomatic patients. Second, we longitudinally 

All nasopharyngeal swab samples were collected using cotton swabs and placed in 3 mL of viral transport media (VTM) obtained from Copan Diagnostics (Murrieta, CA, USA), and 700 μL of the VTM were used for the antigen test immediately after sample collection. The residual VTM was temporarily stored at 4°C and 200 μL of the VTM were used for nucleic acid extraction within 2 hours after sample collection.

The sample antigen levels were determined quantitatively with the LUMIPULSE SARS-CoV-2 Ag test (Fujirebio, Inc., Tokyo, Japan) according to the manufacturer's instructions [11, 14] .

In brief, 700 μL of the VTM samples were briefly vortexed, transferred into a sterile tube, and centrifuged at 2,000 ×g for 5 min. Aliquots (100 μL) of the supernatant were used for testing on the LUMIPULSE G600II automated system (Fujirebio). For samples with an antigen level > 5,000 pg/mL, the samples were diluted with the kit diluent and re-tested, and the antigen level was calculated taking the dilution factor into account. Samples with an antigen level ≥ 10 pg/mL were considered positive, samples with ≥ 1.0 pg/mL and < 10.0 pg/mL antigen were labeled inconclusive, while a result of ≤1.0 pg/mL was considered negative as per the manufacturer's guidelines. Prime System (Thermo Fisher Scientific) as we previously described [15, 16] . Briefly, we added 200 µL of VTM, 5 µL of proteinase K, 265 μL of binding solution, 10 μL of total nucleic acid-binding beads, 0.5 mL of wash buffer, and 0.5-1 mL of 80% ethanol to each well of a deep-well 96-well plate. The nucleic acids were eluted with 70 μL of elution buffer. The total nucleic acids were immediately subjected to RT-qPCR.

According to the protocol developed by the National Institute of Infectious Diseases (NIID) in Japan [3, 15, 17] , we performed one-step RT-qPCR to detect SARS-CoV-2. This PCR amplifies A threshold cycle (Ct) value was assigned to each PCR reaction and the amplification curve was visually assessed. According to the national protocol (version 2.9.1) [17] , we deemed a sample to be positive when a visible amplification plot was observed, whereas a sample was deemed negative when no amplification was observed.

The absolute copy number of viral loads was determined using serial diluted DNA control targeting the N gene of SARS-CoV-2 (Integrated DNA Technologies, Coralville, IA) as previously described [3] . The limit of detection of RT-qPCR using the primer/probe was considered as 2 copies according to the previous report [17] .

To clarify the discordant results observed between RT-qPCR and the antigen test in seven individuals, we measured the pan-immunoglobulin level against the full-length of SARS-CoV-2 nucleocapsid recombinant protein expressed in Escherichia coli. The serum samples were subjected to the Elecsys Anti-SARS-CoV-2 test (Roche Diagnostics, Basel, Switzerland) on the cobas 8000 automated platform (Roche Diagnostics) [18] . This assay utilizes the J o u r n a l P r e -p r o o f electrochemiluminescence immunoassay (ECLIA) principle. Samples with a cut-off Index (COI; electrochemiluminescent signal of the test sample/cut-off value of the calibration sample) < 1.0 were considered negative, while those with a COI ≥ 1.0 were considered positive.

Sensitivity, specificity, positive predictive value and negative predictive value were calculated when the results of RT-qPCR were considered as reference and inconclusive results with antigen test were excluded. Student's t-test was calculated to determine significant differences among the groups, and p-value < 0.05 was considered statistically significant. Cohen's kappa coefficient of results between the two tests with 95 % confidence intervals (CI) was calculated using the R statistical package (version 3.1.2) (http://www.r-project.org/). Cohen's kappa values greater than 0.81 were interpreted to indicate almost perfect agreement [19] .

We collected a total of 1,308 nasopharyngeal swab samples (1,033 initial and 275 followup samples) from 1,033 individuals. Each sample was subjected to both RT-qPCR and the LUMIPULSE SARS-CoV-2 antigen test.

Among the 1,033 initial samples, RT-qPCR identified 990 as negative and 43 as positive J o u r n a l P r e -p r o o f (symptomatic, n = 36; asymptomatic, n = 7; Figure 1 ). The antigen level of RT-qPCR-positive patients was significantly higher than that of RT-qPCR negative patients (p = 0.78x10 -31 , Student's t-test) (Figure 1) . The mean antigen level was -0.7 log10 pg/mL (range, -2.0 to 0.4 log10 pg/mL) for the RT-qPCR-negative patients and 4.4 log10 pg/mL (range, 0.7 to 5.5 log10 pg/mL) for the RT-qPCR-positive patients (Figure 1 ). There was no significant difference in the antigen levels between the symptomatic and asymptomatic patients (p = 0.36, Student's t-test). To determine the accuracy of the antigen test, we compared the results from RT-qPCR with those from the antigen test for each sample. The antigen test identified 992 as negative, 37 as positive, and 4 as inconclusive (Table 1) . Thus, the rate of inconclusive results with the antigen test was 0.4% (4/1,033). 

There were seven discordant results (0.7%, 7/1,033) between the RT-qPCR and the antigen test (Table 2 ). Of these, three were negative by the antigen test and positive by RT-qPCR (Cases #1-3), three were inconclusive by the antigen test and positive by RT-qPCR (Cases #4-6), and one was inconclusive by the antigen test and negative by RT-qPCR (Case #7).

Cases #1-3 were returnees from overseas (two from Brazil and one from India) who tested positive by RT-PCR at the airport quarantine in Japan. Cases #4-7 lived in our district and were admitted to our hospital. Case #7 was found in cardiopulmonary arrest at home and was emergently admitted to our hospital but passed away shortly afterward.

At the beginning of hospitalization, the antibody level showed positive in Cases #1-5 but the viral loads were low (range, 1.0-2.2 log10 copies/L), indicating that these patients were seropositive for SARS-CoV-2 at the late phase of infection (Table 2) . Conversely, the antibody response had likely only begun for Case #6, as we previously reported [20] (Table 2) .

In this prospective study, a total of 318 nasopharyngeal swab samples were collected from 43 COVID-19 patients. These 318 samples included 43 initial samples and an additional 275 follow-up samples collected during hospitalization. Out of 318 samples, 43 initial samples were already analyzed aforementioned in Figure 1 , and these data were used for the following analysis.

Consistent with our previous report [11] , the viral load determined by RT-qPCR and the antigen level were correlated for both the initial samples (R 2 = 0.835) and the follow-up samples (R 2 = 0.773) (Figure 2A) . Similarly, there was a correlation between the Ct value and antigen level in both the initial samples (R 2 = 0.884) and follow-up samples (R 2 = 0.774) ( Figure 2B ). The coefficient of determination for the initial samples was slightly higher than that for the follow-up samples. These results suggested that the variability increased in follow-up samples because these samples includes lower viral load samples collected from hospitalized patients who were in late phase of infection or recovery. Positive results were observed in a high proportion for both the antigen test (71%-98%) and RT-qPCR (94-100%) for samples collected within 0-9 days of onset ( Figure 3A) . However, the positive result ratio of the antigen test declined over 10-30 days after onset ( Figure 3A) . Overall, the positive result ratio of RT-qPCR was higher than that of the antigen test over the entire observation period ( Figure 3A ). The concordance ratio between RT-qPCR and the antigen test was high (95%-100%) among samples collected within 0-9 days of onset, low (50%-64%) within 10-24 days of onset, and slightly increased (67%-80%) within 25-30 days ( Figure 3B ).

Among the 250 follow-up samples from the 36 symptomatic patients, the overall concordance rate was 80.6% (162/201), the sensitivity was 76.1% (124/163), and the specificity was 100% (38/38) ( Table 3) . Sensitivity was high (94.4%-100%) for the samples collected within 0-9 days of symptom onset; however, this gradually declined after 10 days. Notably, the specificity was 100% throughout the observation period (Table 3 ).

In this study, we prospectively validated the performance of the LUMIPULSE SARS-CoV-2 antigen test. Compared with RT-qPCR, the accuracy of the antigen test was high when the test was conducted using initial samples from 1,033 patients who visited the hospital. The sensitivity of the LUMIPULSE antigen test (92.5%) was higher than that of a conventional rapid antigen test (approximately 30%-70%) [5] [6] [7] [8] [9] . The positive detection rate of the antigen test decreased gradually in follow-up samples collected from COVID-19 hospitalized patients. To our knowledge, this is the first longitudinal and prospective study to provide real-world data illustrating the clinical validity of antigen tests for COVID-19 screening.

In 1,033 individuals, three (0.3%) were judged as positive by RT-qPCR and negative by antigen test (Case #1-3 in Table 2 ). In this study, we revealed that these false negative results were attributed to the samples with very low viral load obtained from seropositive patients.

Furthermore, in hospitalized patients, antigen test tended to judge negative after 10 days of onset of symptom, but RT-qPCR often remained to be positive ( Figure 3 ). Consistently, we previously observed discordant results in samples collected from a persistent viral-shedding patient [14] . These results suggested the viral load was low and protein translation was likely to be attenuated in host cells. We consider that assessing the immune response with an antibody test is useful for interpreting the discrepant results [20] . In the sample with antigen-negative and RT-PCR-positive, we need to carefully examine the two possibilities: namely, sample is collected from patient in late or recovery phase, and in extremely very early phase of infection.

False-positive results are a burden to both patients and healthcare workers, necessitating patient quarantine and surveys of sometimes up to 100 individuals who had close contact with these patients. To prevent misleading false-positive results, a highly specific test is needed. Previously, a case report showed a false-positive result with the LUMIPULSE antigen test [21] . However, the percentage of false positive result is 0.3% (1/301) according to the data of the kit's package insert, suggesting that the LUMIPULSE antigen test yields a robust result. In the present study, the specificity was 100% for both the initial samples (989/989 samples) and follow-up samples (38/38 samples). We previously encountered fluctuating results when using viscous samples in our preliminary study but resolved this issue by sufficiently centrifuging the samples and using the supernatants for the antigen test (data not shown). It is hoped that further scrutiny of false-positive samples will lead to increased accuracy regardless of the nature of the sample.

There are accumulating data on SARS-CoV-2 infectivity [22] . RT-PCR yields positive results for a long duration (6-7 weeks following infection) even in the presence of an extremely low viral load [14, 23] . However, RT-PCR cannot directly indicate the presence of viable and infectious SARS-CoV-2. In vitro studies have revealed that the infectivity of SARS-CoV-2 is maintained in clinical samples for only around 8-10 days after the onset of symptoms [24] [25] [26] [27] [28] [29] .

Therefore, RT-PCR-positive results can reflect the presence of non-infectious viral "debris" in samples collected several weeks after symptom onset or recovery. Notably, we observed that the J o u r n a l P r e -p r o o f rate of positive results with the antigen test rapidly declined around 9 days after disease onset.

Based on in vitro studies, the timing of the drop in antigen levels appears to mark the point when the levels of infectious virus particles diminish [24] [25] [26] [27] [28] [29] . Further studies using cell-based and nonhuman primate models are required to clarify the relationship between antigen levels and virus infectivity [24, 30] and to investigate whether monitoring the antigen level will help determine the length of quarantine needed, likely response to treatment, and the timing of hospital discharge.

A possible limitation of this study is that the portability of the LUMIPULSE antigen test.

Although it is easier to conduct the antigen test than RT-qPCR, it needs the special equipment and centrifugation step. Therefore, it is difficult to test outside of the clinical laboratory. There is also a serious concern about how to manage this test in low-income countries. It would be necessary to develop tests that can be applied to a wide range of circumstance by making the equipment more accessible and improving the protocol.

In conclusion, the LUMIPULSE SARS-CoV-2 antigen test system can be easily applied 

YH contributed to study design, data collection, data analysis and writingreview & editing.

MM, MS, KA, YN, KH, and HS contributed to sample preparation and data collection, data analysis.

MH and HM contributed to supervision.

TT, YK and YM contributed to provide resources.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

J o u r n a l P r e -p r o o f J o u r n a l P r e -p r o o f 

How does SARS-CoV-2 cause COVID-19?

Detection of 2019 novel coronavirus (2019-nCoV) by realtime RT-PCR

Double-quencher probes improve detection sensitivity toward Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) in a reversetranscription polymerase chain reaction (RT-PCR) assay

Rodriguez-Villalobos H: the first 7 days after the onset of symptoms

Field evaluation of a rapid antigen test (Panbio COVID-19 Ag Rapid Test Device) for COVID-19 diagnosis in primary healthcare centers

Comparison of Automated SARS-CoV-2 Antigen Test for COVID-19 Infection with Quantitative RT-PCR using 313 Nasopharyngeal Swabs Including from 7 Serially Followed Patients

Analysis of a persistent viral shedding patient infected with SARS-CoV-2 by RT-qPCR, FilmArray Respiratory Panel v2.1, and antigen detection

Pooling RT-qPCR testing for SARS-CoV-2 in 1000 individuals of healthy and infection-suspected patients

Environmental cleaning is effective for the eradication of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) in contaminated hospital rooms: A patient from the Diamond Princess cruise ship

Development of Genetic Diagnostic Methods for Novel Coronavirus 2019 (nCoV-2019) in Japan

Development and Validation of the Elecsys Anti-SARS-CoV-2 Immunoassay as a Highly Specific Tool for Determining Past Exposure to SARS-CoV-2

The Measurement of Observer Agreement for Categorical Data

The dynamic change of antibody index against Covid-19 is a powerful diagnostic tool for the early phase of the infection and salvage PCR assay errors

Another false-positive problem for a SARS-CoV-2 antigen test in Japan

Duration of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) Infectivity: When Is It Safe to Discontinue Isolation?

Prolonged Persistence of SARS-CoV-2 RNA in Body Fluids

Virological assessment of hospitalized patients with COVID-2019

Viral RNA load as determined by cell culture as a management tool for discharge of SARS-CoV-2 patients from infectious disease wards

Early treatment of COVID-19 patients with hydroxychloroquine and azithromycin: A retrospective analysis of 1061 cases in

Predicting infectious SARS-CoV-2 from diagnostic samples

Shedding of infectious virus in hospitalized patients with coronavirus disease-2019 (COVID-19): duration and key determinants

SARS-CoV-2 Virus Culture and Subgenomic RNA for Respiratory Specimens from Patients with Mild Coronavirus Disease

Respiratory disease in rhesus macaques inoculated with SARS-CoV-2

We thank Shintaro Yagi and Satoshi Kojima (Fujirebio) for technical discussion and all of the medical and ancillary hospital staff and the patients for consenting to participate. We also thank Natasha Beeton-Kempen, Ph.D., from Edanz Group (https://en-author-services.edanz.com/) for editing a draft of this manuscript.