SARS‐CoV‐2 rapid antigen test: Fast‐safe or dangerous? An analysis in the emergency department of an university hospital

Abstract The use of Antigen point of care tests (AgPOCT) might be an essential tool to fight the coronavirus disease 2019 (COVID‐19) pandemic. Manufacturer information indicates a specificity of about 95% and there is a growing interest to use these tests area‐wide. Therefore, it is necessary to clarify whether AgPOCT can be used safely for “rule‐in” (detection of positive patients) and for “rule‐out” (valid negative testing). Two thousand three hundred and seventy‐five patients received polymerase chain reaction (PCR) testing and AgPOCT for severe acute respiratory syndrome coronavirus 2 (SARS‐CoV‐2) regardless of symptoms. The positive predictive value of symptomatic and asymptomatic patients was compared with a cut‐off threshold cycle (C t) value of ≤30 and in total. Five hundrded and fifty‐one patients tested positive for the SARS‐CoV‐2 virus by PCR, of whom 35.2% presented without symptoms. In all patients, regardless of their symptoms or C t values, a sensitivity of 68.9% and a specificity of 99.6% were calculated for AgPOCT. In patients with C t values ≤30, a sensitivity of 80.5% (95% confidence interval: ±1.62) and a specificity of 99.6% were shown for all tests (symptomatic/asymptomatic). Highly infectious patients (C t ≤ 20), regardless of symptoms, were reliably detected by the AgPOCT. In infectious patients with C t values ≤30, the test has a sensitivity of about 80% regardless of COVID‐19 typical symptoms, which is apparently less than the 96.52% specificity indicated by the manufacturer. Relevant improvement in test sensitivity by querying the patients who are symptomatic and asymptomatic is also not feasible. We strongly suggest that we critically question the use of AgPOCT for “rule‐out,” as they only provide a supposed safety.


| INTRODUCTION
The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic continues to obsess the world. Different strategies from "zero-covid" with lockdown tightening up to a "life with the virus" are currently discussed. In the context of the so-called "second wave" of the pandemic with the SARS-CoV-2 virus (order Nidovirales, family Coronaviridae, genus Betacoronavirus, Species Severe acute respiratory syndrome-related coronavirus), the question arises if rapid detection of infected persons, including asymptomatic ones, is possible. 1,2 In addition to the gold standard, the reverse transcription polymerase cahin reaction (RT-PCR), a variety of antigen tests (so-called "antigen point of care tests," AgPOCT) are available. 3 So far, these tests can be considered as an additive measure if standards are met. 4,5 The current literature shows weak documentation of the application of the test strategy in practice. Many validations are still in preprint 6 or include smaller cohorts. 7 There is an ongoing discussion for an antigen rapid-test strategy to enable parts of "common" life, including, for example, tests at airports, nursing homes, and major events. In addition, hospitals are a central part of this discussion. Due to an increasing number of patients in the emergency room and scarce resources, both in terms of personnel and PCR testing, the question comes up if AgPOCT is an effective screening tool to initiate treatment of SARS-CoV-2 infection, accelerate de-isolation of patients who do not have SARS-CoV-2 infection or prevent an outbreak in an inpatient setting. In hospitals and nursing homes, outbreaks not only pose organizational problems in a confined space but also pose serious health risks for a large proportion of patients and residents, as a more severe infection course is to be expected due to the risk profile. The aim of this study is to show whether AgPOCT is reliable and whether it could be used as the sole measure in the practice environment of a hospital and consecutively in public life not only for rule-in (detection of positive patients), but also for rule-out (valid negative testing).

| Design and participants
We retrospectively analyzed data from patients who presented for acute complaints in an emergency department of a German university hospital between October 16, 2020 and January 14, 2021.
The general test strategy for detecting SARS-CoV-2 infection during this period involved testing each patient in the emergency department without and with SARS-CoV-2 symptoms (breathing difficulties, temp. >38°C and/or chills; cough, ansomy, ageusie, sore throat, fatigue, headache, limb pain, colds, gastrointestinal symptoms [nonspecific abdominal discomfort, diarrhea, vomiting]). Each patient received a nasopharynx smear according to the manufacturer's information for the performance of a PCR test and a simultaneous AgPOCT (Standard Q; Roche). We compared the positive and false negative tested patients based on their positive predictive value and set the cut-off at a C t value of 30. From the literature and the general approach, subjects with a C t value below 30 are considered infectious. We focused on the correlation between CT value less than 30 and test results of AgPOCT and PCR testing.

| Implementation
The RT-PCR tests used were carried out on the basis of the manufacturers using Alinity in the SARS-CoV-2 assay (Abbott), Cepheid (GeneXpert), and RealStar SARS-CoV-2 RT-PCR (Altona Diagnostics).
The AgPOCT used was the standard Q (Roche). According to the manufacturer of the antigen test, a nasopharynx swab was performed over a nostril 7 with repeated slight rotation after reaching resistance at the area of the posterior nasopharynx. The swab was then inserted into an extraction buffer tube where the liquid from the swab was extracted. Finally, the sample was applied to the test device and the result was read after 15-30 min. The collecting of the sample material for PCR testing was done in the same way via nasopharyngeal swab. Material collecting for both antigen and PCR testing was always carried out in patients by the same qualified nurse or physician.

| Data collection
For this study, data from patients who received an AgPOCT in the emergency department were compared with PCR diagnostics performed in the emergency department and the Institute of Virology.
In addition, the patient's history was used to ask about symptoms of SARS-CoV-2 infection. The patient history was evaluated according to symptoms (e.g., respiratory/thoracic, gastrointestinal, and general flu-like symptoms, as well as additional symptoms, such as loss of smell and taste) and duration of symptoms. The data performed by the Institute of Virology included the type of testing, the manufacturer, and the determined C t value. To achieve comparability of the C t values of the different manufacturers, the testing was stan-

| Statistics
We created four-field panels to calculate the specificities, sensitivities, and negative predictive values. For all continuous variables, the mean values ± standard deviation and the 95% confidence intervals were reported. Study prevalence was used to calculate negative predictive value. The calculations were performed using the scientific statistical modules scipy, pandas, and numpy of the programming language Python (version 3.0).
The representations were created using the matplotlib module of the programming language Python. 5324 | HOLZNER ET AL.

| Ethics
This study was approved by the ethics committee of the faculty of medicine of the University of Duisburg-Essen on December 17, 2020.
The study number is 20-9769-BO. ±0.0660) and for C t values greater than 30 a sensitivity of 12.63%

| 5325
Assuming infectivity and thus the potential risk of transmission to others at C t values below 30 11,12 it is worthwhile to take a closer look at the intervals and take them into account in the evaluation. Overall, the antigen test is able to detect positive patients with a sensitivity of 80.48% (95% CI: ±1.62) within patients who had a positive PCR with a C t value of ≤30. Thus, one in five infectious patients is not recognized, with a specificity of 99.56%. A more representative examination outside a hospital should be considered to be able to make better statements about the test reliability of asymptomatic people, since like in this study, more symptomatic than asymptomatic patients are present in an emergency department.
Within the hospital, the strategy has changed on the basis of this data to ensure that only patients with negative PCR are de-isolated and therefore an antigen test is not sufficient. There must be no risk that one in five infectious patients reaches an inpatient setting undetected and thus an outbreak is risked.
In our opinion, the same applies in other areas of life, the regular "free-testing" is currently being discussed, which should thus enable people to access nursing homes and other care facilities, airports, major events, etc. Due to the low sensitivity in asymptomatic people, large-scale testing using this antigen rapid test method is certainly epidemiologically useful to recognize infected persons, but must not lead to a reduction of infection protection measures with a negative rapid test. In our view, this requires significant information measures for users and decision-makers alike in order not to create new risks.

| Limitations
This study retrospectively analyzes a collective of patients presented in a hospital. Therefore, the numbers of asymptomatic patients are not representative of the overall infections. Due to the sample collection by a large number of qualified staff, pre-analytical factors must be considered. However, the same employee has taken the two tests compared and this setting reflects the real everyday situation of extensive tests, in which there are also no perfect laboratory conditions.
A second restriction is that too broad and loose periods of less than 7 and greater than 14 days were chosen for the documentation of the onset of symptoms. As a result, the analysis cannot relate to meaningful shorter periods of time and has been omitted

| CONCLUSION
We conclude that the AgPOCT shows its strength in the reliable detection of highly infectious (C t value to 20) patients, regardless of whether they have symptoms of SARS-CoV-2 infection or not. The higher the C t value, detached from the symptoms, the less sensitive is the AgPOCT. An improvement in test sensitivity is also not relevant by distinguishing or querying the patients in symptomatic and asymptomatic. In both groups, similarly, low sensitivity is shown, although this is even lower among asymptomatic patients (95% CI: 62.0 ± 0.32) than in symptomatic patients. In all infectious patients regardless of COVID-19-like symptoms, defined as patients with a C t -value of ≤30, the AgPOCT achieves a sensitivity of 79.7% and therefore is clearly lower than the sensitivity reported by the man- enables the rapid and easy detection of approximately two-thirds of infectious patients in the context of large-scale screening, even without or with only minor nonspecific symptoms, which otherwise go undetected. However, this enormous epidemiological advantage must be supplemented by an informed and responsible handling of negative test results.

CONFLICT OF INTERESTS
The authors declare that there are no conflict of interests.

AUTHOR CONTRIBUTIONS
Carola Holzner, Dirk Falk, and Maximilian Pabst contributed to statistic analysis, data interpretation, writing. Olympia Anastasiou, David Fistera, Randi Manegold, and Joachim Risse contributed to data acquisition. Clemens Kill and Ulf Dittmer contributed to proofreading and supervision.

PEER REVIEW
The peer review history for this article is available at https://publons. com/publon/10.1002/jmv.27033

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available on request from the corresponding author, (Dr. med. Carola Holzner).
The data are not publicly available due to patient privacy (European general data protection regulation).