Study design
For both studies described here, a prospective collection of randomly selected upper respiratory specimens sent to the microbiology laboratory were used. Samples were obtained from different COVID-19 suspected patients (emergency room patients (ERP), nursing home patients (NHP) and health care workers (HCW) in the periods from October 6 to October 12 (study 1) and October 24 to November 15, 2020 (study 2), respectively. During the study period the seven-day average of newly confirmed RT-qPCR positive COVID-19 cases was 23%. The study was conducted in a teaching hospital in Rotterdam, the Netherlands.
Sample collection and storage
Patients and HCW with suspected COVID-19 infection were sampled by collection of a combined throat nasopharyngeal swab. After this, swabs were placed in 3 ml Virus Transport Medium (VTM), and stored at 4⁰C until sample preparation after which positive samples were stored at -20⁰C. All specimens were examined for SARS-CoV-2 viral RNA by routine RT-qPCR on the day of collection and for antigen detection by three of five LFAs within 72 hours and by two LFAs one month after collection.
Study 1 (Comparison of five rapid antigen tests)
The first study involved a comparison of five different COVID-19 rapid antigen tests for the detection of SARS-CoV-2 viral antigens. A total of 40 RT-qPCR negative and 40 RT-qPCR positive samples were prospectively selected from October 6 to October 12 2020. In the first part, three LFAs were evaluated: Certest SARS-CoV-2 (Certest Biotec S.L., Spain), Roche SARS-CoV-2 Rapid Antigen Test (Roche, Switzerland) and Romed Coronavirus Ag Rapid Test (Romed, The Netherlands). In the second part, three LFAs were evaluated with 40 RT-qPCR positive samples, 35 samples stored at -20⁰C from the first part, completed with five RT-qPCR positive samples with corresponding Ct values of the missing samples: BD Veritor SARS-CoV-2 point-of-care test (Becton, Dickinson and Company, USA), PanbioTM COVID-19 Antigen rapid test (Abbott, USA) and Romed Coronavirus Ag Rapid Test (Romed, The Netherlands).
Study 2 (Romed - RT-qPCR comparison)
In the second prospective study the clinical performance of the best performing LFA was compared to RT-qPCR in an extended cohort of patients and HCW which were selected between October 24 and November 15, 2020. A total of 900 patients were included, 300 for each of the three defined groups of ERP, NHP and HCW.
Detection of viral RNA by direct RT-qPCR methods
Samples from NHP and HCW were tested on two different RT-qPCR methods by either a validated in-house RT-qPCR assay or on the ELITe InGenius® (Elitech, France) platform [3]. Samples from ERP were tested with the GeneXpert Xpress SARS-CoV-2 PCR assay (Cepheid Inc, Sunnyvale, USA) according to the instructions of the manufacturer.
Detection of SARS-CoV-2 viral antigen by LFAs
For antigen extraction 350 μl of VTM was added to 300 μl of each respective extraction buffer and mixed for 10 seconds. Subsequently, a number of drops of the mixture was added to the sample port of the antigen assay according to the instructions of the manufacturer. The result was read visually after 15 minutes whereby any shade of color in the test line region was considered positive. All tests were independently assessed by two investigators who were blinded to all other test results and in case of discrepancy an additional assessment was performed by a third investigator.
Ethical statement
The Institutional Review Board waived the need for informed consent because tests were performed on samples that had been required for routine microbiological investigation (IRB protocol number 2020-109). Also according to hospital procedure all patients were informed about the possibility of an opt-out if they had objections against the use of left-over material for research to improve or validate diagnostic testing procedures. The study was performed in accordance with Helsinki Declaration as revised in 2013.
Data collection and statistical analysis
The primary outcome measures for both studies were clinical specificity and clinical sensitivity in relation to different CT values of the RT-qPCR. For the second study positive predictive value (PPV) and negative predictive value (NPV) were also calculated as secondary outcomes in order to develop a diagnostic algorithm in different patient groups. All data were analyzed using Microsoft Excel, GraphPad Prism version 8 and R version 3.3.2 (R Foundation for Statistical Computing). Groups were compared by using non-parametric tests for continuous variables and chi-square test or Fisher’s exact test for categorical variables as appropriate. Values of p that were <0.05 were considered to be statistically significant.