SARS-CoV-2 rapid test versus RT-qPCR on noninvasive respiratory self-samples during a city mass testing campaign.

A few months ago, Baro et al. published in this Journal an interesting study comparing different rapid lateral-flow tests (RLFT) dedicated to the direct diagnosis of SARS-CoV-2 in nasopharyngeal aspirates in a large population of asymptomatic subjects in Catalonia, Spain. They showed that the analytical performances of some of them were compatible with their use in routine, notably with an objective of mass screening.¹ In accordance with these observations, we conducted in February 2021 a city-scale screening to evaluate the prevalence of SARS-CoV-2 infection in the town of Saint-Etienne (the 14th most populated city in France in 2018 with 173,000 inhabitants) where the circulation of the virus had exhibited the highest incidence of positive cases in France in October-November 2020. By contrast to other mass testing campaigns,1, 2, 3, 4, 5, 6, 7 the choice was done of noninvasive tests based on salivary specimens. This vast operation led to the inclusion of more than 7000 individuals within a single week, which represents approximately 4% of the whole indoor population of Saint-Etienne.

We report the performance of the COVID19 Speed-Antigen Test (Biospeedia, France) on self-sampled material combining anterior nose and saliva, by comparison to RT-qPCR on saliva, in the context of this campaign. Data regarding acceptability of the different sampling methods were recorded by optional questionnaire. Adults and children over 10 years of age able to provide a self-sample were eligible for inclusion, after oral approval of an informative note by the subject and/or his/her guardian. Participants provided information on demographic characteristics, occupation and symptoms (protocol IDRCB number 2021-A00390–41). They were invited optionally to complete an on-line questionnaire (Limesurvey® software, Hamburg, Germany) including scales in relation with the acceptability of the different modes of specimen collection.

Participants were asked to collect at least 3 ml of saliva in a 50 ml sterile plastic tube more than 30 min after drinking, eating, smoking, brushing teeth or rinsing teeth. This sample was used for RT-qPCR analysis that served as gold standard in this study. In parallel, 2 drops of saliva were transferred into a second tube previously filled with 8 drops of the buffer of the RFLT kit. An anterior nasal sample was then performed by the participant. The content of this swab was then discharged in the tube containing saliva and buffer. This mixture was used to perform the COVID19 Speed-Antigen Test (Biospeedia, France) by the assistants specially trained for this technique on the different sites of the mass testing campaign, according to the manufacturer's instructions.

The RT-qPCR tests were performed in three PCR independent platforms located at the University Hospitals of Lyon (4278 specimens) and Clermont-Ferrand (1291 specimens) and at the bioMérieux Research Laboratory of Grenoble (1068 specimens). Invalid tests were submitted to a second round of extraction/amplification; in case of two successive failures, the result was declared invalid. As different RT-qPCR tests were used, a correction was done to homogenate the cycle threshold (C_T) values according to the concordance table of the French Society of Microbiology.⁸ Positive samples were typed by the Lyon laboratory for the identification of variant strains of SARS-CoV-2.

From February 22 to 28 of 2021, 7020 subjects participated to the mass testing campaign organized in Saint-Etienne but 381 individuals were not included, mainly for young age (N = 132) or missing specimens or data (N = 249). From the 6639 included volunteers, 56% were female and the median age was 48 years (interquartile range: 30–64); 90.5% of participants were asymptomatic with regard to SARS-CoV-2 infection.

Overall, 165 participants were tested positive by RT-qPCR and 114 by RLFT. The number of positive specimens by RLFT was almost significantly lower in asymptomatic than in symptomatic individuals (P = 0.05). The sensitivity, specificity, positive predictive value and negative predictive value of RLFT were 59.4 (CI₉₅: 51.5–67.0), 99.8 (CI₉₅: 99.6–99.9), 86.0 (CI₉₅: 78.2–91.8) and 99.0 (CI₉₅: 98.7–99.2), respectively, by reference to RT-qPCR. The sensitivity of RLFT dropped to 48.6% (CI₉₅: 36.9–60.6) when only asymptomatic participants were considered. This latter sensitivity was close to that observed with the best kits evaluated by Baro et al.¹ and a bit superior to that observed with similar commercial tests in Liverpool (40.0%),² in the Pima county of Arizona (35.8%)³ or in Kuwait City (42.3% with the first RFLT and 30.6% with the second one).⁴

The 165 strains recorded by RT-qPCR were submitted to molecular typing: 13 were not typeable, 85 belonged to the original phenotype whereas 68 exhibited a variant of concern (VOC) profile (66 Alpha and 2 Beta or Gamma). The median C_T value was significantly higher in variant than in non-variant isolates (21.8 vs. 25.7, P < 0.001 by Wilcoxon-Mann-Whitney test).

A strong inverse correlation (Kendall's τ coefficient of −0.910; P < 10⁻⁸) was observed between the C_T value measured by RT-qPCR and RLFT sensitivity. The sensitivity dropped from 100% to 15.4% for C_T values ≤ 18 and >28, respectively (Fig. 1 ). Three variables were associated to RLFT positivity: low C_T value, detection of SARS-CoV-2 variant and presence of Covid-19 symptoms; however, after multivariate analysis, the C_T value remained the sole explicative variable (Table. 1 ).

Fig. 1.

Sensitivity of the rapid lateral-flow test with reference to RT-qPCR taken as gold standard. The bar on the left correspond to the overall sensitivity. Bars on the right correspond to stratified sensitivity according to the Cycle threshold (C_T) value of the RT-qPCR assay.

Table 1.

Variables associated to the positivity of the rapid lateral flow test (RLFT).

	PCR+/RLFT+ (N=98) N (%)	PCR+/RLFT-(N=67) N (%)	Crude Odds Ratio (CI95%)	Adjusted Odds Ratio (CI95%)	Adjusted P value
CT value					< 0.001
- >28	8 (15.4)	44 (84.6)	-	-
- ≤28	90 (79.6)	23 (20.4)	21.52 (9.34–55.30)	23.96 (7.75–94.03)
Typing					0.870
- non variant	49 (57.6)	36 (42.4)	-	-
- variant	48 (70.6)	20 (29.4)	1.76 (0.90–3.51)	1.08 (0.41–2.83)
Covid-19 symptoms					0.991
- no	36 (48.6)	38 (51.4)	-	-
- yes	32 (66.7)	16 (33.3)	2.11 (1.00–4.56)	1.01 (0.38–2.65)

C_T: cycle threshold. CI₉₅: confidence interval 95%.

Different criteria of acceptability (pain, sampling convenience) of the tests were assessed on about two-thirds of the tested population who accepted to fill out the corresponding questionnaire. In agreement with others,^{9 , 10} our results (data not shown) indicate that salivary auto-tests (associated or not to anterior nose) were considered less painful, especially in women and young people, and as convenient as conventional nasopharyngeal swabbing.

In agreement with others using similar RLFT,1, 2, 3, 4, 5, 6, 7 this study comforts the pertinence of the COVID19 Speed-Antigen Test to be used for detecting through mass testing asymptomatic individuals at high risk to disseminate COVID-19 due to their high viral load. The result availability within minutes could allow the rapid implementation of isolation measures and immediate set-up of contact tracing. The use of samples combining anterior nose and saliva is another innovation tested in the study. We confirm the feasibility and the excellent acceptability of noninvasive self-sampling by asymptomatic volunteers involved in large campaigns of screening.

AutoCov study group composition (city)

Amandine Baudot (Saint-Etienne), Geneviève Billaud (Lyon), Thomas Celarier (Saint-Etienne), Laure Choupeaux (Paris), Hélène Chabrolles (Clermont-Ferrand), Laura Cinieri (Saint-Etienne), Constance Delaugerre (Paris), Christine Forissier (Saint-Etienne), Emilie Frobert (Lyon), Alexandre Gaymard (Lyon), Cécile Henquell (Clermont-Ferrand), Isabelle Martin (Saint-Etienne), Agathe Mattei (Saint-Etienne), Audrey Mirand (Clermont-Ferrand), Aline Tchaplyguine (Saint-Etienne) and Jean-Marc Treluyer (Paris).

Funding

This research was supported mainly by the France Ministry of Solidarity and Health. Additional funds were obtained from the City of Saint-Etienne and the University Hospital of Saint-Etienne. The PCR results were provided free of charge by the University Hospitals of Lyon and Clermont-Ferrand and by the BioMérieux company. All these institutions are warmly acknowledged for their generous support.

Authors’ contributions

Conceptualization: JG, PM, TB, FC, PhB, EBN & BP. Data curation: JG, PaB, EV, SG & BP. Formal analysis: JG, PaB, EV, SG & BP. Funding acquisition: JG, PM, FC & BP. Investigation: JG, BB, RL & BP. Methodology: RL, SG, SP, AB, FM, TB, AutoCoV study group & BP Project administration: JG, FC, PhB, EBN & BP Resources: JG, FC & BPSoftware: PaB. Supervision: JG, BB, RL & BP. Validation: JG, PaB, EV, RL, SG, AB, FM, AutoCoV study group & BP. Visualization: PaB. Writing - original draft: JG, EBN & BP. Writing - revew and editing: all authors.

Conflict of Interest

None declared.