Analytical Sensitivity and Specificity of Two RT-qPCR Protocols for SARS-CoV-2 Detection Performed in an Automated Workflow

Abstract

WHO declared the novel coronavirus (COVID-19) outbreak a global pandemic on 11 March 2020. The establishment of standardized RT-qPCR protocols for respiratory secretions testing, as well as sharing of specimens, data, and information became critical. Here, we investigate the analytical performance of two interim RT-qPCR protocols (Charité and Centers for Disease Control (CDC)) for the qualitative detection of SARS-CoV-2 executed in a fully automated platform. Analytical specificity, PCR amplification efficiency, analytical sensitivity (limit of detection), and cross-reactivity were evaluated using contrived samples. The on-going accuracy was evaluated by retrospective analysis of our test results database (real clinical samples). N1, E, and a modified version of RdRP assays presented adequate analytical specificity, amplification efficiency, and analytical sensitivity using contrived samples. The three assays were applied to all individuals who requested the SARS-CoV-2 molecular test assay in our laboratory and it was observed that N1 gave more positive results than E, and E gave more positive results than RdRP (modified). The RdRP and E were removed from the test and its final version, based on N1 assay only, was applied to 30,699 Brazilian individuals (from 19 February 2020 to 8 May 2020). The aggregated test results available in the database were also presented.

Keywords: RT-qPCR; SARS-CoV-2; validation.

Figure 1

Assays’ analytical specificity analysis. N, N2, and N3 were excluded from the validation due to lack of analytical specificity when applied to nasopharyngeal samples of healthy volunteers.

Figure 2

Figure depicting quality control targets. Left panel (sample control RPP30) and right panel—(process control AEC)—n = 60, 30 collected with Rayon swabs and 30 collected with cotton swabs. No difference was observed between both Rayon and cotton.

Figure 3

Assays′ amplification efficiency analysis. Left panel—linear regression analysis; right panel—raw data. The better amplification efficiency was observed for N1 (close to 100%), followed by RdRP (modified), E, RdRP, and E (modified).

Figure 4

Determination of assays′ limit of detection. Left panel—probit regression analysis (inserted values unit are copies/reaction); right panel—raw data. N1 and RdRP (modified) showed better LOD, followed by E, RdRP, and E (modified).

Figure 5

Cq values observed for the first 23 positive samples detected by the first version of the proposed method: N1 (red), E (magenta) and RdRP modified (R) (light green). N1 and E showed lower Cq values than RdRP (modified). Dashed line depicts the cut-off value of >40.

Figure 6

Cq values observed for the first 75 positive samples detected by the second version of the proposed method: N1 (red) and E (magenta). Dashed line depicts the cut-off value of >40. Three samples were positive only for N1 suggesting a better diagnostic capability for this assay.