Validation of the colorimetric RT-LAMP assay for SARS-CoV-2 RNA detection
To determine the specificity and sensitivity of the RT-LAMP assay, we continued to analyze more RNA samples. We assayed a total of 768 RNA samples obtained on different days (fig. S1). Visualization of the RT-LAMP assay results 30 min after the start of the incubation at 65°C showed comparable behavior of the samples in a total of ten 96-well test plates (Fig. 3A and Table 1), indicating that the RT-LAMP assay was reproducible from day to day and from plate to plate.
Fig. 3  Detection of SARS-CoV-2 RNA using the RT-LAMP assay.
(A) Scatter plot shows a comparison of RT-LAMP assay results and RT-qPCR results for RNA samples tested on 10 96-well plates. The RNA extraction method (QC, QiaCube, a column-based method; QS, QiaSymphony, a bead-based method) is indicated. The time point for measurement by the colorimetric RT-LAMP assay was 30 min after the start of the 65°C incubation. The 96-well plate shown in Fig. 2 is not included here. Table 1 shows numbers of samples stratified according to the results of the RT-LAMP and the RT-qPCR assays. (B) Sensitivity (right) and specificity (left) of the RT-LAMP assay [derived from data in (A) and Table 1] are shown. The specificity is the fraction of RT-qPCR–negative samples correctly identified as negative by the RT-LAMP assay. For sensitivity, the RT-qPCR–positive samples were stratified by CT values into three bins (as indicated by x axis labels), and for each bin, the sensitivity is given as the fraction of qPCR-positive samples in the respective CT bin that have also given a positive result in the RT-LAMP assay. The thick black lines indicate the values of these fractions (i.e., the specificity and sensitivity estimates); the black boxes indicate the corresponding 95% confidence intervals (Wilson’s binomial confidence interval). (See also table S2).
Table 1  Shown is RT-qPCR and RT-LAMP testing of 768 clinical samples stratified into CT value bins (see Fig. 3A).
Fig. 3B and table S2 show specificity and sensitivity values calculated from these numbers.
RT-LAMP
CT   Pos   Neg  Sum
RT-qPCR   Pos   0–25  51  0  51
25–30  28  2  30
30–35  4  16  20
35–40  0  16  16
Neg   Neg  2  649  651
Sum  85  683  768 The consistency of the results during the analysis confirmed a threshold of ΔOD > +0.3 as a robust measure to identify samples that were positive for SARS-CoV-2 RNA (Fig. 3A). RT-qPCR–positive samples with a CT < 30 scored positive in the RT-LAMP assay (79 of 81), whereas almost all samples with CT values between 30 and 40 scored negative (only 4 positive of 36) (Fig. 3B). This confirmed the sensitivity of the RT-LAMP assay for detection of SARS-CoV-2 RNA in samples corresponding to a CT < 30. We observed small differences between different plates on the exact sensitivity threshold, probably caused by slight variability in plate or reagent handling. We found two RT-qPCR–negative samples that scored positive in the RT-LAMP assay (Fig. 3A and Table 1) and one sample that scored just below the ΔOD cutoff of +0.3. The overall specificity of the RT-LAMP test was 99.7% (Wilson’s 95% confidence interval: 98.9 to 99.9%), and the sensitivity for samples with CT < 30 on RT-qPCR was 97.5% (Wilson’s 95% confidence interval: 91.4 to 99.3%) (Fig. 3B and table S2).