Analytical and Clinical Performance of Droplet Digital PCR in the Detection and Quantification of SARS-CoV-2

Abstract

Background and objective: Since the initial coronavirus disease outbreak in late 2019 (COVID-19), reverse-transcription real-time polymerase chain reaction (RT-qPCR) has become the gold standard test to detect severe acute respiratory syndrome-related coronavirus 2 (SARS-CoV-2). However, a more sensitive and accurate diagnostic tool was required. Therefore, droplet digital polymerase chain reaction (ddPCR) was suggested as an alternative method. Here, we evaluated the performance of ddPCR to detect SARS-CoV-2 and compared it to the performance of RT-qPCR.

Methods: The analytical performances, including limit of blank and limit of detection, were established using positive and negative SARS-CoV-2 reference materials. A total of 366 RNA extracts (173 positive and 193 negative by RT-qPCR) were collected from four institutions and tested with a Bio-Rad SARS-CoV-2 ddPCR kit that detects the SARS-CoV-2 genome using primers for N1 and N2.

Results: Limit of blank was set at 0, and the limits of detection of N1 and N2 were 1.99 copies/μL and 5.18 copies/μL, respectively. Linearity was evaluated using serial dilution samples, which demonstrated good results (R²: 0.999, linear range: 5.88-6825.25 copies/μL for N1 and R²: 0.999, 5.53-5855.47 copies/μL for N2). The results of ddPCR and RT-qPCR revealed substantial agreement (Cohen's kappa: 0.639, p < 0.01). The 63 samples with positive ddPCR but negative RT-qPCR showed low copy numbers, and 55% of them had COVID-19-related symptoms.

Conclusions: Droplet digital polymerase chain reaction demonstrated excellent sensitivity for SARS-Cov-2 detection and consistently agreed with the results from conventional RT-qPCR. Furthermore, ddPCR provided quantitative data that can be used to monitor changes in the viral load of patients with COVID-19.

Fig. 1

Linearity evaluation of droplet digital polymerase chain reaction performance to measure N1 and N2. The horizontal axis is the dilution ratio of the pooled sample (1:x). The vertical axis is the log-scaled copy value. Regression equation was denoted with y, 10 log-scaled copy value (copies/mL), and x, dilution ratio. The linear model and cubic model of N1 (a) and N2 (b) was evaluated with four serial dilutions. The differences between the measured and predicted values in N1 (c) and N2 (d) was presented as the deviation plot with the log-scaled copy value unit (an allowable deviation from linearity was shown as 1% in a dotted line)

Fig. 2

Scatter plot and linear regression of the copy value of droplet digital polymerase chain reaction (ddPCR) for the quantification cycle (C_q) of reverse-transcription real-time polymerase chain reaction (RT-qPCR) and Bland–Altman plots of the difference against the mean for measured and predicted C_q values. All positive RNA extract samples with definite C_q values and positive copy values within a linear range of ddPCR were analyzed. Log-scaled copy values (log copies/mL) of N1 (a–c) and N2 (d–f) were dotted for the C_q of each RT-qPCR target gene, RdRp, E, and N with paired Bland–Altman plots. The linear regression line, equation, and R2 are displayed on each graph. All pairs showed significant linear correlation (p < 0.01). In each Bland–Altman plots, the x-axis was the average of measured C_q values by RT-qPCR and predicted C_q values using linear regression function with the log-scaled copy value, and the y-axis was the difference of two values. The upper and lower limit of agreement was displayed with a horizontal dotted line, with the 95% confidence interval presented in square brackets. SD standard deviation

Fig. 2

Scatter plot and linear regression of the copy value of droplet digital polymerase chain reaction (ddPCR) for the quantification cycle (C_q) of reverse-transcription real-time polymerase chain reaction (RT-qPCR) and Bland–Altman plots of the difference against the mean for measured and predicted C_q values. All positive RNA extract samples with definite C_q values and positive copy values within a linear range of ddPCR were analyzed. Log-scaled copy values (log copies/mL) of N1 (a–c) and N2 (d–f) were dotted for the C_q of each RT-qPCR target gene, RdRp, E, and N with paired Bland–Altman plots. The linear regression line, equation, and R2 are displayed on each graph. All pairs showed significant linear correlation (p < 0.01). In each Bland–Altman plots, the x-axis was the average of measured C_q values by RT-qPCR and predicted C_q values using linear regression function with the log-scaled copy value, and the y-axis was the difference of two values. The upper and lower limit of agreement was displayed with a horizontal dotted line, with the 95% confidence interval presented in square brackets. SD standard deviation

Fig. 3

Severe acute respiratory syndrome-related coronavirus 2 copy value changes of confirmed patients in follow-up test samples. The vertical axis reflects the copy value (copies/μL) and the horizontal axis reflects the post-admission days. In each graph, the droplet digital polymerase chain reaction test results were presented by full or empty dots (nasopharyngeal [NP] swab, N1, and N2) and squares (sputum, N1 and N2). Three patients, SCMC116, SCMC117, and SCMC118 (a–c) were improved and transferred to residential treatment centers. However, one patient, SCMC122 (d) worsened during hospital admission and was transferred to the COVID-19 intensive care system