Graphical observations of the Sigmoid Fluorescence Amplification plots of Calibrators, Diluted Calibrators and patients’ sample
All four validated calibrators showed a typical sigmoid plot with an exponential increase immediately above the baseline fluorescence (Y0), increasing to a maximum slope at the inflection point (b), followed by plateauing and reaching the maximum fluorescence (Fmax - Y0 = ‘a’) which was considered as the 100% normalised fluorescence intensity, used also as the reference fluorescence intensity for comparisons (Fig. 1a, Table 1). Yo was nearly the same for all samples. The normalisation to 100% fluorescence intensity of the validated calibrators was done as reference for calculations of the relative fluorescence intensities of the diluted calibrators and the patients’ samples with Cq nearest to a validated calibrator. The sigmoid fluorescence amplification plot was near symmetrical at the inflection point. NTC had no Cq as it did not cross the threshold (Fig. 1a).
The repeated one-tenth dilutions of calibrators at 1 IU/µl and 0.1 IU/µl showed a decreased normalised fluorescence intensity, compared to the 100% fluorescence intensity of the 10 IU/µl validated calibrator and a decreased slope at the inflection point (Fig. 1a).
Patients’ RNA RT-PCR at medium viral load (Cq of 30 to 33) with Cq at 31.62 showed near 100% normalised fluorescence intensities and good slope. At low plasma viral load (Cq of 33 to 36), the slope and fluorescence intensity decreased markedly (Fig. 1b). These results from the patients’ RNA RT-PCR confirmed that there was a decreased performance at Cq >33.19 of the 10 IU/µl validated calibrator.
Data analysis of Linear Calibration Plots of Validated and Diluted Calibrators
Calibration plot of validated calibrators had coefficient of correlation, R of 0.99963, and the coefficient of determination, R2 of 0.99925 (99.925%), indicating good linearity. The slope (M), efficiency (E) and Y-intercept (B) were within acceptable limits (Box in Fig. 1c), indicating good performance of the calibration plot.
The calibration plot with 10 IU/µl and its diluted concentrations had decreased numerical value of slope (less steep) and y-intercept, and increased efficiency, indicating decreased performance and unacceptable calibration plot (Box in Fig. 1d). The R = 0.9924 and R2 = 0.9849 (or 98.49%) were decreased but had good linearity.
Data Analysis of four-parameter sigmoid fluorescence amplification plots of Calibrators, Diluted Calibrators and RNA from patients’ plasma
The graphical description of the sigmoid amplification was given in Fig. 1a and b. SigmaStat software analysed the four-parameter sigmoidal or logistic curve fit model (Table 1; Equation-2) for validating the sigmoid amplification plot. The lowest SigmaStat validated Cq for patient’ sample was 39.9. The numerical values obtained by this model was nearly comparable, but not exactly, with the graphical observation in Fig.1a and b, and its data measured manually in Table 2. The fluorescence intensity change (‘a’) and slope (b) of the diluted calibrators decreased when compared to the nearest validated calibrator (10 IU/μl). These two parameters were selected for differentiating quantitative and qualitative RT-PCR.
The fluorescence intensity decreased sharply in the patient sample at Cq >33.2, but the slope did not show decrease up to Cq of 36.0. At Cq >36.0 there was large decrease in slope. The slope at inflection points from SigmaStat software appeared to be less reliable from Cq of 33 – 36 (Table 1) than that by graphical observation (Table 2).
There was a good relationship between the LinRegPCR software data with graphically observed data at various concentrations and Cq
Slope and efficiency, E of the validated calibrators were 0.297 to 0.260 and 1.98 to 1.82, respectively, from LinRegPCR software (Fig. 2; Table 2), and were within the acceptable range of RT-qPCR.
The diluted calibrators showed decreased slope (0.243 and 0.230), fluorescence intensity (89% and 78%), and E (1.75 and 1.7). The manually measured slope and fluorescence intensity were also decreased (1.477 and 1.08) (Table 2). Therefore, from these data the 10 IU/μl calibrator concentration and its corresponding Cq at 33.2 could be considered as the point of differentiation between quantitative and qualitative RT-PCR. The diluted calibrator concentration of 0.1 IU/μl corresponding to 42.8 IU/ml of plasma viral load, was the limit of detection, as further dilutions did not give a fluorescence amplification. Therefore, the validated qualitative RT-PCR range of calibrators were from 10 to 0.10 IU/μl.
Slope, Efficiency, and fluorescence intensity of RNA from patients’ plasma: The LinRegPCR software data of patients’ plasma at high viral load (Cq <30) showed an efficiency 1.98 to 1.80 and samples with medium viral load (Cq 30 to 33) showed an efficiency of 1.708 to 1.797, nearly correspond with the acceptable validated calibrator efficiency and slope (Table 2). At low plasma viral load (Cq >33), the slopes decreased to 0.238 and 0.173, and efficiency decreased to 1.728 and 1.489 (Table 2). Similarly, the manually measured normalised fluorescence intensity and slope at inflection point decreased at Cq >33.2 (10 IU/μl calibrator). Again, confirming that the Cq of 33.2 was the cut off between quantitative and qualitative RT-PCR.
Bias/Uncertainty and analytical variation (%CV) of Cq obtained from RT-PCR of validated calibrators and diluted calibrators were presented as the mean±SD, 95% CI of mean and %CV (n = 6). The 95% CI of mean of diluted calibrators were overlapping, indicating increased variations. The %CV of validated calibrators ranged from 1.49 to 2.53, and that for the diluted calibrators were 2.087 to 3.169 (Table 3). Although the diluted calibrator %CV was higher, both were within the acceptable analytical performance of <7% (Table 4). The lower %CV was possible due to the machine and reagent performance.
Comparison of X-Y scatter, trendline and correlations of selected RT-PCR parameters of diluted calibrators with patients’ RNA
The RT-PCR parameters compared were slope, fluorescence intensity and efficiency of diluted calibrators (Fig. 3, a, b, c) with patients’ sample (Fig. 3, d, e, f). The RT-PCR performance of diluted calibrators and patients’ sample showed decreased performance with decrease in slope, fluorescence intensity and efficiency, and the decreased performance was comparable. The correlations coefficient, R and the coefficient of determination, R2 were found to be high for diluted calibrators (R = -0.830 to 0.882; R2 = 0.639 to 0.790) and for patient sample (R = -0.619 to -0.838; R2 = 0.473 to 0.742) and were comparable. The correlations that increased with a quadratic trendline were preferred over that with linear trendline. As expected, the correlations were negative with Cq and positive with concentrations of diluted calibrators; when concentration decreased, Cq increased. The results showed that the performance decreased at calibrator concentration <10 IU/μl and at Cq >33.2 and were comparable with that of patients’ sample.
Decision criteria for determining the range of qualitative real-time RT-PCR
A. The graphical observations and the manually measured data from fluorescence amplification plots were found to be the most useful method for determining the range of Cq for qualitative RT-PCR of the patients’ sample. As the cut-off for differentiating quantitative and qualitative RT-PCR was Cq 33.2 (may vary from assay to assay) of 10 IU/μl calibrator, it might be used with each assay set of patient samples (Table 4).
B. Four-parameter numerical data from SigmaStat software was most suited for validation of the sigmoid amplification data and to differentiate it from non-sigmoidal plots, especially with Cq >39. The highest validated Cq of the qualitative RT-PCR was 39.9 in this study (Table 1). Fluorescence intensity and slope comparisons were more reliable with the graphical observations.
C. LinRegPCR software demonstrated that the cut-off calibrator concentration was 10 IU/μl with Cq of 33.2. The data from LinRegPCR software matched well with the manually measured data (a) at all concentrations or Cq of quantitative and qualitative RT-PCR, (b) and between the calibrator data and patients’ data (Table 2, Table 4 A and C).
D. Data analysis of Calibration plots should be in the acceptable range of slope, efficiency and sensitivity for reporting both quantitative and qualitative RT-PCR. If the values were outside this range, patient’s RT-PCR was unacceptable. This problem can be solved by recalibration or by using fresh calibrators and other reagents (Fig. 1c, 1d and Table 4 D).