Normalizing real-time PCR results in routine testing

Abstract

Normalization, the process of controlling for normal variation in sampling and testing, can be achieved in real-time PCR assays by converting sample quantification cycles (Cqs) to “efficiency standardized Cqs” (ECqs). We calculated ECqs as E^−ΔCq, where E is amplification efficiency and ΔCq is the difference between sample and reference standard Cqs. To apply this approach to a commercial porcine reproductive and respiratory syndrome virus (PRRSV) RT-qPCR assay, we created reference standards by rehydrating and then diluting (1 × 10⁻⁴) a PRRSV modified-live vaccine (PRRS MLV; Ingelvac) with serum or oral fluid (OF) to match the sample matrix to be tested. Sample ECqs were calculated using the mean E and reference standard Cq calculated from the 4 reference standards on each plate. Serum (n = 132) and OF (n = 130) samples were collected from each of 12 pigs vaccinated with a PRRSV MLV from −7 to 42 d post-vaccination, tested, and sample Cqs converted to ECqs. Mean plate Es were 1.75–2.6 for serum and 1.7–2.3 for OF. Mean plate reference standard Cqs were 29.1–31.3 for serum and 29.2–31.5 for OFs. Receiver operating characteristic analysis calculated the area under the curve for serum and OF sample ECqs as 0.999 (95% CI: 0.997, 1.000) and 0.947 (0.890, 1.000), respectively. For serum, diagnostic sensitivity and specificity of the commercial PRRSV RT-qPCR assay were estimated as 97.9% and 100% at an ECq cutoff ≥ 0.20, and for OF, 82.6% and 100%, respectively, at an ECq cutoff ≥ 0.45.

According to the minimum information for publication of quantitative real-time PCR experiments (MIQE) guidelines, ⁷ publications on quantitative real-time PCR (qPCR) technology should include estimates of analytical sensitivity and specificity, accuracy, precision within (repeatability) and between (reproducibility) laboratories, and the criteria for generating reportable results. To these can be added other crucial test performance parameters (e.g., diagnostic sensitivity, diagnostic specificity, and limits of detection).

Less celebrated, but vital to consistent performance as tests move from the research laboratory to routine diagnostic use, are assay standardization and data normalization. ²⁹ “Standardization” means using test protocols and procedures that reliably produce results that meet agreed-upon specifications. ⁶⁴ Simply put, “the results of laboratory tests must be uniform to inspire confidence, and uniform results can only be obtained by uniform methods.” ⁶⁵

Standardization per se will improve laboratory repeatability and reproducibility, ⁶⁰ but “normalization” is necessary to produce results that are directly comparable among laboratories. Normalization is typically achieved by re-expressing results in the context of an agreed-upon reference standard.^11,32,40 Normalization has been described as the means “to distinguish between real (biological) variation and deviations resulting from measurement processes such as run-to-run variation, disparities in mRNA quantity and quality among samples, pipetting errors, and efficiencies of the enzymes used for the reverse transcription and amplification steps.” ¹

In veterinary testing, the widespread use of commercial PCR kits has achieved de facto assay standardization for major pathogens. In contrast, normalization in routine PCR testing remains uncommon. A robust normalization method that re-expresses the quantification cycle (Cq) as a function of plate-specific amplification efficiency (i.e., efficiency standardized Cqs [ECqs]) has been proposed.^10,43 Our objective was to adapt and apply the ECq methodology to routine PCR testing of serum and oral fluid (OF) using a commercial porcine reproductive and respiratory syndrome virus (PRRSV; Betaarterivirus suid 2) reverse-transcription qPCR (RT-qPCR) assay.

Materials and methods

Experimental design

To adapt the ECq methodology to routine PRRSV RT-qPCR testing, target- and matrix-specific (serum or OF) reference standards were created by rehydrating and then diluting (1 × 10⁻⁴) a PRRSV modified-live virus (MLV) vaccine (Ingelvac PRRS MLV, serial 2451416A; Boehringer Ingelheim) with PRRSV-free serum or OF. Serum or OF reference standards (n = 4) were included on each plate to match the sample matrix to be tested. Technically, the process does not use plates, but for practical convenience, a “plate” will be defined as a block of ≤ 48 reactions loaded and tested in 1 qPCR program. Reference standards were used to calculate the mean plate amplification efficiency (E) and mean reference standard Cq, which were then used to calculate the individual sample ECqs:

$$\begin{array}{l}EfficiencystandardizedCq=ECq=E^{-\mathrm{\Delta }Cq}\\ =E^{-(sampleCq-meanreferencestandardCq)}\end{array}$$ (1)

To illustrate the use of ECqs, 132 serum and 130 OF samples collected over time from a group of 12 pigs vaccinated with a PRRSV MLV vaccine (Ingelvac PRRS MLV) ²⁶ were tested with a PRRSV RT-qPCR assay. Sample Cqs were first converted to ECqs and then analyzed by receiver operating characteristics (ROC) analysis to obtain diagnostic sensitivity and specificity estimates for the PRRSV RT-qPCR assay.

Reference standards

Target-specific (PRRSV) and matrix-specific (serum or OF) reference standards were generated by rehydrating a lyophilized PRRSV MLV vaccine with serum or OF known to be free of PRRSV. In brief, 45 mL of PRRSV-free serum or OF were centrifuged in 50-mL centrifuge tubes (Nunc; Thermo Scientific) at 3,300 × g for 3 h (IEC Centra CL3 centrifuge; Thermo Fisher). The supernatant was used to rehydrate a 10-dose vial of PRRSV MLV vaccine (Ingelvac PRRS MLV) to a final volume of 20 mL. Thereafter, this solution was 10-fold serially diluted (1 × 10⁻¹ to 1 × 10⁻⁶) using the appropriate matrix, and all dilutions were tested in triplicate with the RT-qPCR assay; the dilution that achieved a consistent Cq of 30 for each matrix was chosen (1 × 10⁻⁴). Matrix-specific PRRSV reference standards at the selected dilution were included in the extraction and amplification processes such that each plate (40 serum plates; 28 OF plates) included 4 reference standards.

Serum and oral fluid samples

A panel of matched OF and serum samples was created by collecting samples from a study conducted with the approval of the Iowa State University Office of Research Ethics, Institutional Animal Care and Use Committee (IACUC, Log 3-16-8214-S) involving 12 pigs over a period of 50 d (i.e., before and after vaccination with a PRRSV-2 MLV vaccine). ²⁶ In brief, 12 individually housed 14-wk-old PRRSV-naïve pigs were intramuscularly administered a PRRSV-2 MLV vaccine (2 mL, Ingelvac PRRS MLV) at day post-vaccination (dpv) 0. Individual serum samples (n = 132) were collected weekly at dpv −7, 0, 14, 17, 21, 28, 35, and 42. Subsets of pigs (n = 4) were rotated through the sampling schedule to obtain daily serum samples for dpv 3–11. Individual OF samples (n = 130) were collected daily from dpv −7 to 42 by providing each pig access to a 3-strand twisted 100% cotton rope for 30 min. OFs were recovered from the rope and then aliquoted into 5-mL tubes (Corning; Fisher). Serum and OF samples were stored at −80°C until tested. Matrix-specific PRRSV reference standards (n = 4 per plate) were included in each serum and OF plate (n = 8 plates per sample type).

Nucleic acid extraction

Total nucleic acid extraction was performed (RealPCR DNA/RNA spin column kit; Idexx) following the instructions of the manufacturer for the alternative protocol for all sample types. For OFs, 200 µL of OFs were added to 200 µL of a lysis working solution consisting of 190 µL of lysis buffer, 5 µL of carrier RNA, and 5 µL of proteinase K. For sera, 200 µL of serum were mixed with 5 µL of proteinase K, and then 200 µL of a lysis working solution consisting of 195 µL of lysis buffer and 5 µL of carrier RNA. Binding conditions were adjusted with 200 µL of 96% ethanol solution (Thermo Fisher). Thereafter, 600 µL of lysate was transferred to a spin column and washed. Nucleic acids were eluted with 50 µL of elution water pre-heated to 70°C. Every nucleic acid extraction run included 1 positive and 1 negative extraction control and 4 replicates of matrix-specific PRRSV reference standards. All extracts were directly subjected to qPCR.

RT-qPCR

All qPCRs were performed using RealPCR RNA master mix and RealPCR PRRS types 1-2 RNA mix (Idexx). The target mix for PRRSV detects pathogen-specific RNA by a cyanine (Cy) 5 probe and an endogenous pig-specific RNA internal sample control (ISC) by a hexachlorofluorescein (HEX) probe. All RT-qPCRs were performed with 5 µL of nucleic acid extract using tubes compatible with the Mic qPCR system (Bio Molecular Systems) and a magnetic induction cycler (Mic qPCR cycler; Bio Molecular Systems) programmed as: 50°C (15 min) for reverse transcription, denaturation at 95°C (1 min), 45 cycles at 95°C (15 s), followed by 60°C (30 s) for amplification. In addition to extraction controls, every run included an amplification control. Results were analyzed using the Mic qPCR cycler software v.2.10.4 (Bio Molecular Systems).

Amplification efficiency estimates

Estimates of E were calculated using the Mic PCR software (Bio Molecular Systems). Initially, the dynamic analysis feature was applied to the amplification data of the reference standards (n = 4) run on each plate. In brief, the “dynamic” analysis performed a baseline correction by 1) determining the reaction-specific average fluorescence values measured before the start of the exponential phase; 2) defining the linear phase by iteratively searching a fixed number of data points to determine which of these showed exponential growth; and 3) subtracting the values from (1) and (2) to baseline-correct each sample. Thereafter, the software determined the window-of-linearity (WoL) in the exponential phase using the LinRegPCR method.^47,49 This process involved setting the upper limit of the WoL at the mean fluorescence level found at the maximum of the second derivative (i.e., the beginning of the linear phase) and downward selecting 4–6 data points from the baseline-corrected fluorescence of each reaction to set the lower limit. Then, an iterative modification was applied to this initial WoL by systematically lowering the data points by half per cycle. For each window, the CV was calculated from the x̄ and the SD of the amplification efficiencies. The minimum CV marked the WoL in which the individual sample amplification efficiencies showed the least variation relative to the mean E. Individual Es were then calculated from the slope of the regression line within the WoL:

$$E={10}^{-1/slope}-1$$ (2)

These values were displayed in the output for each reference standard as a ratio of 1 (or 100%). Likewise, a cycle threshold was automatically established at 75% of the WoL. The rest of the plate dataset (i.e., samples and controls) were baseline-corrected under the dynamic analysis feature, and the previously obtained cycle threshold was manually set to obtain samples and reference standard Cqs.

Efficiency standardized Cqs

Sample Cqs were converted to ECqs (equation 1) to account for plate-specific Es and to normalize the results relative to matrix-specific reference standards.^10,43 E can be expressed either as a ratio (number of target amplicons at the end of a PCR cycle divided by the number at the beginning) or as a percentage. Thus, a doubling at each cycle would represent an E of 2 or 100%. Equation 1 requires that E be expressed as a ratio, which was done by first calculating the plate E (%) from the 4 reference standards and then adding 1 to the arithmetic mean of the 4 reference standards run on each plate. E estimates > 100% were truncated at 100%. ΔCq (sample Cq – mean reference standards Cq) was calculated using the arithmetic mean from the individual Cqs of the plate reference standards (n = 4). For example, a sample with a Cq of 28, run on a plate with a reference standard with mean E of 85% (E = 1.85) and mean Cq of 31, would have an ECq of 6.33 (ECq = E^–ΔCq = 1.85^–(28–31) = 6.33). ECqs for samples with indeterminate Cqs were assigned a Cq of 45 (i.e., the total number of cycles in the PCR program).

Analysis

All ECq values were transformed to their cube root for the analysis. For illustrative purposes, ROC analyses were conducted (R v.4.2.1., package pROC, https://www.r-project.org/) for each specimen (serum and OF) to estimate the area under the curve (AUC) and the diagnostic specificities and sensitivities for a range of ECq cutoffs. The ROC analyses of the PRRSV RT-qPCR results were based on ECq results for 72 sera and 70 OFs. Samples collected before vaccination (≤ 0 dpv; 24 sera; 24 OFs) were assigned PRRSV-negative status. Samples collected at dpv 3–14 (48 sera; 46 OFs) were assigned PRRSV-positive status. To account for the correlated nature of the data, the nonparametric DeLong method was used to estimate 95% CIs for the AUC. ¹⁵ Estimation of 95% CIs for diagnostic sensitivities and specificities was performed using the nonparametric stratified bootstrapping method with 10,000 iterations, in which lower and upper bounds were computed as the 5th and 95th percentiles of the sensitivities or specificities derived from 10,000 iterations. ⁸

Results

Reference standards

Mean reference standard plate Es were 1.8–2.6 for 40 serum plates and 1.7–2.3 for 28 OF plates. Mean reference standard Cqs were 29.1–31.3 for serum and 29.2–31.5 for OF (Fig. 1).

Figure 1.

PRRSV RT-qPCR testing results for matrix-specific (serum or oral fluid [OF]) reference standards. Standards were created by rehydrating and diluting (1 × 10⁻⁴) a PRRSV modified-live virus vaccine (Ingelvac PRRS MLV; Boehringer Ingelheim) with serum or OF. A. Distribution of mean plate amplification efficiency for serum and OFs. B. Distribution of mean Cqs for serum and OFs. In each case, the means were calculated from the 4 reference standards run on each plate.

PRRSV RT-qPCR

ECqs represent target concentrations of PRRSV RNA in a sample relative to the target concentration in the reference standards included on each qPCR plate. We detected PRRSV RT-qPCR positives on dpv 3–5 in serum (x̄ ECq: 1.39) and OF (x̄ ECq: 1.32) in all pigs. That is, in both serum and OF, the concentration of virus was approximately the same as in the reference standards (Table 1). Thereafter, PRRSV positivity peaked on dpv 7 in serum (x̄ ECq: 2.24) and on dpv 9 in OF (x̄ ECq: 2.00) in all pigs (Fig. 2).

Table 1.

PRRSV RT-qPCR response (geometric mean of 12 pigs) expressed in ECqs and Cqs by day post-vaccination and specimen.

Days post-vaccination	Serum		Oral fluid
	x̄ ECq (SE)	x̄ Cq (SE)	x̄ ECq (SE)	x̄ Cq (SE)
−7	0.05 (0.01)	44.5 (0.50)	0.13 (0.02)	44.6 (0.39)
0	0.03 (0)	45.0 (0)	0.11 (0.02)	44.6 (0.33)
3, 4, 5	1.39 (0.32)	28.9 (0.85)	1.32 (0.21)	33.2 (0.65)
6, 7, 8	2.49 (0.44)	27.3 (0.85)	1.29 (0.19)	31.5 (1.44)
9, 10, 11	1.78 (0.43)	28.0 (0.90)	1.35 (0.47)	32.2 (1.05)
14	1.06 (0.26)	30.2 (0.70)	1.03 (0.18)	33.6 (0.81)
17	0.90 (0.15)	31.0 (0.70)	0.92 (0.26)	33.8 (1.02)
21	0.88 (0.19)	31.0 (0.70)	0.97 (0.14)	33.7 (0.91)
28	0.69 (0.15)	32.6 (0.80)	0.96 (0.17)	33.5 (0.74)
35	0.51 (0.08)	33.6 (1.00)	0.71 (0.10)	35.9 (0.90)
42	0.31 (0.04)	36.9 (0.20)	0.59 (0.05)	37.5 (0.33)

ECq = E–ΔCq, where E is the mean reference standard amplification efficiency (by plate), and ΔCq is the difference in raw Cq values between sample Cq and mean reference standard Cq (E estimates > 100% were truncated at 100% before calculation of ECqs). PRRSV = porcine reproductive and respiratory syndrome virus; RT-qPCR = reverse-transcription quantitative real-time PCR.

Figure 2.

PRRSV RT-qPCR performance by day post-vaccination (n = 12 pigs). Left y-axis denotes positivity rate (%) based on serum- and oral fluid–specific cutoffs derived from a receiver operating characteristic analysis. Right y-axis denotes numeric response (efficiency standardized Cq [ECq]) based on geometric means of 12 pigs.

The AUCs estimated for serum and OF samples were 0.999 (95% CI: 0.997, 1.000) and 0.947 (0.890, 1.000), respectively. The optimal ECq cutoff for serum (ECq ≥ 0.20) provided a diagnostic sensitivity of 97.9% (95% CI: 93.8%, 100%) and diagnostic specificity of 100% (95% CI: 100%, 100%). For OF, a cutoff of ECq ≥ 0.45 provided a diagnostic sensitivity of 82.6% (95% CI: 71.7%, 93.5%) and diagnostic specificity of 100% (95% CI: 100%, 100%; Table 2).

Table 2.

PRRSV RT-qPCR (RealPCR PRRS Types 1-2 RNA mix; Idexx) diagnostic sensitivity (%) and specificity (%) by specimen and ECq cutoff.

Cutoff (ECq)	Serum		Oral fluid
	Diagnostic sensitivity (95% CI)	Diagnostic specificity (95% CI)	Diagnostic sensitivity (95% CI)	Diagnostic specificity (95% CI)
0.20	97.9 (93.8, 100)	100 (100, 100)	93.5 (84.8, 100)	70.8 (50.0, 87.5)
0.25	97.9 (93.8, 100)	100 (100, 100)	93.5 (84.8, 100)	91.7 (79.2, 100)
0.30	95.8 (89.6, 100)	100 (100, 100)	89.1 (80.4, 97.8)	95.8 (87.5, 100)
0.35	95.8 (89.6, 100)	100 (100, 100)	89.1 (80.4, 97.8)	95.8 (87.5, 100)
0.40	93.8 (85.4, 100)	100 (100, 100)	87.0 (76.1, 95.7)	95.8 (87.5, 100)
0.45	91.7 (83.3, 97.9)	100 (100, 100)	82.6 (71.7, 93.5)	100 (100, 100)
0.50	91.7 (83.3, 97.9)	100 (100, 100)	80.4 (69.6, 91.3)	100 (100, 100)

Diagnostic sensitivity and specificity point estimates derived from ROC analysis (R v.4.2.1, package pROC). ECq = efficiency standardized Cq = Cq values re-expressed as a function of plate amplification efficiency that represent target concentration relative to matrix-specific PRRSV reference standards (ECq values were transformed to their cube root). PRRSV = porcine reproductive and respiratory syndrome virus; RT-qPCR = reverse-transcription quantitative real-time PCR.

Discussion

The introduction of a new diagnostic platform is typically followed by a period of mixed success as researchers explore its capabilities, but over time, the procedures that reliably produce accurate results are identified and adopted. ⁶⁵ The ELISA is a case in point. Development of ELISA technology was followed by its wholesale application to a variety of infectious diseases and an exploration of the boundaries of the technology. ¹⁷ For example, in developing a rubella virus serum antibody ELISA, the response was affected by the type of plastic used in the ELISA plate, conjugate dilutions, and incubation times. ³ This led to a focus on ELISA optimization and standardization. ⁵

Standardization of qPCR technology has followed a similar path. For example, in an inter-laboratory comparison of 13 influenza A virus RT-qPCR assays using samples of known status, a 100-fold difference was found between the most sensitive and least sensitive assays. ²² However, as predicted, time and the adoption of standardized reagents and protocols have improved qPCR assay consistency. ⁶⁵ Thus, in a 2022 comparison of 12 African swine fever virus commercial PCRs, mean Cq values were in general agreement. ⁴⁵

Standardization is crucial to reliable test performance, but normalization is needed to control for the run-to-run variation arising from differences in samples^19,61,66,68 and in sample handling and processing,^9,24,25,67 as well as from differences among technicians ²⁰ and among manufacturing lots. ⁶ For ELISAs, normalization was achieved by expressing the sample response (i.e., absorbance values) relative to positive and/or negative reference sera run on the same plate (i.e., sample:positive or sample:negative ratios).^14,56 Likewise, methods for normalizing PCR results were developed in the 2000s and are now used routinely in gene expression research.^2,12,33,42 We explored a practical method to normalize routine PCR test results.

Normalization of PCR results is necessary to control for variation and produce results that are comparable within and between laboratories. However, unique to PCR, the rate of the reaction (E) is also an important source of variation that must be accounted for in routine testing. In general terms, efficiency is a measure of the performance of a system. ¹⁶ In equation 3 of PCR kinetics:

$$\left(N_c=N_0\times E^c\right),$$ (3)

the relationship between N₀ (initial target concentration) and N_c (final target concentration) is a function of E to the power of the number of the cycle (^c). ⁵³

It follows that the measure of efficiency is the fraction of target molecules copied in one PCR cycle (i.e., the ratio of N_c:N₀, where 2 = 100% efficiency). ³¹ When PCR results are reported as raw Cqs, it is assumed that E is 100%, but there are both technical and theoretical reasons why this cannot be true. From the laboratory perspective, E is affected by elements specific to the test procedure, including amplicon size, ¹³ probe chemistry,^4,30 plate color, ⁴⁸ and thermocycler. ⁵⁸ Further, reaction kinetics modeling has shown that efficiency is not a constant, but varies throughout the amplification phase giving a sigmoidal shape to the process (i.e., ~100% efficiency during the exponential phase and then decreasing until it reaches zero at the plateau phase).^{21,28,35,36,53,54} Finally, from the perspective of thermodynamics, a PCR reaction is an isolated and irreversible system that tends towards equilibrium when a single-stranded target is denatured, hybridized, and polymerized to the double-stranded state.^39,52 Throughout the PCR reaction, energy is released to the environment, which means that efficiency must be < 100%.^41,70 Ultimately, failure to account for E is a critical oversight because of the impact of E on N_c. For example, based on equation 3 and under ceteris paribus conditions, a test sample with an initial concentration of 1 × 10² (= 100) copies (N₀) amplified with an efficiency of 100% (E = 2) will produce 1 × 10^6.52 = 3,276,800 copies (N_c) after 15 cycles of amplification (c = 15). However, if efficiency is 80% (E = 1.80), N_c = 1 × 10^5.38 = 674,664 copies. That is, a 0.2 reduction in E would result in ~79% reduction in the number of target copies.

In qPCR assays, normalization can be done using either absolute or relative quantification. Absolute quantification is done by constructing a calibration curve using a reference standard of known concentration and then estimating the concentration of the target in the sample by interpolation. ⁵⁰ Because differences in PCR efficiency result in calibration curves with different slopes, for absolute quantification to be reliable, a calibration curve must be generated concurrently with the sample, thereby making it unmanageable for most service laboratories.^7,50,55,59 Alternatively, relative quantification is achieved by expressing the sample Cq as the fold difference relative to the Cq of the reference standard run on the same plate. ²³ Unlike absolute quantification, it is not necessary to establish the target concentration in the reference standard, which makes it a practical option for routine qPCR testing in service laboratories.

For both absolute and relative quantification, specific criteria for reference standards have been described^23,44,69:

1. The standard should be pure, stable, and reproducible.

2. Inherent Es in the sample and the standard should be equivalent.

3. The amplification characteristics of the matrix in which the standard is suspended should resemble the sample matrix.

4. The reference standard should be amplified concurrently with the test samples.

We generated reference standards by rehydrating a commercial PRRSV vaccine with the matrix corresponding to the specimens to be tested (i.e., serum or OF free of PRRSV), and including the reference standard on each plate in quadruplicate. This approach was a practical solution to the problem of identifying a reference standard that met the requirements listed above and was likewise widely available to service laboratories performing PRRSV RT-qPCR assays:

1. Each commercial vaccine manufacturing serial produced in the United States and the European Union must meet sterility, purity, safety, and potency standards given in Title 9 §113.6 of the U.S. Code of Federal Regulations and the European Union Regulation (EU) 2019/6.^18,63,69

2. The requirement for equivalent inherent E was satisfied by using the same target (PRRSV RNA) in both the sample and the reference standard.

3. Requirement 3 was addressed by rehydrating the vaccine with the specimen matrix (i.e., serum or OF, as recommended by others ¹¹ ).

4. Concurrent amplification of reference standards and test samples was done by including 4 technical replicates of the reference standard on each qPCR plate.

Notably, positive extraction controls and positive amplification controls consisting of non-native target suspended in a synthetic matrix do not meet the criteria and, therefore, are not appropriate reference standards.

The 2 most common approaches to relative quantification are the double-delta Cq (^ΔΔCq) ³⁸ and ECqs.^10,43 These 2 approaches differ most importantly in that the ΔΔCq method assumes that E is 100%, whereas the ECq adjusts for actual E. Given the impact of E on N_c (final target concentration), the ECq methodology is preferable and was used in our study. E may be calculated by linear regression analysis of the exponential phase fluorescence data,^47,49 from the slope of the standard curve,^50,57 or through mathematical model-fitting.^37,51,62 We obtained plate-specific estimates of E from the exponential phase of the amplification curve of the plate reference standards using the LinRegPCR method. ⁴⁷ This approach did not include data from the initial or terminal ends of the amplification curve and, therefore, could produce E estimates > 100%.^21,23 Following the fundamentals of E described earlier, E cannot be > 100% and thus, plate E estimates > 100% were truncated at 100% prior to the calculation of ECqs.

ECqs represent a semiquantitative measure of target concentration relative to the reference standards, and their interpretation is simple: the larger the ECq value, the more nucleic acid there is in the sample relative to the reference standard (e.g., an ECq of 13.5 means that the target concentration in the sample was 13.5 times the concentration of target in the reference standard). Further, accounting for E produces values with distinct differences that would be otherwise unnoticeable as Cqs. For example, the serum response at dpv 17 and dpv 21 produced the same raw Cq of 31, whereas when converted to ECq, dpv 17 and dpv 21 ECqs were 0.90 and 0.88, respectively. In practice, these differences could be of significance for classifying sample results as positive or negative based on a statistically valid cutoff. Thus, our PRRSV ECq estimates were consistent with the reported dynamics of PRRSV over time (i.e., high RT-qPCR positivity rates in serum and OF in the first week post–PRRSV-inoculation or -vaccination, followed by a gradual decline thereafter).^27,34,46

There are several benefits to converting Cqs to ECqs. ECqs produce test results that are normalized by accounting for differences in Es among plates. Thus, results obtained within one laboratory or between several laboratories using the same reference standard are comparable. Strictly speaking, this general approach to test data normalization is already in use (e.g., ELISAs, with reference standards provided by the manufacturers and normalization calculations automated within the testing software). Additionally, ECqs also provide a solution to the problem of estimating qPCR test performance. That is, when converted to ECqs, all samples (including samples with Cq values greater than the cutoff) have a numeric value. Thus, test performance can be evaluated using well-established statistical techniques (e.g., ROC analysis, as was done in our illustration). Overall, we demonstrated that PCR normalization could be achieved using a simple, practical method and with immediate benefits to our clients.