Comparison of the Use of Serum and Plasma as Matrix Specimens in a Widely Used Noncommercial Dengue IgG ELISA

Abstract.

Although sera are most commonly used in serological diagnostic tests for dengue, sometimes only plasma containing ethylenediaminetetraacetic acid (EDTA) may be available. When we studied the performance of a widely used noncommercial dengue ELISA in the detection of reactive IgG in sera and plasma from the same individuals, we found no significant differences in the diagnostic performance of the assay. The inter-specimen coefficient of variation (CV) of the optical density was 0.081 and the concordance correlation coefficient (CCC) was 0.92, showing a very strong agreement between the two matrix specimens. The intra-specimen CV and CCC were higher for plasma than for serum at low dilutions, but similar at high dilutions. Overall, our results show that the performance of a widely used in-house ELISA using plasma containing EDTA is equivalent to serum with the recommendation to assay the plasma specimens in duplicate to reduce variability of results at lower dilutions.

Dengue virus (DENV) IgG ELISAs are used extensively in epidemiological surveys^1–4 as these are a relatively quick and inexpensive method for primary screening for dengue and require minimal training,⁵ despite the known limitations of cross-reactivity with other circulating flavivirus such as Zika virus. Most protocols recommend serum as the sample matrix^6–8 as anticoagulants may affect the performance of assays.⁹ However, in instances where sera are not available or where peripheral blood mononuclear cells are required for other studies,¹⁰ plasma from blood collected in EDTA may only be available. To the best of our knowledge, only one commercial dengue ELISA (Dengue IgG capture, catalog no. DEN02G; Alere, Brisbane, Australia) has been evaluated for use with plasma and sera.¹⁰ A long-established in-house DENV IgG ELISA protocol⁸ that is widely used has only been evaluated with sera. To determine whether plasma could also be tested in this ELISA assay, we conducted a study approved by the Institutional Review Board (IRB# 14-01-XP) of Ross University School of Veterinary Medicine.

Paired samples of plasma and serum were collected by phlebotomists from the medial cubital vein of consenting apparently healthy volunteers with unknown histories of DENV infection. With a single venipuncture, blood was collected into a Vacutainer BD tube containing K3 EDTA 15% solution, 0.081 mL, 12.15 mg (Cat# 366450; Covidien, Minneapolis, MN), and a Monoject™ Blood Collection Tube (Cat# 301710; Covidien) with no additives. The tubes were left at room temperature and, when the blood in the Monoject tubes had clotted, sera and plasma were separated by centrifugation at 4,000 g for 5 minutes and stored at −80°C. Samples were thawed at room temperature and the assay was performed simultaneously for both specimen samples as described elsewhere.⁸ Briefly, 96-well microtiter plates (Immulon II, ImmunoChemistry, cat# 227) were coated with 100 µL of monoclonal antibody 4G2¹¹ and incubated overnight at +4°C in a humidified chamber. All serum and plasma samples were tested in duplicate on the same microtiter plate using eight 4-fold serial dilutions beginning from 1:40 and recombinant COS-1 derived DENV 1–4 virus antigens originally developed, produced, and distributed by the Arbovirus Reference Collection from the CDC, National Center for Emerging and Zoonotic Infectious Diseases, Division of Vector-Borne Diseases, Arboviral Diseases Branch, Diagnostic and Reference Team, Reference Reagents Laboratory^12–14 were added to each sample. After a series of incubations for 1 hour at 37°C, the plates were washed with phosphate-buffered saline in an automatic ELISA plate washer (ELx405™ Microplate Washer; Biotek Instruments, Winooski, VT); 40 μL of horseradish peroxidase-conjugated goat antihuman IgG (Jackson ImmunoResearch Laboratories, West Grove, PA) were added, incubated for 1 hour at room temperature, and washed before the addition of 100 µL of substrate 2,2′-azino-bis(3-ethylbenzothiazoline-6-sulphonic acid) to each well. Following incubation, the plates were read in an optical spectrophotometer (ELx800™ Absorbance Reader; Biotek Instruments) at 405 nm. Each plate contained a negative and a positive control. The optical density (OD) of each dilution (“initial OD”) was subtracted from the corresponding dilution of the negative control to give the “normalized OD,” and mean of the OD from each duplicate plasma and serum sample was calculated for each dilution. Test results in serosurveys are normally expressed as the dilution number at the end-point titration⁸ and defined as the highest dilution in which OD was greater than 0.15.

Both the “initial OD” and the “normalized OD” values for all paired sera and plasma samples were compared using the Wilcoxon test to analyze equality of performance for each sample pair. The precision of the assay was assessed for all the plasma and serum samples separately to obtain an indication of the intra-specimen variability and then compared with each other for agreement between the specimens by calculating the coefficient of variation (CV). The level of precision and agreement between the specimens was also compared by the concordance correlation coefficient (CCC)^15–17 and by observation of Bland–Altman plots.¹⁸

All of the samples collected from 102 individuals showed an end-point titration smaller than 1:2,560. The OD values of dilutions greater than 1:2,560 were only marginally above zero and were removed from further analysis to avoid underestimating variation between specimens with little biological significance. The intra-specimen variability between sera and plasma ODs was higher overall for the plasma samples when compared with that of the serum samples (CV_plasma = 0.102; CV_serum = 0.061), especially at low dilutions (1:40) (Table 1). The variability of plasma at dilution 1:160 and higher was similar to that of sera, and in all cases, the SD among the replicate samples was less than 9.2% of the mean of the replicate samples. This greater variability at low dilutions can be explained by the prozone effect that occurs typically when an excess of antibodies inhibits agglutination reactions due to inhibitory substances or to competition by low-avidity antibody that are washed off in subsequent steps.⁸ This finding was also supported by the CCC. While a coefficient value of one indicates perfect agreement between two sets of samples, the intra-specimen CCC for serum was 0.94 (95% CI: 0.91–0.96) and the intra-specimen CCC for plasma was 0.84 (95% CI: 0.78–0.89), suggesting that the precision for serum specimens was slightly higher than that for plasma. The CCC for both specimens showed little variation at dilution 1:160 and higher, indicating consistency in the performance of the test (Table 1).

Table 1.

Intra-specimen variability: coefficient of variation (CV) and concordance correlation coefficient (CCC) between plasma and serum specimens

Dilution	CVplasma	CVserum	CCCplasma	95% CI	CCCserum	95% CI
d1 (1:40)	0.210	0.078	0.75	0.65–0.82	0.94	0.91–0.96
d2 (1:160)	0.092	0.065	0.84	0.77–0.89	0.94	0.92–0.96
d3 (1:640)	0.061	0.065	0.91	0.87–0.94	0.93	0.90–0.95
d4 (1:2,560)	0.043	0.035	0.88	0.84–0.92	0.95	0.92–0.96
Mean	0.102	0.061	0.84	0.78–0.89	0.94	0.91–0.96

The inter-specimen variability between plasma and sera was nonsignificant overall. There was little difference in the mean OD values for both initial OD and normalized OD values between plasma and serum, and the Wilcoxon signed-rank test indicated that the differences in the OD between replicates and between specimens were nonsignificant at all dilutions (P > 0.05) (Table 2). The mean inter-specimen variability (CV_plasma-serum = 0.081) was lower than the mean intra-specimen variability for plasma (CV_plasma = 0.102), indicating that any variability observed between paired plasma and serum samples was lower than the variability expected because of individual differences. Also, the inter-specimen CCC between paired plasma and serum samples for individuals (Table 2) showed an almost perfect agreement between the OD at dilutions 1:160 and higher (CCC = 0.95, 95% CI: 0.93–0.97). Overall, the concordance between the paired plasma and serum samples showed very good agreement (CCC = 0.92, 95% CI: 0.88–0.95) and the intra-specimen CCC was slightly lower at the lower dilution (1:40) due to the greater variability between plasma replicates. When the samples showing evidence of dengue infection (i.e., OD greater than 0.150) were analyzed, the intra-specimen variation (CV_plasma = 0.098, CV_serum = 0.066, CCC_plasma = 0.71, CCC_serum = 0.93) and the inter-specimen variation (CV = 0.082; CCC = 0.89, 95% CI: 0.83–0.94) were in agreement with the overall results.

Table 2.

Inter-specimen variability of paired samples (n = 102) at each dilution level: Wilcoxon signed-rank test (W), coefficient of variation (CV), and concordance correlation coefficient (CCC) estimated by variance components with 95% CI

Dilution	OD plasma	OD serum	W	P-value	CVplasma-serum	CCCplasma-serum	95% CI
d1 (1:40)	0.263	0.294	4,587	0.14	0.144	0.86	0.81–0.91
d2 (1:160)	0.153	0.142	5,590	0.36	0.078	0.95	0.93–0.97
d3 (1:640)	0.063	0.064	5,104	0.82	0.063	0.95	0.93–0.97
d4 (1:2,560)	0.021	0.023	4,865	0.42	0.039	0.91	0.87–0.94
Mean	0.125	0.131	–	–	0.081	0.92	0.88–0.95

OD = optical density.

The Bland and Altman limits of agreement ranged from −0.14 to +0.14 and most values outside this range corresponded to the first dilution (Figure 1A–C). These plots indicated that the range and distribution of differences between the results from the two sets of specimens was analogous. Also, the plots were consistent with the aforementioned findings and showed that plasma replicates had greater variability than sera replicates as a slightly greater number of the former were outside the 95% limits of agreement.

Figure 1.

Bland-Altman plots. (A) Inter-specimen plasma vs. serum, (B) intra-specimen plasma, and (C) intra-specimen plasma.

In conclusion, a comparison of the use of plasma as a sample matrix in the in-house DENV IgG ELISA showed no significant difference in the diagnostic result from paired plasma and serum specimens from the same individuals. Although there was higher variability between plasma replicates at a low dilution of 1:40 than between sera replicates at the same dilution, this finding was nonsignificant and testing plasma samples in duplicate and averaging the results improves test accuracy and repeatability. Although our samples were not initially characterized as dengue positive, a limitation identified in this study, the separate analysis of positive samples showed no significant deviations, indicating that these conclusions can be extrapolated to dengue-positive samples. Overall, our data suggested that plasma and sera are both suitable samples to use in a noncommercial dengue IgG ELISA distributed by the Arboviral Disease Branch at the CDC.