For each sample, the MFI value obtained when the samples were reacted against each of the viral antigens (V) was divided by the MFI value obtained for the corresponding negative antigen reaction (N) to yield a V/N value, which was used in further computations. The negative serum control pool gave MFIs of <350 when the samples were reacted against the viral antigens, and positive controls were always >1,000 MFIs when they were reacted against the viral antigens. Negative samples gave MFIs generally in the range of 50 to 350 and positive samples gave MFIs generally in the range of 1,000 to 25,000 when they were reacted against the viral antigens.
The ROC plots generated for each virus (Fig. ) gave AUCs of >90%, indicating the good discrimination abilities of the tests. The initial results indicated that a small number of species were susceptible to background reactivity with the negative antigens in the WN/SLE virus biotin-MIA when PBS-1% BSA was used as the final 1:10 diluent. This problem was mitigated by the use of Candor Low Cross buffer and did not alter the results obtained for the other species. For each species exhibiting high background values, the mean MFIs for WN virus- and SLE virus-negative antigens with the original diluent versus the mean MFIs for the WN virus- and SLE virus-negative antigens with the Candor Low Cross buffer is as follows: for Alligator mississippiensis, 486 and 351 versus 46 and 46, respectively; for Ardea alba, 1,556 and 901 versus 8 and 7, respectively; for Equus caballus, 3,917 and 5,175 versus 159 and 151, respectively; for Megascops asio, 929 and 370 versus 230 and 160, respectively; for Petrochelidon pyrrhonota, 950 and 668 versus 50 and 47, respectively; and for Quiscalus quiscula, 1,857 and 1,614 versus 427 and 403, respectively. Therefore, this reagent was adopted as the standard in the protocol. Approximately equal numbers of antibody-positive and -negative samples were used in the development of these tests, with at least half in the WN/SLE virus biotin-MIA originating from wild-caught species. The biotin-MIA results compared to those of PRNT are shown in Fig. for the WN/SLE virus biotin-MIA and Fig. for the EEE virus biotin-MIA.
FIG. 1. Receiver operator characteristic curves for WN, SLE, and EEE viral antigens in the biotin-MIAs. The sensitivities and the specificities for each antigen were calculated by using the results of PRNT as the gold standard. Areas under the curve and 95% (more ...)
FIG. 2. V/N ratios for samples used to develop the tests are shown. •, PRNT negative; +, PRNT positive; dashed vertical lines, positive V/N cutoff for that antigen. Species are listed in the following order: birds (the normal type above the boldface (more ...)
The sensitivity and specificity data for each virus, as plotted on the ROC curves, were used to derive V/N cutoff values above which the samples were deemed positive for the respective viruses. Because all the sensitivity and specificity values were >90%, cutoffs were chosen such that the sensitivity was equal to the specificity. The cutoff values were 10.00 for WN virus, 10.23 for SLE virus, and 8.97 for EEE virus. Figure shows the biotin-MIA results for all viruses species tested compared with those of the PRNTs by application of the calculated cutoffs for each virus. The rates of false-positive and false-negative results computed by using these cutoffs were 5.5% and 5.8%, respectively, for WN virus; 0.0% and 0.0%, respectively, for SLE virus; and 0.0% and 4.0%, respectively, for EEE virus. The overall accuracy of the WN/SLE virus biotin-MIA was 94.4%, and the overall accuracy of the EEE virus biotin-MIA was 98.0%. Cross-validation estimates of the predictive errors mirrored these empirical results, with prediction errors given as overall, false-positive, and false-negative results, as follows: for WN virus, 5.5%, 6.4%, and 4.9%, respectively; for SLE virus, 1.8%, 0.0%, and 4.8%, respectively; and for EEE virus, 6.0%, 8.3%, and 3.8%, respectively. Repeatability was measured to be 100% (95% CI, 80.6 to 100.0%) for 16 samples tested on the same plate at the same time, and reproducibility was measured to be 95.2% (95% CI, 77.3 to 99.8%) for 21 samples tested on different days, on different plates, and on different instruments. The last two parameters were for the WN/SLE virus biotin-MIA only by use of a variety of samples that were known to be WN or SLE virus antibody positive or negative. For the WN/SLE virus biotin-MIA, the samples used comprised 99 samples from WN virus infections, 17 samples from SLE virus infections, and 140 negative samples, as determined by PRNT. Two samples from confirmed WN virus infections gave higher V/N values for the SLE virus antigen, leading to false-positive results for SLE virus. Overall, 27% of all positive samples exhibited reactions to the heterologous antigen in this test.
The 239 WN virus antibody-positive and -negative samples were used to identify a potential equivocal zone surrounding the V/N cutoff value of 10 for the WN virus component of the WN/SLE virus biotin-MIA. Results inconsistent with those of PRNT had V/N values of <5 for 3/131 samples, 5 < 10 for 3/7 samples, 10 < 15 for 3/15 samples, 15 < 25 for 4/19 samples, and >25 for 1/67 samples. For these samples, the total proportion of misclassified results was 5.9%, with 71% of these falling between V/N values of 5 and 25.
Quantified PRNT-positive results were available for 72 of the serum samples taken from individuals with WN virus infection. Of these, 8 were of low titer (1:10/1:20), 18 were of moderate titer (1:40/1:80), and 46 were of high titer (1:160 or above). Positive results for WN virus were obtained in the WN/SLE virus biotin-MIA for 75%, 94%, and 96% of the samples with low, moderate, and high titers, respectively.
The cross-reactivity results are shown in Fig. . Both the WN and SLE viral antigens in the WN/SLE virus biotin-MIA (Fig. ) showed significant cross-reactivity against sera from humans with flavivirus infections: for JE virus, 1 of 6 samples, and for DEN virus, 4 of 10 samples. No cross-reactivity was seen with sera with antibodies to the alphaviruses EEE virus (0/10) and CHIK virus (0/6) or against the bunyavirus LAC virus (0/10). Human sera positive for anti-CHIK virus antibodies showed minimal cross-reactivity (1/10), and chicken anti-WN virus antibodies showed no cross-reactivity (0/13) in the EEE virus biotin-MIA (Fig. ).
FIG. 3. Cross-reactivity of the WN/SLE virus biotin-MIA (a) and the EEE virus biotin-MIA (b) with antisera to other confirmed arboviral infections. No dual infections were included in the analyses. The cutoffs for a positive reaction were V/N ratio values of (more ...)
Validation of the WN virus portion of the WN/SLE virus biotin-MIA was performed with samples that had previously been tested for antibodies to WN virus by PRNT but that were not included in the cutoff determination. Samples were not available to do the same for SLE and EEE virus infections. By using the V/N cutoff value of 10.00 for WN virus, the WN/SLE virus biotin-MIA gave a rate of false-positive results of 0% (95% CI, 95.6 to 100.0%; 84/84 negative samples were correctly identified) and a rate of false-negative results of 10% (95% CI, 78.6 to 95.6%; 45/50 samples positives for WN virus were correctly identified).
A previous study compared specimens from WN virus-infected humans in a WN virus IgM ELISA and an SLE virus IgM ELISA (12
). It was consistently found that the optical densities (OD) of the test specimen divided by the OD of the negative control serum sample obtained when it was reacted against the WN viral antigen (positive/negative [P/N] ratio) was at least three times greater than the P/N ratio obtained when the sample was reacted against SLE viral antigen. The reverse was not true of SLE viral infections. A paired t
test for the WN/SLE virus biotin-MIA during the development phase showed that the V/N values for the WN viral antigens were consistently twofold or greater than those obtained with the SLE viral antigen when WN virus was the infecting virus (P
< 0.001; extrapolated 95% CI, 2.2 to 3.7). In humans, if SLE virus was the infecting virus, a twofold difference in the V/N values was not observed, although the V/N values were consistently higher for SLE virus infections for the limited number of samples tested. For chickens, however, SLE virus infections were distinguishable with a minimum SLE virus-to-WN virus V/N ratio of 6:1 (P
< 0.001; extrapolated 95% CI, 8.1 to 11.9).
To confirm that IgM was also detectable by the WN/SLE virus biotin-MIA, 46 human serum samples that were IgM positive, IgG negative, and PRNT positive for WN virus in previous tests were analyzed by the WN/SLE virus biotin-MIA (data not shown). Forty-three were found to be positive (93%), and the remaining three that tested negative were from IgM-equivocal samples.