Assessment of Immunoglobulin M Enzyme-Linked Immunosorbent Assay Ratios to Identify West Nile Virus and St. Louis Encephalitis Virus Infections During Concurrent Outbreaks of West Nile Virus and St. Louis Encephalitis Virus Diseases,Arizona 2015

Abstract

West Nile virus (WNV) and St. Louis encephalitis virus (SLEV) are closely related mosquito-borne flaviviruses that cause clinical disease ranging from febrile illness to encephalitis. The standard for serological diagnosis is immunoglobulin M (IgM) testing followed by confirmatory plaque reduction neutralization test (PRNT) to differentiate the infecting virus. However, the PRNT is time-consuming and requires manipulation of live virus. During concurrent WNV and SLEV outbreaks in Arizona in 2015, we assessed use of a diagnostic algorithm to simplify testing. It incorporated WNV and SLEV ratios based on positive-to-negative (P/N) values derived from the IgM antibody-capture enzyme-linked immunosorbent assay. We compared each sample's ratio-based result with the confirmed WNV or SLEV sample result indicated by PRNT or PCR testing. We analyzed data from 70 patients with 77 serum and cerebrospinal fluid samples, including 53 patients with confirmed WNV infection and 17 patients with confirmed SLEV infection. Both WNV and SLEV ratios had specificity ≥95%, indicating a high likelihood that each ratio was correctly identifying the infecting virus. The SLEV ratio sensitivity of 30% was much lower than the WNV ratio sensitivity of 91%, likely because of higher cross-reactivity of SLEV antibodies and generation of lower P/N values. The standard for serological diagnosis of WNV and SLEV infections remains IgM testing followed by PRNT. However, these results suggest the ratios could potentially be used as part of a diagnostic algorithm in outbreaks to substantially reduce the need for PRNTs.

Introduction

West Nile virus (WNV) and St. Louis encephalitis virus (SLEV) are closely related mosquito-borne flaviviruses endemic in the United States. The clinical spectrum of illness for both viruses is similar, and can range from acute febrile illness through severe encephalitis. While the medical management of infections caused by these viruses does not vary, differentiating them is important to understand their epidemiology, including changing disease patterns over time. Diagnosis is achieved primarily via antibody testing. The immunoglobulin M (IgM) antibody-capture enzyme-linked immunosorbent assay (MAC-ELISA) is commonly used, particularly in public health laboratories, and provides a result expressed as a positive-to-negative (P/N) value. A P/N value is derived by dividing the average optical density reading of the patient's sample against the positive antigen by the average optical density reading of a negative control human sample tested against the positive antigen (Martin et al. 2004). Because of cross-reactivity of flaviviral antibodies in the MAC-ELISA, the more specific confirmatory plaque reduction neutralization test (PRNT) is performed to differentiate the infecting virus. However, the PRNT is a time-consuming test requiring manipulation of live virus and is only conducted at a limited number of public health laboratories. An approach to serologic testing that could accurately and rapidly identify WNV and SLEV infections would be valuable.

Using data from the WNV epidemic of 2002, a previous evaluation was conducted of a diagnostic algorithm using a ratio of MAC-ELISA P/N values to differentiate WNV and SLEV infections (Martin et al. 2004). This was soon after the introduction of WNV into the United States in 1999 and in the setting of many decades of SLEV transmission (Nash et al. 1999, Reisen 2003). Investigators assessed the sensitivity and specificity of a WNV P/N-to-SLEV P/N (W/S) ratio for detection of WNV infection (Martin et al. 2004). Results indicated that use of the ratio as part of a testing algorithm could assist with identifying WNV infections and substantially reduce the need for PRNT.

Since 2002, the epidemiology of WNV and SLEV infections in the United States has changed. Following the introduction of WNV, SLEV disease has been reported at a very low incidence and WNV is now the leading cause of arboviral infection (Curren et al. 2018). Exposure and background immunity to these viruses among the U.S. population likely has changed. It is unknown whether this change might have impacted WNV and SLEV MAC-ELISA P/N values and thus the findings from the previous evaluation. In 2015, the first known concurrent outbreaks of WNV and SLEV disease occurred, focused in Maricopa County, Arizona (Venkat et al. 2015). We used the opportunity of the concurrent outbreaks to assess use of W/S and SLEV P/N-to-WNV P/N (S/W) ratios to diagnose WNV and SLEV infections.

Materials and Methods

Laboratory testing

Sample testing followed normal diagnostic testing procedures. Serum and cerebrospinal fluid (CSF) samples submitted for routine diagnostic testing for acute WNV or SLEV infection were tested by MAC-ELISA at the Arizona State Public Health Laboratory based on methods previously described (Martin et al. 2000). In brief, coating antibody, conjugate, and antigens were independently titrated against a positive control serum sample and standardized by choosing reagent dilutions that yielded an A450 in the range of 0.8 to 1.2. Plates were coated with goat anti-human IgM, incubated overnight at 4°C, and subsequently blocked with blocking buffer. Bound IgM was detected by stepwise addition of either positive or negative antigen, followed by the addition of a flavivirus group-reactive monoclonal antibody conjugated to horseradish peroxidase. Bound conjugate was detected by adding 3,3′5,5′-tetramethylbenzidine substrate, and the A450 was measured with a microplate reader. A P/N value <2.0 was interpreted as negative, a P/N from 2.0 to <3.0 was equivocal, and a P/N ≥3.0 was presumptive positive. Any sample with a positive or equivocal P/N value for either WNV or SLEV and sufficient remaining sample was forwarded to the Centers for Disease Control and Prevention (CDC) for confirmatory PRNT testing.

WNV and SLEV 90% PRNT testing was performed at CDC using methods previously described (Lindsey et al. 1976). A sample with a positive or equivocal IgM against WNV or SLEV and a PRNT titer against WNV ≥4-fold higher than the titer against SLEV was considered a confirmed WNV infection according to standard interpretations used at CDC's Arboviral Diseases Branch Reference Laboratory (Johnson et al. 2009). A sample with a positive or equivocal IgM against WNV or SLEV and a PRNT titer against SLEV ≥4-fold higher than the titer against WNV was considered a confirmed SLEV infection. A sample with a positive or equivocal IgM against WNV or SLEV and a less than fourfold difference between the WNV and SLEV titers was considered an unspecified flavivirus infection.

PCR testing is not routinely performed for diagnosis of flaviviral infections as humans usually have low levels of transient viremia and detectable neutralizing antibodies by the time of clinical presentation. However, PCR testing was performed on samples if clinically indicated (e.g., immunosuppressed patients). Detection of WNV or SLEV RNA in a sample indicated a confirmed WNV or SLEV infection, respectively.

Definitions

A W/S ratio was calculated by dividing a sample's WNV P/N value by its SLEV P/N value. Similarly, a S/W ratio was calculated by dividing a sample's SLEV P/N value by its WNV P/N value. Based on these ratios, infections were defined as follows: (1) A W/S ratio of ≥3.0 was defined as a WNV infection, (2) A S/W ratio of ≥3.0 was defined as an SLEV infection, and (3) Both ratios <3.0 were defined as an unspecified flavivirus infection. We selected a ratio cutoff of ≥3.0 because in the previous 2002 evaluation this W/S ratio cutoff was shown to have a favorable balance of a specificity of 90% and a sensitivity of 81% for detection of WNV infection (Martin et al. 2004).

Data management and analysis

P/N data for samples from patients with confirmed WNV or SLEV infection by PRNT or PCR testing were used in the analysis. Samples were excluded if PRNT testing indicated an unspecified flavivirus infection, or if complete laboratory data were unavailable.

We collected deidentified data, including sample type, interval between illness onset and sample collection, and WNV and SLEV test results. We calculated the W/S and S/W ratios and compared those with the WNV or SLEV result by PRNT or PCR testing. Ratio accuracy was calculated by dividing the number of ratios that correctly indicated WNV or SLEV infection by the total number of WNV- or SLEV-confirmed cases. For the overall ratio accuracy calculation, if cases had both serum and CSF samples and one ratio indicated a WNV or SLEV infection and the other indicated an unspecified flavivirus infection, we used the ratio indicating either WNV or SLEV infection, respectively. We determined sensitivity and specificity for the ratio in CSF and serum samples. For cases with both serum and CSF samples, consistency between the ratios was evaluated by comparing the infection indicated by each sample ratio. We performed logistic regression to examine the relationship between the disease onset and sample collection interval and the accuracy of the ratio. All data were analyzed using Excel version 2016 (Microsoft, Redmond, WA) and SAS statistical software version 9.4 (SAS Institute, Cary, NC). Data were collected as part of a public health investigation that underwent institutional review and was determined not to be research.

Results

Fifty-three confirmed WNV disease cases and 17 confirmed SLEV disease cases were included. For WNV disease cases, confirmatory testing was by neutralizing antibody testing (N = 50; 12 with both WNV and SLEV neutralizing antibodies but WNV PRNT titer ≥4-fold higher than the SLEV titer, and 38 with WNV neutralizing antibodies alone) and PCR (N = 3); all of the SLEV cases were confirmed by neutralizing antibody testing (6 with both SLEV and WNV neutralizing antibodies but SLEV PRNT titer ≥4-fold higher than the WNV titer, and 11 with SLEV neutralizing antibodies alone). Of the 53 confirmed WNV disease cases, 30 (57%) had only a serum sample available for evaluation, 19 (36%) had only CSF, and 4 (8%) had both serum and CSF. Of the 17 confirmed SLEV disease cases, 10 (59%) had only a serum sample available for evaluation, 4 (24%) had only CSF, and 3 (18%) had both serum and CSF. Fifty additional patients with samples submitted for testing were excluded, including 48 who lacked complete laboratory data and 2 who had unspecified flavivirus infections.

The median and range of the interval between illness onset and sample collection was the same for both WNV and SLEV disease cases (median 5 days; range 0–44 days). Overall, among the 70 WNV and SLEV cases, the W/S and S/W ratios correctly identified the infecting virus in 54 (77%) cases. Among 53 confirmed WNV disease cases, 49 (92%) cases were classified correctly as WNV infections by the W/S ratio, and among 17 confirmed SLEV disease cases, 5 (29%) cases were classified correctly as SLEV infections by the S/W ratio.

Accuracy of ratio of P/N values for diagnosis of infection in serum samples

Among 34 confirmed WNV cases with serum samples, 31 (91%) IgM antibody P/N value ratios indicated WNV infection (Table 1). Three (9%) serum ratios suggested unspecified flavivirus infection; although the ratio of P/N values was higher for WNV than SLEV, it was <3.0.

Table 1.

Accuracy of Ratio of Positive-to-Negative Values for Diagnosis of Infection by Sample Type

	WNV cases^a				SLEV cases^a
	Serum (N = 34)		CSF (N = 23)		Serum (N = 13)		CSF (N = 7)
	N	%	N	%	N	%	N	%
W/S ratio ≥3	31	91	21	91	1	8	0	0
S/W ratio ≥3	0	0	0	0	5	38	1	14
W/S and S/W ratio <3	3	9	2	9	7	54	6	86

Confirmed with neutralizing antibody or PCR testing.

CSF, cerebrospinal fluid; P/N, positive-to-negative; SLEV, St. Louis encephalitis virus; S/W, SLEV P/N-to-WNV P/N; W/S, WNV P/N-to-SLEV P/N; WNV, West Nile virus.

Among 13 confirmed SLEV cases with serum samples, 5 (38%) IgM antibody P/N value ratios indicated SLEV infection. One (8%) ratio indicated WNV infection. Seven (54%) serum ratios suggested unspecified flavivirus infections; among these, four had a ratio of P/N values that was higher for SLEV than WNV but <3.0, and three had a ratio of P/N values that was higher for WNV than SLEV but <3.0.

Accuracy of ratio of P/N values for diagnosis of infection in CSF samples

Among 23 confirmed WNV cases with CSF samples, 21 (91%) IgM antibody P/N value ratios indicated WNV infection (Table 1). Two (9%) CSF P/N ratios suggested a flavivirus unspecified infection; although the ratio of P/N values was higher for WNV than SLEV, it was <3.0.

Among seven SLEV cases with CSF sample data, one (14%) CSF ratio indicated SLEV infection. Six (86%) CSF ratios suggested an unspecified flavivirus infection; among these, for four, the ratio of P/N values was higher for SLEV than WNV, but <3.0; and for two, the ratio of P/N values was higher for WNV than SLEV, but <3.0.

Sensitivity and specificity of ratios of P/N values for diagnosis of infection with all samples

Among the 53 confirmed WNV cases, there were 57 total samples, including 34 (60%) serum samples and 23 (40%) CSF samples. Overall, the W/S ratio had a sensitivity of 91% (confidence interval [95% CI] 81–96) and specificity of 95% (95% CI 76–99) for diagnosis of WNV infection (Table 2).

Table 2.

Sensitivity and Specificity of Ratios to Identify West Nile Virus and St. Louis Encephalitis Virus Infections

	WNV infections^a			SLEV infections^b
	N	%	95% CI	N	%	95% CI
Sensitivity	52/57	91	81–96	6/20	30	15–52
Specificity	19/20	95	76–99	57/57	100	94–100

WNV infections confirmed with neutralizing antibody or PCR testing; in specificity calculations, negatives (non-WNV infections) were confirmed SLEV infections.

SLEV infections confirmed with neutralizing antibody testing; in specificity calculations, negatives (non-SLEV infections) were confirmed as WNV infections.

CI, confidence interval.

Among the 17 confirmed SLEV cases, there were 20 total samples, including 13 (65%) serum samples and 7 (35%) CSF samples. Overall, the S/W ratio had a sensitivity of 30% (95% CI 15–52) and specificity of 100% (95% CI 94–100) for diagnosis of SLEV infection.

Consistency of the ratio of P/N values in serum and CSF samples

Among the seven cases with serum and CSF samples, one WNV case had discordant results (i.e., the CSF ratio indicated a WNV infection, but the serum ratio suggested an unspecified flavivirus infection). The other six cases had concordant results.

Interval between disease onset and sample collection and ratio accuracy

There was no correlation between timing of sample collection and the accuracy of the ratio of P/N values (p = 0.32).

Discussion

Analysis of data from the 2015 concurrent WNV disease and SLEV disease outbreaks showed that specificity of W/S and S/W ratios was ≥95%. W/S ratios of ≥3.0 were >90% sensitive in identifying confirmed cases of WNV infection, but S/W ratios of ≥3.0 had low sensitivity for diagnosis of SLEV infection. The timing of sample collection did not have an impact on accuracy of the ratio, and ratio accuracy was similar for serum and CSF samples.

Because of cross-reactivity among flaviviruses, IgM ELISA results frequently are positive for both WNV and SLEV in patients with infections caused by either of these viruses. ELISA results alone generally are insufficient for correct classification of the infecting virus and the accepted standard for serologic diagnosis of WNV and SLEV infections remains IgM testing followed by a confirmatory PRNT. However, in future outbreaks, W/S and S/W ratios could be used to reduce the amount of PRNT testing required. Once the outbreak is confirmed by PRNT testing, consideration could be given to limiting use of this test to only those samples with W/S and S/W ratios <3.0, given the high likelihood a patient with WNV or SLEV infection diagnosed using a ratio ≥3.0 truly has that infection. Among the subset of samples included in this analysis from the 2015 outbreak, 1 patient with SLEV infection would have been misclassified as having WNV infection, but only 18 (23%) of the 77 samples would have needed PRNT testing, substantially reducing the time and effort required of the diagnostic laboratories. An additional benefit would be a presumed diagnosis for patients in whom sample quantity is insufficient for additional confirmatory testing. The proportion of WNV and SLEV infections could have an effect on the positive predictive value of the W/S ratio, but the impact is likely to be small (Martin et al. 2004).

Despite the changes in WNV and SLEV epidemiology during the last 15 years, our findings are similar to the previous assessment of this diagnostic algorithm in 2002 (Martin et al. 2004). In that investigation a W/S ratio of ≥3.0 in serum was 90% specific and 81% sensitive in identifying confirmed cases of WNV infection. The authors similarly concluded that use of a W/S ratio as part of a testing algorithm could simplify diagnostic testing procedures during outbreaks.

The main weakness of the diagnostic approach was the poor sensitivity of the S/W ratio for SLEV infections. There are likely two reasons for its lower accuracy compared with the W/S ratio. The P/N values generated in the MAC-ELISA for SLEV infection are typically lower than P/N values generated with WNV infection, and SLEV IgM antibodies are more cross-reactive in the CDC MAC-ELISA (Martin et al. 2002).

There are two limitations to this analysis. First, the findings are based on a relatively small number of samples. Despite the 2015 outbreak being the largest SLEV disease outbreak in over a decade, there was still a relatively small number of SLEV disease cases, and samples from only 17 confirmed SLEV cases were available for the analysis. In addition, there were 48 WNV or SLEV samples that could not be included in the analysis because of incomplete laboratory data. Most were missing confirmatory test results, likely because there was insufficient sample remaining for PRNTs, or the sample was initially tested at a laboratory unaware of the need to forward the samples for PRNTs. Second, we included only confirmed WNV and SLEV infections in the analysis; including unspecified flavivirus infections that might have lowered the ratio's specificity.

While medical management does not vary between these two flaviviral diseases, differentiating them is important in understanding their epidemiology, especially in the context of concurrent outbreaks. Given that simultaneous transmission of the viruses is both possible and unpredictable, the use of W/S and S/W ratios could potentially be used to public health benefit to streamline testing during future outbreaks.

Footnotes

Acknowledgments

Jennifer Adair, Alice Carrigan, Helen Houser, Tammy Kafenbaum, Ronald Klein, Kenneth Komatsu, Lia Koski, Nicole LaMantia, Janeen Laven, Tracey Lussier, James Matthews, William McConahey, Margaret Mills, Kathryn Putman, Karen Rose, Andrew Strumpf, Tammy Sylvester, and Carrie Walker.

Disclaimer

The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the U.S. Centers for Disease Control and Prevention.

Author Disclosure Statement

No conflicting financial interests exist.

Funding Information

No funding was received for this article.

References

Curren

, Lindsey

, Fischer

, Hills

. St. Louis encephalitis virus disease in the United States, 2003–2017. Am J Trop Med Hyg, 2018; 99:1074–1079.

Johnson

, Kosoy

, Hunsperger

, Beltran

, et al. Evaluation of chimeric Japanese encephalitis and dengue viruses for use in diagnostic plaque reduction neutralization tests. Clin Vaccine Immunol, 2009; 16:1052–1059.

Lindsey

, Calisher

, Matthews

. Serum dilution neutralization test for California group virus identification and serology. J Clin Microbiol, 1976; 4:503–510.

Martin

, Biggerstaff

, Allen

, Johnson

, et al. Use of immunoglobulin M cross-reactions in differential diagnosis of human flaviviral encephalitis infections in the United States. Clin Diagn Lab Immunol, 2002; 9:544–549.

Martin

, Muth

, Brown

, Johnson

, et al. Standardization of immunoglobulin M capture enzyme-linked immunosorbent assays for routine diagnosis of arboviral infections. J Clin Microbiol, 2000; 38:1823–1828.

Martin

, Noga

, Kosoy

, Johnson

, et al. Evaluation of a diagnostic algorithm using immunoglobulin M enzyme-linked immunosorbent assay to differentiate human West Nile virus and St. Louis encephalitis virus infections during the 2002 West Nile Virus epidemic in the United States. Clin Diagn Lab Immunol, 2004; 11:1130–1133.

Nash

, Mostashari

, Fine

, Miller

, et al. West Nile Outbreak Response Working Group, 2001. The outbreak of West Nile virus infection in the New York City area in 1999. N Engl J Med, 1999; 344:1807–1814.

Reisen

. Epidemiology of St. Louis encephalitis virus. Adv Virus Res, 2003; 61:139–183.

Venkat

, Krow-Lucal

, Hennessey

, Jones

, et al. Concurrent outbreaks of St. Louis encephalitis virus and West Nile virus disease—Arizona, 2015. MMWR Morb Mortal Wkly Rep, 2015; 64:1349–1350.