Criteria for Selection of Cases

Data were collected from laboratory submission forms that accompanied blood samples from equids suspected of having WNV during 2003. Equids had to have lived in Colorado in 2003 and been tested for IgM directed against WNV by use of IgM capture ELISA to be included in the study. Four veterinary diagnostic laboratories in Colorado performed the IgM capture ELISA test⁷ for detection of WNV antibody in 2003: Rocky Mountain Regional Animal Health Laboratory in Denver, Colorado State University Veterinary Diagnostics Laboratory in Fort Collins, Western Slope Animal Diagnostics Laboratory in Grand Junction, and Rocky Ford Animal Diagnostics Laboratory in Rocky Ford. The cost of the test at each of the laboratories was approximately the same. Copies of submission forms were acquired through the state veterinarian and the individual veterinary diagnostic laboratories.

Procedures

The submission form submitted with each blood sample was completed by the veterinarian requesting the test, and their contact information was included on the submission form. Information requested on each form included signalment (age, sex, and breed); clinical signs with date of onset; history of vaccination against WNV, western equine encephalitis virus, eastern equine encephalitis virus, and Venezuelan equine encephalitis virus; date of sample collection; location of the equid; and clinical status at time of testing. The owner's contact information was requested on the submission form. All available data were entered into a software system.^b,c

History of vaccination against WNV (yes, no, or unknown) and the date of vaccination for 2 initial WNV vaccines as well as the date of a booster vaccination were recorded. Clinical signs compatible with WNV infection, including muscle fasciculation, ataxia or paralysis, hyperexcitability, convulsions or twitching, recumbency, incoordination, depression, and lack of appetite, were included on submission forms. Forms also included a space to record other clinical signs and a space to indicate that an equid was not having clinical signs of WNV infection at the time of testing.

Criteria for determining that an equid had WNV infection included a positive equine anti-WNV IgM capture ELISA result and 1 or more clinical signs compatible with WNV infection (case equid).⁹ Equids with a positive test result in which lack of appetite was the only reported clinical sign were not considered as having WNV infection. Equids with clinical signs compatible with WNV infection that had negative test results were considered as controls. Breeds of equids tested were classified as Appaloosa, Arabian, Paint, Quarter Horse, Thoroughbred, and other (including ponies, donkeys, and mules).

Equids were considered fully vaccinated (satisfying the vaccination recommendation) against WNV if they received either 2 WNV vaccines 3 to 6 weeks apart in 2002 and at least 1 booster vaccination in 2003 or 2 vaccines 3 to 6 weeks apart in 2003. Partial vaccination included equids that received at least 1 vaccine at any time. Equids were considered unvaccinated if they were reported as never having received a vaccination against WNV. Equids with unknown WNV vaccination status were excluded from any analysis that included WNV vaccination status as a factor.

The submission form was considered complete unless data important for analysis, such as signalment, WNV vaccination status, or geographic location of the equid, were missing. One of the authors (JMT) contacted the submitting veterinarian via telephone or fax to collect missing data. Submission forms with incomplete data after follow-up were excluded from statistical analysis.

Statistical analysis—For purposes of statistical analysis, regardless of test result, equids were considered to have clinical signs compatible with WNV infection if they had 1 or more clinical signs (muscle fasciculation, ataxia or paralysis, hyperexcitability, convulsions or twitching, recumbency, incoordination, and depression), with the exception of lack of appetite.

Descriptive statistics were calculated from summarized data by use of Pearson χ² test and are reported as ORs with associated 95% CIs and a P value^d; values of P < 0.05 were considered significant. Factors potentially associated with being a case of WNV were screened by use of univariable logistic regression and were eligible for inclusion in multivariable logistic regression^e if they improved model fit (P < 0.10 on likelihood ratio test in univariable logistic regression). The best-fitting factor on univariable analysis was used as the basis for the multivariable model. Additional factors were added in decreasing order of significance and were retained if they significantly (likelihood ratio testing; P < 0.05) improved model fit. Factors were considered to be significantly associated with being a case of WNV if the 95% CI for the factor did not include 1.

Results of statistical analysis are reported as ORs with associated Wald χ² P values and 95% CIs. For continuous variables in which a Wald χ² was not calculated, the P value for the likelihood ratio is reported. Age was examined as both a continuous variable and in quartiles. Breed was examined as a categoric variable. Sex was considered a dichotomous outcome (male vs female). West Nile virus vaccination status was examined as a categoric outcome (eg, fully vaccinated, partially vaccinated, and unvaccinated). Equids with unknown vaccination status were excluded from the portion of the multiple logistic regression model concerning vaccination status but were included in the portion of the model concerning sex. The duration from onset of clinical signs until sample collection was considered as both a continuous variable and in categories (0 to 1 days, 2 to 4 days, and > 4 days). Severity of clinical signs was considered in categories of mild, moderate, and severe. Equids were considered to have mild clinical signs if they were reported to have only signs of depression or signs of depression and lack of appetite. Clinical signs were considered severe if the equid was reported to be recumbent and had ataxia or convulsions. Equids were considered to have moderate clinical signs if they were reported to have any other combination of clinical signs. Vector season categories were based on the likely presence of competent vectors. The nonvector season was considered as January 1, 2003, through April 15, 2003, and November 1, 2003, through December 31, 2003. The early vector season was considered as April 16, 2003, through June 30, 2003; the mid vector season was considered as July 1, 2003, through August 31, 2003; and the late vector season was considered as September 1, 2003, to October 31, 2003.

When paired samples were submitted for testing, data associated with the first sample submitted were included in the statistical analysis. One sample was randomly chosen for exclusion in paired samples in which the order of submission was unknown.

Equine population data for Colorado were acquired from the USDA National Agriculture Statistics Service 2002 Census of Agriculture. This data set describes population density of horses and ponies on premises defined as “farms” by the USDA at the county level.¹⁰ On the basis of zip code for the location of the horse, mapping and spatial analysis was performed for all tested equids, confirmed cases of WNV infection in horses and ponies, and equine population at risk by county and by 7 multicounty regions^b,f (Figure 1).

View Full Size

Regional case rate of WNV infection in horses and ponies in Colorado in 2003.

Citation: Journal of the American Veterinary Medical Association 228, 3; 10.2460/javma.228.3.414

• Download Figure
• Download figure as PowerPoint slide

Results

There were 1,125 samples submitted for IgM capture ELISA. The Rocky Mountain Regional Animal Health Laboratory had the largest number of submissions during 2003. Samples from 21 equids were tested twice; thus, a total of 1,104 equids were tested. Fifty-one of 1,104 (4.6%) submission forms lacked information concerning history of vaccination against WNV. All 51 equids with missing WNV vaccination history reportedly had clinical signs compatible with WNV infection. Sixty-nine (6.3%) of 1,104 submission forms lacked information concerning clinical signs. After attempts were made to collect missing data, approximately 10% of records remained incomplete, and these submission forms were excluded from further data analysis.

Of the 1,125 serum samples tested, 634 (56.4%) had a positive IgM capture ELISA result, 19 (1.7%) had a nonspecific test result, and 472 (42.0%) had a negative test result. Only those equids that were reported to have clinical signs compatible with WNV infection (n = 889) were included in the subsequent detailed data analysis and mapping. Twenty-six equids were reported to be clinically normal, 13 had clinical signs that were not compatible with WNV infection, and clinical signs were not reported on 69 laboratory submission forms.

Of the equids tested that had complete WNV vaccination history and clinical signs compatible with WNV infection (n = 838), 64% were considered as case equids on the basis of clinical signs and positive IgM ELISA results; 36% of equids tested negative for WNV. Among case equids with WNV, most had not been vaccinated against WNV infection. Among equids testing negative for WNV, distribution among the 3 categories of vaccination was approximately equal (Table 1).

Of the 1,017 equids tested that had clinical signs at the time of sample collection, 113 were dead at the time of sample submission. Determination of the long-term outcome of equids with WNV infection was not pursued as part of our study. Vaccination history was known for 111 of 113 equids that died. Of those 111 equids, 33 tested negative for WNV; 9 were reportedly fully vaccinated, 9 had been partially vaccinated, and 15 had no history of vaccination against WNV. Of those equids that died, 80 met the criteria for case equids. Of those 80 case equids, 8 had been fully vaccinated, 9 had been partially vaccinated, and 63 (78.8%) had no history of vaccination against WNV. Among equids that died, those that were classified as case equids were more likely to be unvaccinated than those not classified as case equids (OR, 4.0; 95% CI, 1.63 to 9.58; P = 0.002).

The mean age of equids tested was 11.4 years (range, 2 months to 40 years), and the mean age of case equids was 11.9 years (range, 3 months to 35 years). Approximately equal numbers of males and females were tested for WNV (Table 1). Sex of equids was not associated with vaccination status (OR, 1.1; 95% CI, 0.8 to 1.4; P = 0.53). Most equids tested were Quarter Horses; donkeys and mules accounted for < 1% of the equids tested.

Factors significantly associated with testing positive for WNV infection, compared with testing negative for WNV infection, included having not been vaccinated against WNV, severity of clinical signs, region in Colorado, duration of illness, and season of illness (Table 2). Factors that were not significantly associated with testing positive for WNV infection, compared with control equids, included age, breed, and diagnostic laboratory performing IgM capture ELISA.

The initial sample was submitted for IgM capture ELISA on January 8, with the first positive test result on June 25, 2003. The final sample was submitted on November 12, which was also the last positive test result for the year. Both case equids and sample submissions followed a curve, which began to increase the week of June 22 and peaked the week of August 17, 2003 (Figure 2). There was no significant difference in the proportion of case equids versus control equids by diagnostic laboratory.

View Full Size

Temporal distribution of IgM capture ELISA results in equids with clinical signs compatible with WNV infection in Colorado during 2003.

Citation: Journal of the American Veterinary Medical Association 228, 3; 10.2460/javma.228.3.414

• Download Figure
• Download figure as PowerPoint slide

Discussion

The 2003 equine WNV data from Colorado represent the first vector season in which equids would likely have been vaccinated in sequential years against WNV and had adequate time to develop an immune response prior to the onset of exposure to infection. Results of 1 study² indicate that vaccination against WNV decreases the likelihood of severe disease among equids with WNV infection. However, to the authors’ knowledge, factors associated with equids with clinical signs compatible with WNV infection testing positive via IgM capture ELISA versus equids with clinical signs compatible with WNV infection testing negative via IgM capture ELISA have not been examined.

In our study, females were more likely than males to be case equids. Conflicting reports^2,9,11,12 exist concerning sex among equids with WNV infection. However, results of 1 study² indicate that among cases of WNV in equids, females were at an increased risk for developing severe disease and death. Possible explanations include a sampling bias toward 1 sex of equid being more likely to be tested for the disease on the basis of value or severity of clinical signs. In our study, sample submissions from males and females were not significantly different from the estimated Colorado equine population.¹³ Results of several studies^2,9,11 indicate that females are more likely than males to develop WNV infection. Detailed investigation into the severity of clinical signs in male and female equids for which samples are submitted for diagnostic testing would help determine whether clinical signs in 1 sex are consistently more severe, which may predispose those equids to test positive for WNV infection.¹² Results of a study¹² suggest that females are more likely to become infected and die from WNV infection. Sex was not found to be a factor in the mounting of an immune response through natural exposure or vaccination.¹⁴ When considering results of that study and our results indicating that sex was not associated with vaccination status, there is little explanation for the finding that females are more likely than males to be case equids. Differences in management practices or other confounding factors may have accounted for this difference.

Compared with horses that had WNV infection in 2002, fewer horses were fully or partially vaccinated against WNV infection in 2003.² A large number of equids that had severe clinical signs compatible with WNV infection were not considered as case equids because they had negative IgM ELISA results. There are several possible explanations for the negative IgM ELISA results in equids with severe neurologic signs during the WNV epidemic. First, another disease such as other vector-borne viral encephalitides and zoonoses, metabolic disorders, organ failure, or intoxications may have caused illness in those equids. The possibility of a vector-borne disease was indicated by the seasonal pattern of equids with clinical disease. Surveillance for other likely viral encephalitides indicated that western equine encephalitis virus and other zoonotic diseases such as rabies virus were not circulating in Colorado during the 2003 vector season.¹⁵

Second, clinical signs in these equids may have been caused by WNV infection, but samples were obtained when IgM antibodies were not present. In humans, evaluation of viremia in relation to the subsequent immune response indicates that there is a brief period after the onset of clinical signs when IgM concentrations are not sufficient to result in a positive ELISA result.^g Third, these equids may have had clinical signs attributable to WNV infection but did not develop an IgM response adequate for detection via IgM capture ELISA. Results of 1 study¹¹ indicate that some horses will test positive for WNV via plaque reduction neutralization test but will not develop sufficient IgM responses to be detected by IgM capture ELISA at the standard 1:400 dilution, even when paired samples are used.¹¹

The IgM subclass of immunoglobulin is the primary immunoglobulin produced in the immune response of an individual on initial exposure to an antigen.¹⁶ Hypothetically, in the first year that WNV is present in a geographic area in which the equine population is primarily unvaccinated or only partially vaccinated and naïve to exposure to the wild form of the virus, most infected equids may develop an adequate IgM response; therefore, the ELISA results would be positive when they are infected with the wild form of the virus.⁵

To the authors’ knowledge, studies evaluating how previous vaccination affects the immune profile and subsequent antibody production when naturally exposed to the wild-type WNV have not been performed. Vaccination could affect an equid's ability to mount an adequate IgM response for detection of WNV by use of the current dilution of 1:400 for the IgM ELISA. Thus, it would be hypothesized that in most horses, some immune response would be induced after exposure to an inactivated strain of WNV through vaccination.⁶ Therefore, any subsequent exposure to the antigen whether through natural exposure or vaccination should be considered as a secondary exposure. In a secondary exposure, only a small amount of IgM antibodies may be produced.¹⁶ The IgG antibodies are made earlier and at much higher concentrations in response to a secondary exposure to an antigen than in response to an initial exposure.¹⁶ This would explain why many equids vaccinated against WNV but with neurologic signs had negative IgM capture ELISA results and thus were not considered as case equids. However, if an IgG ELISA alone was used as a diagnostic tool, it would not permit differentiation of an immune response from vaccination verses that from natural exposure. Further research is needed to help determine whether the IgM capture ELISA presently being used is adequate to detect disease in equids receiving multiple WNV vaccines.

There are unavoidable limitations when performing a retrospective study, one of which is incomplete reporting. In our study, compliance with completion of submission forms and follow-up with veterinarians to collect missing data was reasonably good. As a result, only 10% of the total submission forms had missing data. There were a limited number of paired serum samples submitted; however, because of the retrospective study design, this was not avoidable. Estimating case fatality rate was not the main focus of our study, and as a result, it was difficult to make conclusions because long-term follow-up with veterinarians submitting samples or owners of equids tested concerning outcome of disease in equids was not performed. To the authors’ knowledge, the study reported here represents the first effort to evaluate confirmed cases of WNV in equids and in equids with similar clinical disease that had negative results for WNV via IgM capture ELISA. Reasons for such a large number of equids with neurologic signs that tested negative for WNV infection during the WNV epidemic in Colorado are not known; however, several potential explanations require further investigation.