Efficacy of Vaccination for West Nile Virus in Saskatchewan Horses

Preliminary analysis of studies conducted during the 2003 outbreak of West Nile Virus in Saskatchewan showed that vaccination was strongly associated with the prevention of clinical disease at both the individual horse and the farm level.

1. Introduction

In North America, West Nile Virus (WNV) was first diagnosed in horses in the late summer of 1999 [1]. Symptoms ranged from mild to severe with a case fatality rate of 25 - 40% [1,2]. Not all infected horses developed clinical symptoms. One report suggests that only 10% of horses infected with WNV will exhibit neurological symptoms [3]. Other than geographic location and age, factors associated with the development of disease have not been reported in previous studies [4]. Factors associated with seroconversion include geographic location, age, housing, and use of horses [4,5]. A variety of methods of mosquito control and housing of horses have been suggested as means of decreasing the exposure risk of susceptible horses to WNV [6]. Controlled, challenge trials have shown protection against WNV in vaccinated horses.

Before the summer of 2003, it was recognized that most horses in the province of Saskatchewan were housed outside with little protection against exposure to mosquitoes. Based on the spread of WNV across the continent during the previous year, horses in the province were expected to be at high risk of exposure to the virus in the coming summer. As a consequence, vaccination was recommended as the most logical means of preventing the disease in susceptible animals.

In anticipation of these events, several studies were undertaken to examine the impact of this disease on both animals and people in the province. The objectives of these studies were to identify risk factors for WNV in horses at both the farm level and the individual horse level and to assess the efficacy of a provisionally licensed killed WNV vaccine [a].

2. Materials and Methods

Three related studies, including a case-control study, were carried out to assess the impact of WNV among horses in Saskatchewan. During these investigations, we were able to obtain data on the occurrence of disease and vaccination status among a sample of 875 horses. A confirmed case of clinical disease was defined as any equid with symptoms consistent with WNV and a positive serum IgM enzyme-linked immunosorbent assay (ELISA) test. Vaccination was defined as two doses of vaccine received before the occurrence of disease on a given farm.

For the case-control study, a subset of case farms was randomly selected from affected properties across the province along with an equal number of control farms. Case farms had at least one clinical case of WNV on the premise in 2003, a minimum of five horses on the property, and an owner willing to participate in our investigations. A control farm was the nearest property to a case farm with at least five horses, an owner willing to participate in the study, and no confirmed or suspected clinical cases of WNV disease.

In the presence of a member of the research team, study participants completed a form that provided the following information:legal land location, individual characteristics of all sampled horses, herd and individual management methods, and mosquito control measures. Data on 5 - 10 horses was collected from each farm. All data were collected within 3 mo of the onset of disease on the case farm. The disease status of each farm was monitored until the end of the mosquito season.

Horse-Level Data

Horses were classified as either light horses, draft and draft crosses, mules, miniatures, or ponies. Gender was classified as female, intact male, or gelding. Color was divided into three categories:darks were chestnut, sorrel, brown, black, or bay; lights were gray, white, palomino, buckskin, or dun; multi-colored were roan, paint, pinto, or appaloosa. Shelter was classified as an enclosed barn accessible to the horses, a simple three-sided shelter with corral or pasture, or an open pasture with or without trees.

Farm-Level Data

Primary shelter of the majority of horses on the premises was recorded using the above categories. Farms were classified as either a farm that used a method of mosquito control and a farm that did not. The proportion of sampled horses vaccinated on each farm was also recorded.

Environmental Confounders

Mosquito data were obtained from the 2003 Saskatchewan Health mosquito-trapping program. Cases and controls were assigned a value equal to the highest average weekly number of C. tarsalis mosquitoes trapped during the period between June and September in the nearest New Jersey Light trap (NJLT).

Daily precipitation data for all climate stations in Saskatchewan were obtained from Environment Canada. Season average precipitation was calculated for each climate station using values for June, July, and August 2003. These variables reflect the principal growing period that was available for the mosquito development. Cases and controls were assigned values from the nearest climate station with complete daily precipitation information.

Growing degree days (GDD) are the number of degrees above the base temperature of 16°C (the threshold of development temperature for C. tarsalis) calculated using average daily temperature. Season GDD was calculated for each climate station by adding all positive values accumulated across the period of May to September 2003. Cases and controls were assigned GDD from the nearest climate station with complete data.

Statistical Analysis

Each potential risk factor was assessed for association with individual clinical disease status in a three-level hierarchical logistic regression model [b] to account for clustering of horses within herds and pairing of case and control farms. Each variable was also assessed for its association with farm disease status in a two-level hierarchical model [b] to account for herds within pairing of case and control farms. All independent variables with p > 0.25 were considered in developing the final multivariable models. Variables were removed from the model by manual backward selection based on statistical significance or the effect of the outcome of interest on the estimates. GDD, precipitation, and C. tarsalis numbers were maintained in both farm and individual disease status models as confounders.

3. Results

Of the 875 horses in all of the studies, 527 were vaccinated, and 348 were not vaccinated. Of the vaccinated horses, nine were symptomatic; of the non-vaccinated horses, 121 were symptomatic. The crude odds of disease in unvaccinated horses were 31 times greater (95% CI = 15 - 61) than the odds of disease in vaccinated horses. This equates to a crude estimate of vaccine efficacy (attributable fraction) of 96.7% (95% CI = 94 - 98%).

Of the 123 farms with clinical cases in 2003, 44 (36%) met the definition for a "case farm". Of these, 23 (53%) were randomly selected and matched by location to 23 control farms. A total of 300 horses were sampled on the 46 farms. The locations of the case-control pairs largely corresponded to the distribution of disease across the province of Saskatchewan.

After controlling for the effect of all confounding environmental, geographic, and horse-level variables, only vaccination was retained in the final farm and individual-animal models. The odds of disease on farms that did not vaccinate any animals were more than 85 times greater than the odds of disease on farms that vaccinated all animals. Additionally, after controlling for the effect of all confounding variables at the level of the individual horse, the odds of disease in horses that were not vaccinated were more than 20 times greater than they were among horses that were vaccinated.

4. Discussion

Vaccination has been reported to have a protective effect with respect to disease outcome (recovery or death) when horses show neurological symptoms; however, no published study has looked at whether vaccination can protect against the development of clinical disease [7,8]. Analysis of the data from all 875 animals in our studies showed that the odds of disease in unvaccinated horses were 31 times greater than among vaccinated animals. Data from our case-control study in which we controlled for the effect of relevant environmental, geographic, and horse-level variables showed that vaccination had a substantial effect on disease at the level of both the farm and individual horse.

WNV prevention strategies are presently based on vaccination, housing, and mosquito-control measures employed at the farm or individual level. Methods such as on-farm mosquito control or barn housing have been shown to be effective at reducing the risk of infection [5]. Failure of our study to show an effect of mosquito control on the odds of disease may have been caused by the limited adoption of these strategies in the study population. However, our data do show that vaccination provides substantial protection against the development of disease in the absence of concerted efforts to reduce the risk of exposure to the virus.

The authors gratefully acknowledge funding and support from the Western College of Veterinary Medicine Equine Health Research Fund, Saskatchewan Health, Western College of Veterinary Medicine Interprovincial Graduate Student Fellowship Fund, Wyeth Animal Health, Saskatchewan Agriculture, Health Canada, and the Canadian Co-operative Wildlife Health Center.

Footnotes

[a] West Nile Innovator vaccine, Fort Dodge Animal Health, Fort Dodge, IA 50501.
[b] MlwiN, Centre for Multilevel Modeling, Institute of Education, London WC1H 0AL, UK.