JVDI Advertisement
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Journal of Veterinary Diagnostic Investigation Vol. 18 Issue 5, 479-482
Copyright © 2006 by the American Association of Veterinary Laboratory Diagnosticians
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Bicout, D. J.
Right arrow Articles by Sabatier, P.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Bicout, D. J.
Right arrow Articles by Sabatier, P.

Brief Communication

Distribution of Equine Infectious Anemia in Horses in the North of Minas Gerais State, Brazil

Dominique J. Bicout1, Regina Carvalho, Karine Chalvet-Monfray and Philippe Sabatier

Correspondence: 1 Corresponding Author: Dominique J Bicout, Biomathematics and Epidemiology Unit – TIMC, National Veterinary School of Lyon, 1 avenue Bourgelat, B.P. 83, 69280 Marcy l'Etoile, France, bicout{at}ill.fr


   Abstract
TOP
 Sources and manufacturers
 Abstract
Introduction
 Appendix
 References
 
The paper examines the prevalence of equine infectious anemia (EIA) in horse populations in the northern part (comprising 89 cities) of Minas Gerais State, Brazil, from January 2002 to December 2004. Data on 8,981 agar gel immunodiffusion test results from the region were used as input for a statistical and autoregressive analysis model to construct a city-level map of the distribution of EIA prevalence. The following EIA prevalence (P) levels were found: 49 cities with 0 < P ≤ 0.5%, 26 with 0.5% < P ≤ 1.5%, 10 with 1.5% < P ≤ 5%, and 4 with 5% < P ≤ 25%.

Key Words: Brazil • equine infectious anemia virus • prevalence • relative risk


   Introduction
TOP
 Sources and manufacturers
 Abstract
 Introduction
Appendix
 References
 
Equine infectious anemia (EIA), also known as swamp fever, is a persistent infection caused by equine infectious anemia virus, a member of the lentivirus subfamily of retrovirus, affecting all members of the Equidae, including horses, mules, and donkeys.6 Equine infectious anemia virus-infected equids may develop fatal viremia, but most survive and remain viremic for life.3 Transmission of the EIA virus from an infected horse to a susceptible one requires mechanical vectors such as insects, mainly horseflies (Tabanus fuscicostatus Hine) and stable flies (Stomoxys calcitrans L.), contaminated syringes, needles, blood-contaminated instruments, and blood transfusions.4 The monitoring of EIA is currently based on detection of anti-EIA virus antibodies, and the most widely used test is the agar gel immunodiffusion (AGID).2

Equine infectious anemia virus is distributed worldwide and exists in an enzootic form in about 23% of countries. In Brazil, the first confirmed EIA case was reported in Minas Gerais State (MGS) in 1968,2 and the AGID has been used as the official diagnostic test for EIA since 1974. In 1998, 131,991 horses out of 8,391,942 were tested in all Brazil for EIA, and 3,689 of them were classified as positives, resulting in a prevalence of 3%. According to the local office of the Brazilian Ministry of Agriculture, the EIA prevalence in MGS was 0.88% from 1973 to 1991 and 0.69% from 1995 to 2001, whereas it was about 25% in the Pantanal, the largest swamp area in Brazil.7 The EIA status remains unknown for most (about 98%) horses in Brazil (Boletim de Defesa Agropecuária, Anemia Infecciosa Eqüina, Brasilia, DF, 1998) because the AGID tests conducted by accredited laboratories are performed only for horse transportations (interstate and international trades) and horse shows. Therefore, a quantitative risk assessment of EIA could be very informative and useful for decision makers and horse industries.

The main objective of the study reported here was to provide a city-level distribution of EIA prevalence in the north of MGS, based upon a comprehensive analysis of raw data from AGID test results collected between January 2002 and December 2004.

The northern region (NR) of MGS is situated in the southeast of Brazil, between latitudes 14°13'57''S and 22°55'22''S and longitudes 39°51'23''W and 51°02'45''W. The NR, with an area of 126,521 km2 (22% of the MGS), is composed of 89 municipalities or cities and includes about 19% of the MGS horse population (211,636 horses).a It is geographically divided into 3 river basins: São Francisco River (SFR), Pardo River (PR) and Jequitinhonha River (JR), which cover 80%, 10.5%, and 9.5%, respectively, of the total area. The dominant ecosystem is the tropical savannah or Cerrado, the yearly temperatures average between 21°C and 24°C with temperatures greater than 24°C in the rainy season, and the average annual rainfall is between 1,000 mm and 1,500 mm.a

From January 2002 to December 2004, a total of 8,981 AGID tests were conducted by the Animal Surveillance Service of Minas Gerais, 37% of which were performed during the rainy season (from October to March) and 63% during the dry season. A total of 284 horses were found to be positive for EIA (AGID-positive), 2.9% of positive horses during the rainy season and 2.7% positives in the dry season. As summarized in Table 1, all the samples tested originated from 68 cities, while the remaining 21 cities (accounting for 13% of the horse population) contributed no samples. The distribution of samples according to river basins was: 5,108 (57%) from SFR, 200 (2%) from PR basins, and 3,673 (41%) from JR basin. The distribution of the positive horses was 279 (98.2%) from SFR, 2 (0.7%) from PR, and 3 (1.1%) from JR.


View this table:
[in this window]
[in a new window]
  		Table 1 Classification of the estimated prevalence (P) of equine infectious anemia from January 2002 to December 2004, according to the location of each city by river basin in the north of Minas Gerais State, Brazil.

 
We employ the disease mapping approach to analyze the distribution of EIA prevalence in MGS. Disease mapping is now well established as a tool in the analysis of regional health data.5 The object of such an analysis is to provide an estimate of the relative risk of a disease across a regional study area.5 It was assumed for our analysis that each horse was tested once and that the likelihood of sampling an EIA positive horse more than once was negligible. We applied a statistical and autoregressive analysis to compute the estimated relative risk (defined as the estimated number of positives divided by the number of expected positives under the homogeneous hypothesis) and prevalence (defined for each city as the estimated number of cases over the horse population) map versus the reported cases of EIA. The spatial interpolation and relative risk calculations were done as follows. Starting with the horse population Si for each city "i" (i = 1,2...,N) and the reported data of the number of EIA positives for each city (Ki), the relative risk (RRi) and the averaged prevalence (P_i) of EIA for each city were computed in 2 steps:

1. Step 1: Spatial Interpolation

From the total N = 89 cities in the NR, just a single test was reported for each of 2 cities and no test from 21 cities during the 3-year study period (January 2002 to December 2004) (Table 1). In order to generate the subset of missing data for those 23 cities (Kj, where j = 1,2,...,23), the initial number of positives per city that were available (Ki, where i = 24,25,...,N) were transformed into rescaled data as, Formula . The Yi (see Appendix) and their associated spatial coordinates xi,yi of the centroids of each city (plus those of missing data) were used in the spatial prediction model as implemented in GeoBUGS (a module of WinBUGSb) to produce the interpolated Yj and thus the interpolated number of positives, Kj. About 5,000 iterations were used for the convergence of the interpolation process. The new set of the number of positives (reported plus the interpolated ones) were denoted as K i in the subsequent step.

2. Step 2: Estimated Relative Risk and Prevalence Calculations

The number of positives i for i = 1,2,...,N, the number of expected positives Formula under the homogeneous hypothesis, and the matrix A of adjacent neighbors (Aij = 1 for cities i and j sharing a common border and Aij = 0 otherwise) were used in a log linear model (see Appendix). For each city, the model incorporates both spatial correlations and unstructured variability. The software WinBUGSb was used for Markov chain Monte Carlo simulations to find Bayesian estimates of the model parameters and calculate the estimated relative risk RRi for each city (see Appendix). About 100,000 iterations were used for equilibration of the process. Subsequently, the prevalence for each city was simply obtained from the relation, Pi = Ei x RRi/Si.

The results of the 2-step calculations are presented in Table 1 and the estimated prevalence distribution on the map in Figure 1. In Table 1, the distribution of EIAV prevalence (P) is considered at the level of river basins and sorted into 4 classes of increasing prevalence according to the legend in Figure 1: null (0 < P ≤ 0.5%), low (0.5% < P ≤ 1.5%), medium (1.5% < P ≤ 5%), and high (5% < P ≤ 25%). The number of cities classified with P ≥ 0.5% were 40 (45% of the cities in NR), distributed as 29 (33%) in SFR, 6 (7%) in PR, and 5 (6%) in JR, and corresponding to 31% of farms (average size, 188 ha vs. 120 ha for P < 0.5%) and 48% of the horse population. The number of cities with no reported EIA tests was 4 in the null class of prevalence, 14 in low, 3 in medium, and zero in high classes.


Figure 18051101
View larger version (34K):
[in this window]
[in a new window]
  		Figure 1 Geographical distribution of the estimated prevalence of equine infectious anemia in the north of Minas Gerais State, Brazil. Calculations are based upon data of EIA cases from January 2002 to December 2004.

 
Using statistical modeling as described above, a city-level epidemiological profile of the estimated EIA risk and prevalence in the NR from 2002 to 2004 was constructed. It was found that 40 out of 89 cities were classified as having an EIA prevalence of P ≥ 0.5%. These cities contain 34% of NR active agricultural workers; are characterized by high horse and bovine population densities (Table 1); produce 36% of the beef and 50% of the milk, respectively, for Brazil; occupy 44% of the rural area; and stretch all around and along the major rivers SFR and PR (Fig. 1).

Rather unexpectedly for a subtropical region like NR, no significant difference ({chi}2 test P value > 0.05) was found in the proportion of EIA positives between the rainy and dry seasons. Indeed, because of the subclinical nature of EIA infection it was impossible to accurately time the infection onset, and therefore to study any correlations between the seasonality and the potential risk of EIA virus transmission and spreading by insects as documented in the literature.1 On the other hand, the increase by a factor of 1.7 in the number of tests during the dry season seems to be correlated with an increase in horse events like horse shows and races from March to September, as revealed by inspection of event schedules for 5 horse breeding associations.

The EIA relative risk and prevalence map generated in this study is informative and useful to surveillance programs for intra- and interstate horse transportations, as it indicates areas and cities where control barriers could be reinforced. However, improving our knowledge on the spread and persistence of EIA infection and finding more effective methods to deal with likely outbreaks would require combination of modeling and further investigations, including longitudinal studies on antibody kinetics and establishment of long-term asymptomatic infection.3


   Appendix
TOP
 Sources and manufacturers
 Abstract
 Introduction
 Appendix
References
 
In this section, we describe the steps displayed in Figure 2 and related models used for spatial interpolation of data and the relative risk calculations using the software WinBUGS.b More details can also be found in reference 5.


Figure 18051102
View larger version (16K):
[in this window]
[in a new window]
  		Figure 2 Diagram of the conditional autoregressive smoothing model for the relative risk of equine infectious anemia cases. Model implemented using the software WinBUGS.b Details on the model description and choice of parameters are provided in the Appendix text.

 
Interpolation model. The spatial interpolation as implemented in GeoBUGS uses the combination of 2 functions:

  1. the spatial.exp (s.e) function that allows the fitting of the parameterized covariance function between reported data. In our analysis, the kriging was conducted on the logarithm of Yi, and the spatial fitting red as, log(Yi) ~ s.e(µi, xi, yi, {tau}, {varphi}, {kappa}), where µi = ß is the mean of log(Yi) of each city, (xi,yi) spatial coordinates of the centroids of each city, {tau} a scalar representing the precision of the model, and {varphi} and {kappa} are parameters of the covariance, exp{—({phi}dij)k}, with dij the interpoint distance between cities "i" and "j." The parameters ß, {tau}, {varphi} and {kappa} used for the kriging were chosen as follows: ß ~ dflat() (is a random variable from the whole real line), {tau} ~ dgamma(0.1,0.001) (is Gamma distributed random variable of exponent 0.1 and argument 0.001), {varphi} ~ dunif(0.001,0.8) (is a uniform distributed random variable from the interval [0.001,0.8]), and {kappa} ~ dunif(0.05,1.95) (is a uniform distributed random variable from the interval [0.05,1.95])., The mean values after 5,000 iterations were: ß = 0.718, {tau} = 0.052, {varphi} = 0.407, and {kappa} = 0.092.
  2. the spatial.unipred (s.up) function that generates the unknown data from the covariance function previously determined. The interpolated data log(Yj) were thus obtained from, log(Yj) ~ s.up(ß, xj, yj, log(Yi)), where (xi,yj) are spatial coordinates associated to the Yj.

Relative Risk Estimation. The number of positives i in each city is assumed to follow a Poisson distributions of parameters or means µi (i.e., i. Poisson(µi)) and the unknown relative risk RRi are defined as, RRi = µi/Ei, where Ei are the expected number of positives as defined in the text. And the estimated prevalence Pi for each city is obtained as the ratio of the estimated number of positives µi to the population Si of the city, i.e., Pi = µi/Si = Ei x RRi/Si. To determine the µi‘s, we used the log-linear model for the relative risk as, log(RRi) = log(µi) – log(Ei) = {alpha} + ui + vi, which takes into account the overall random effect {alpha} of the relative risk and the correlated ui and uncorrelated vi heterogeneities. The parameters were chosen as follows: {alpha} ~ dflat() (is a random variable from the whole real line) and the uncorrelated random heterogeneity vi ~ dnorm(0,tau.v) (is a Gaussian distributed random variable of mean zero and inverse-variance tau.v) with tau.v ~ dgamma(0.1,0.001) (a Gamma distributed random variable of exponent 0.1 and argument 0.001). For the spatial correlated heterogeneity ui, we used the clustering structure where the risk in any city depends on neighboring cities as described in the CAR model in GeoBUGS, ui ~ car.normal(A), where A is the matrix of adjacent neighbors (see the text). The mean values after 100,000 iterations were: {alpha} = –3.20, tau.v = 22.56.


   Acknowledgments
 
Thanks to Leticia de Souza Santos and Denise M. Viegas from the Area Office of Animal Surveillance Services of Minas Gerais (DFA-MG-Brazil) for their assistance in organizing the data. We are grateful to the Minister of Agriculture (MAPA-Brazil) for the institutional support and for authorizing the use of their official registers. Dr. Carvalho is the recipient of a fellowship from the Fondation CAPES - MEC - Brazil and is grateful to MAPA-Brazil for her sabbatical at Biomathematics and Epidemiology Unit – ENVL in France.


   Sources and manufacturers
TOP
 Sources and manufacturers
Abstract
 Introduction
 Appendix
 References
 
From the Biomathematics and Epidemiology Unit–TIMC, National Veterinary School of Lyon, 1 avenue Bourgelat, B.P. 83, 69280 Marcy l'Etoile, France (Bicout, Carvalho, Chalvet-Monfray, Sabatier), and the Serviço de Sanidade Animal da Delegacia Federal de Agricultura de Minas Gerais, Belo Horizonte, MG, Brazil (Carvalho). Back

a Data available from Instituto Brasileiro de Geografia e Estatistica (IBGE): http://www.ibge.gov.br. Back

b WinBUGS v1.4, Imperial College and Medical Research Council, London, UK. Back


   References
TOP
 Sources and manufacturers
 Abstract
 Introduction
 Appendix
 References
 

  1. Barros A.T.: 2001, Seasonality and relative abundance of Tabanidae (Diptera) captured on horses in the Pantanal, Brazil. Mem Inst Oswaldo Cruz 96:917–923.[Medline]
  2. Coggins L., Norcross N.L., Nusbaum S.R.: 1972, Diagnosis of equine infectious anemia by immunudiffusion test. Am J Vet Res 33:11–18.[Medline]
  3. Hammond A.S., Li F., McKeon B.M., et al.: 2000, Immune responses and viral replication in long-term inapparent carrier ponies inoculated with equine infectious anemia virus. J Virol 74:5968–5981.[Abstract/Free Full Text]
  4. Issel C.J., Rwambo K., Montelaro R.C.: 1988, A perspective on equine infectious anemia with an emphasis on vector transmission and genetic analysis. Vet Microbiol 17:251–286.[Medline]
  5. Lawson A.B., Browne W.J., Vidal Rodeiro C.L.: 2003, Disease mapping with WinBugs and MLwiN. John Wiley & Sons, Chichester, UK.
  6. Montelaro R.C., Ball J.M., Rushlow K.E.: 1993, Equine retroviruses. In: The Retroviridae, vol. 2 ed. Levy J., pp. 257–360. Plenum Press, New York, NY.
  7. Silva R.A.M.S., DÁvila A.M.R., Abreu U.G.P.: 1999, Equine viral diseases in the Pantanal, Brazil. Studies carried out from 1990 to 1995. Rev Élevage Méd Vét Pays Tropicaux 52:9–12.




This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Bicout, D. J.
Right arrow Articles by Sabatier, P.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Bicout, D. J.
Right arrow Articles by Sabatier, P.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS