de la Rúa-Domènech R, Mohammed H O, Atwill E R, Cummings J F, Divers T J, Summers B A, deLahunta A, Jackson C
Section of Epidemiology, College of Veterinary Medicine, Cornell University, Ithaca, NY 14853-6401, USA.
Am J Vet Res. 1995 Nov;56(11):1433-9.
To examine the regional variations in the distribution of equine motor neuron disease (EMND) in the United States and the factors that might explain those variations.
Cluster investigation and case-control study.
The study population consisted of 97 horses with histopathologically confirmed EMND and 698 controls with diagnosis of other spinal cord disorders at 21 US veterinary teaching hospitals participating in the Veterinary Medical Data Base.
The total horse population of the United States was divided into 21 regions, and the regional incidence rates of EMND from January 1985 through January 1995 were estimated. Moran's index of spatial autocorrelation was calculated to test for spatial clustering of the disease. The 21 regions were then joined in broader areas according to the similarity of their EMND rates by means of the cluster analysis statistical technique. Finally, the role of potential confounding factors (age at diagnosis, month of diagnosis, breed, and sex) in the present distribution of EMND was assessed, using logistic regression analysis.
Differences in estimated rates across the 21 regions resulted in a strong pattern of spatial clustering of EMND in the United States. The geographic units were grouped into 5 risk regions, with the gradient of EMND incidence rates increasing from the western states (almost 0 cases/1,000,000 horse-years) toward New England (20.78 cases/1,000,000 horse-years). Reported risk factors of EMND (age, breed) and other extraneous factors (sex, month of diagnosis) could not explain the observed geographic variations of disease rates. Nevertheless, there is evidence of some confounding attributable to age and breed.
Although the mechanism responsible for the clustering of EMND in northeastern states is still unexplained, it is not an epiphenomenon caused by regional differences in the distribution of the factors investigated.
