Caseous lymphadenitis is an important disease of sheep and goats,1 and results of a study2 published in 1997 indicate this disease is the third most common cause of economic loss in the sheep industry. Results of another study3 in which culled sheep from 9 western US states were examined indicate the prevalence of CL in sheep is 42.41%. To the authors' knowledge, few studies have been conducted to determine the prevalence of CL in goats in the United States. An impediment to determination of the true prevalence of CL is the lack of a definitive test for identification of animals harboring Corynebacterium pseudotuberculosis that have not developed visible lesions. The performance of an SHI test to detect antibodies against C pseudotuberculosis has been evaluated for several populations of sheep and goats to determine the prevalence of CL and the usefulness of the test for CL control programs4; results of that study indicate the SHI test does not have high specificity for identification of animals with CL, which could result in inaccurate determination of the prevalence of CL and culling of genetically valuable seropositive animals that would not have developed active CL lesions (ie, lesions in which C pseudotuberculosis is growing).4
Various techniques have been used for management of CL in large herds or flocks of animals including testing and elimination of animals with positive assay results, vaccination, and culling animals with suspected external CL lesions (ie, externally detectable lesions likely attributable to C pseudotuberculosis infection). Although these CL management techniques have been successful,5,6 methods that can be used to retain genetically valuable animals in herds or flocks are desirable. In another study,7 we compared outcomes of various treatment regimens for sheep and goats with CL. Treatments of animals in that study included lancing, flushing, and draining of lesions and parenteral administration of penicillin G procaine or intralesional or parenteral administration of tulathromycin. Results of that study7 indicated no significant differences among treatment groups regarding proportions of lesions that had resolved by 1 month after initiation of treatment. We believe that the treatments used in that study7 do not affect future development of new lesions in affected animals. In addition, we believe that intralesional or parenteral administration of tulathromycin does not affect the odds of C pseuotuberculosis infection for other animals in a herd without appropriate biosecurity measures. These opinions were determined on the basis of information reported in the literature and clinical experience.
Investigator interest in evidence-based medicine and the use of Bayesian statistics for medical applications is increasing.8 Bayesian statistical approaches assume that data are true values and mean values are derived. By use of Bayesian statistical techniques, credible value intervals can be determined that are centered near the mean values calculated for a sample. Frequentist statistical approaches use population-based inference and assume that a mean value for a population can only be estimated via analysis of data. In such analyses, confidence intervals are centered at calculated mean values.9 Bayesian statistical approaches may be advantageous for clinical research; by use of such methods, a credible interval can be determined that has a 95% probability of including the true mean value. For example, if results of a study indicate that the odds for subjects with serologically positive results to develop lesions in the future has a mean value of 3 and a credible interval of 1.5 to 6, the findings can be directly applied to evaluation of animals with clinical disease; such data would indicate that a patient with serologically positive results is 1.5 to 6 times as likely as a patient with negative results to develop lesions in the future. Furthermore, results of Bayesian analysis can be used to identify a specific probability for every value of the evaluated variables, whereas results of frequentist analyses can be used to accept or reject a null hypothesis. For example, for values of P < 0.05, results of frequentist analyses will typically indicate that the initial or fixed parameter value should be rejected, but results of such analyses do not apply to any other values of variables. Also, Bayesian estimates are influenced by prior beliefs and information determined with means of data for prognostic factors. Prior beliefs are useful in analysis if they are known with confidence and can therefore influence posterior likelihood (ie, evaluation of prognostic factors). When prior beliefs are considered vague or uninformative, the results are primarily influenced by the data and are therefore similar to results determined with frequentist statistical analyses.
The objectives of the study reported here were to evaluate herds of sheep and goats in southwest Texas in which various CL control measures were used to determine the serologic prevalence of antibodies against C pseuotuberculosis, the prevalence of active external CL lesions, and the odds for future development of CL lesions in animals with serologically positive assay results and those with active external CL lesions versus animals with serologically negative results and those without active CL lesions, respectively. Because of characteristics of data and populations of animals evaluated in the study reported here, we chose to use a Bayesian approach for statistical analyses. Our hypothesis was that animals with positive serologic results and those with external lesions attributable to CL would have higher odds for future development of CL lesions versus animals without such findings.
Synergistic hemolysin inhibition
BBL, BVA Inc, San Antonio, Tex.
API Corynebacterium identification system, BioMerieux Inc, Hazelwood, Mo.
Texas Veterinary Medical Diagnostic Laboratory, College Station, Tex.
Washington Animal Disease Diagnostic Lab, Washington State University, Pullman, Wash.
WinBUGS, version 1.4, Imperial College of Science, Technology, and Medicine, Cambridge, England.
3. Stoops SG, Renshaw HW, Thilsted JP. Ovine caseous lymphadenitis: disease prevalence, lesion distribution, and thoracic manifestations in a population of mature culled sheep from western United States. Am J Vet Res 1984; 45: 557–561.
4. Brown CC, Olander HJ, Zometa C, et al. Serodiagnosis of inapparent caseous lymphadenitis in goats and sheep, using the synergistic hemolysis-inhibition test. Am J Vet Res 1986; 47: 1461–1463.
5. Dercksen DP, ter Laak EA, Scheuder BEC. Eradication programme for caseous lymphadenitis in goats in the Netherlands. Vet Rec 1996; 138: 237.
6. Paton MW, Sutherland SS, Rose IR. The spread of Corynebacterium pseudotuberculosis infection to unvaccinated and vaccinated sheep. Aust Vet J 1995; 72: 266–269.
7. Washburn KE, Bissett WT, Fajt VR, et al. Comparison of three treatment regimens for sheep and goats with caseous lymphadenitis. J Am Vet Med Assoc 2009; 234: 1162–1166.
8. Ashby D, Smith AFM. Evidence-based medicine as Bayesian decision-making. Stat Med 2000; 19: 3291–3305.
9. Annis C.Frequentists and Bayesians. What is probability? Confidence intervals vs credible intervals. Available at: www.statisticalengineering.com/frequentists_and_bayesians.htm. Accessed Jan 1, 2008.