Determinação do ponto de corte, sensibilidade e especificidade do ELISA rgp90 para o diagnóstico da anemia infecciosa eqüina utilizando o modelo Bayesiano.

Abstract

Equine infectious anemia (EIA) is caused by a retrovirus. The EIAV infected animal represents the main source of virus and the diagnosis of disease through clinical signs is not reliable. The availability of diagnostic tests marked the beginning of disease control by proof correlation with the presence of virus infection. These methods often have limitations attributable to the uncertainty of sensitivity and specificity estimates. Thus it is necessary studies of validation. In order to study these variations in specificity and sensitivity results of two diagnostic tests AGID and ELISA rgp90 were evaluated by statistical approach to estimate uncertainty for the diagnosis of EIA. The first, ROC curve, assessed values of optical density for the ELISA rgp90 as a parameter to select the optimal cut-off through the positive and negative results likelihood, the correct classification of results, an individual and accumulated sensitivity and specificity. The second, the Bayesian model, estimates the sensitivity and the specificity of two diagnostic tests in the absence of a Gold Standard. The posterior distribution computed with Bayesian analysis was compared considering just one test and both combined as well. The analysis of the area under the ROC curve and Bayesian analysis revealed the best accurate results when cut-off of 0.228 was established enabling the most discrimination between infected and non-infected approaching to improve the diagnosis of equine infectious anemia.