Desenvolvimento e validação de modelo preditivo e avaliação de testes de diagnóstico por classe latente para o parasitismo por L. chagasi em cães atendidos no Hospital Veterinário Universitário da UFPI, Teresina

Abstract

American visceral leishmaniasis (AVL) is an expanding disease in Brazil, for which few, and apparently inefficient, control strategies are available. A major problem for the containment of visceral leishmaniasis is the lack of an accurate test for the identification of infected dogs, which are considered the main reservoirs of the disease in urban areas. In this sense, the characterization of clinical and laboratory markers of the infection in this reservoir, as well as, a more appropriate assessment of the performance of tests for the diagnosis of infection, might increase the effectiveness of control strategies for AVL. Thus, this study has two main objectives: (1) to develop and validate a prediction model for parasitism by Leishmania chagasi in dogs, based on the results of serological tests and clinical signs and (2) to evaluate the sensitivity and specificity of a clinical criterion, and of parasitological and serological tests for canine infection by L. chagasi using latent class analysis. The first goal was developed from a study in which clinical, serological and parasitological data were obtained for all dogs, suspicious or not to AVL, admitted at the Veterinary Hospital at the Federal University of Piauí, Teresina, in the years of 2003 and 2004, totaling 1412 animals. Logistic regression models were constructed with the animals admitted in 2003 with the aim of developing a predictive model for parasitism based on clinical signs and results of serology by indirect immunofluorescence (IIF). This model was validated in dogs admitted at the hospital in 2004. To evaluate the area under the ROC curve (AuROC), sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and overall accuracy, three models were developed: one based only on clinical variables, other only with the serological results and a final model based on both the clinical and serological data. Among the three, the last model showed the best performance (AuROC=90.1%, sensitivity 82.4%, specificity=81.6%, PPV=73.4% NPV=88.2% and overall accuracy=81.9%). It is concluded that the use of predictive models based on clinical and serological data for diagnosis of canine visceral leishmaniasis might be useful in the evaluation of canine infection, allowing the anticipation of ontainment of these animals in order to reduce transmission levels. The second goal was developed through a cross-sectional study of 715 dogs aged between 1 month and 13 years, with varied races evaluated by veterinarians at VUH-UFPI in the period January to December 2003. The sensitivities and specificities of clinical, parasitological and serological tests for detection of canine infection by Leishmania chagasi were estimated using latent class analysis, considering four types of models and different cutoffs. The best sensitivity estimates for the clinical, serological and parasitological tests were 60%, 95% and 66% respectively. The best estimates for the specificity of clinical, serological and parasitological tests were 77%, 90% and 100% respectively. It is concluded that the use of parasitological assays as the gold standard for validation of diagnostic tests is not appropriate and that the indicators of accuracy of the tests evaluated are insufficient to justify that they are used alone for diagnosis of infection in order to control the disease.