Discriminatory accuracy of serological tests for detecting Trypanosoma cruzi using the ROC curve and the standard methodology

espanolIntroduccion. Las pruebas serologicas se utilizan para detectar infeccion por Trypanosoma cruzi y su precision discriminatoria depende del umbral de decision establecido. La metodologia estandar y la de curva caracteristica operativa del receptor (COR) permiten obtener tal umbral. Objetivo. Comparar la precision discriminatoria de la metodologia estandar y la metodologia de curva COR en lo que respecta a pruebas serologicas para detectar infeccion por T. cruzi. Materiales y metodos. Se utilizo un conjunto de valores de anticuerpos contra T. cruzi de individuos clasificados como chagasicos o sanos y se realizaron simulaciones computarizadas en condiciones homocedasticas y heterocedasticas. Se calculo sensibilidad, especificidad, sensibilidad=100%, especificidad=100% y decision-perfecta. Resultados. La precision discriminatoria de la metodologia estandar favorecio la especificidad (98.22% a 99.56%) sobre la sensibilidad (67.25% a 87.14%), mientras que la de la curva COR mostro un equilibrio entre sensibilidad (94.56% y 96.44%) y especificidad (90.35% y 92.11%). Esta ultima metodologia tambien evidencio una mayor proporcion de decision-perfecta, la cual llego a ser >90% en condiciones de homocedasticidad. Los umbrales de decision fueron afectados por las condiciones de heterocedasticidad. Conclusion. La metodologia de la curva COR mostro una mejor precision discriminatoria, por lo que se recomienda su uso para el calculo de umbrales de decision en pruebas serologicas para la enfermedad de Chagas. EnglishIntroduction: Serological tests are used to confirm Trypanosoma cruzi infection and their discriminatory accuracy depends on the established decision threshold. Both, the standard methodology and the receiver operating characteristic (ROC) curve methodology allow obtaining such threshold. Objective: To compare the discriminatory accuracy of the standard methodology and the ROC curve methodology regarding serological tests for confirming T. cruzi infection. Materials and methods: A set of anti-T. cruzi antibodies values from subjects previously classified as healthy or as having Chagas disease were used, and computer simulations were performed under homoscedasticity and heteroscedasticity conditions. Sensitivity, specificity, 100% sensitivity, 100% specificity, and perfect-decision were calculated. Results: The discriminatory accuracy obtained with the standard methodology favored specificity (98.22% to 99.56%) over sensitivity (67.25% to 87.14%), while in the ROC curve methodology a balance between sensitivity (94.56% and 96.44%) and specificity (90.35% and 92.11%) was observed. Also, in the ROC curve methodology a greater perfect-decision ratio was observed, which, under homoscedasticity conditions, was >90%. Decisions thresholds were affected by heteroscedasticity conditions. Conclusion: The ROC curve methodology showed better discriminatory accuracy, therefore its use for calculating decision thresholds in serological tests for detecting Chagas disease is recommended.