LATENT CLASS ANALYSIS IN DIAGNOSTIC TESTS EVALUATION FOR CANINE LEISHMANIA INFANTUM INFECTION.

Abstract

Accurate assessment of diagnostic tests may be biased if an imperfect reference test is used for comparison; such a situation exists for the diagnosis of canine leishmaniasis. We compared classical diagnostic tests for Leishmania infantum with Latent Class Analysis (LCA), to assess whether we could make a more accurate calculation of diagnostic accuracy. Microscopy (Lymph node aspirate), serological test (IFAT), and molecular tests (LAMP and PCR) data were recorded for 75 dogs captured in Tunisian endemic area and suspected of leishmaniasis. Sensitivity and specificity estimates with the 2 x 2 contingency tables (Microscopy as gold standard) were broadly corroborated by LCA. However, the LCA provided a way to control the study limitations (small sample size) as well as for confounding factors. It also produces consistent estimates of the test characteristics. LCA estimation of the sensitivity and specifcity of the LAMP cpb assay (se: 68.7% [95% CI 573-80%] and sp: 86.2 [95% CI 749-975%]) is higher as compared to classical calculations (se: 54.2% [95% CI 38.2-69.5%] and sp: 80% [95% CI 65.2-89.5%). Considering the lack of an adequate reference standard, LCA proved to be a useful tool to independently evaluate diagnostic methods.