Accuracy of diagnostic tests for American tegumentary leishmaniasis: a systematic literature review with meta-analyses.

Abstract

To analyse the accuracy of American tegumentary leishmaniasis (ATL) diagnostic methods and evaluate the quality of the existing publications by means of a systematic review. Diagnostic tests evaluated in at least two studies with common reference standards were included in the sensitivity and/or specificity meta-analyses. Quality and susceptibility to bias were analysed using the QUADAS-2 and STARD tools. The title and abstract of 3387 publications were evaluated after deduplication resulting from database searches. Thirty-eight studies were included in the review, and 26 of them had results inserted in meta-analyses. The diagnostic methods with the highest pooled sensitivity values were ELISA, polymerase chain reaction (PCR), indirect immunofluorescence reaction and Montenegro's intradermal reaction. Cytometry was assessed in only two studies and presented 100% sensitivity in both. Smear slide microscopy and histopathology showed low pooled values of sensitivity. For specificity, the highest pooled values were identified for PCR. High values were also identified for ELISA, except for studies in which the reference standard for defining negative participants included individuals with Chagas' disease or paracoccidioidomycosis, which also occurred for cytometry. IFR had lower specificities than ELISA. There was a predominance of case-control designs of phase 1 or 2 and only four studies were strongly recommended as evidence generators. Several reference standards were adopted, and different methods were assessed in a small number of studies. PCR showed the highest accuracy for the diagnosis of ATL, and its use should be encouraged in clinical practice. ELISA is recommended for the screening of suspected individuals, but the possibility of cross-reactions should be considered. New validation studies for the tests evaluated in few publications and studies of phase 3 with appropriate methods are needed.