Diagnostic accuracy of MAT and ELISA assays in the detection of Leptospira in two, low-prevalence, populations

Leptospirosis, a zoonotic disease that affects both humans and other animals, is caused by the ingestion of Lepstospira bacteria, usually through contaminated water or soil. The purpose of this study was to characterize the performance of assays in detecting prior infection with leptospirosis in low-prevalence populations. Assays investigated included two commercial ELISAs and the micro-agglutination test (MAT). Serum samples were obtained from two previous study populations: urban women from southern Chile (n=393) and children in the Marshfield Epidemiological Study in Wisconsin (n=611). Percent agreements were computed for each pair of assays used. A Bayesian no-gold standard approach was used to estimate prevalence, sensitivity (SE) and specificity (SP) for each test, and allowed for comparison of diagnostic performance of the ELISA assays to the common reference test, MAT. One ELISA assay was systematically unreliable and removed from further analysis. Percent agreements between the ELISA and MAT assays were high for negative samples, but low for positive samples. The Bayesian estimate of SE for MAT was 64% and 74.2% for ELISA. Estimates for SP were 94.3% for MAT and 77.3% for ELISA. The prevalence estimate, (adjusted for SE and SP) was 2.1% for the Chilean study population and 1.0% in Wisconsin. Knowledge of assay SE and SP are useful for correct interpretation of epidemiological studies that rely on serology to ascertain infection status. Improved characterization of the SE and SP of ELISA kits could potentially allow them to substitute for the more laborious MAT assays in epidemiological studies.