Diagnostic accuracy of commercial urinary lipoarabinomannan detection in African tuberculosis suspects and patients.

OBJECTIVE: To evaluate a commercially available antigen capture enzyme-linked immunosorbent assay (ELISA) based on detecting lipoarabinomannan (LAM) in urine for the diagnosis of tuberculosis (TB). DESIGN: Consenting TB suspects and registering TB patients prospectively recruited from three hospitals were asked for two sputum specimens for microscopy and culture urine for LAM testing and blood for human immunodeficiency virus (HIV) testing with radiological and clinical follow-up for 2 months. RESULTS: Of 427 participants complete data were available from 397 (307 adult and 23 adolescent TB suspects and 67 registering TB patients). HIV prevalence was 77%. TB was diagnosed in 195 (49%) including 161 culture-positive patients and confidently excluded in 114 (29%) participants. LAM ELISA sensitivity was 44% (95%CI 36-52) for culture-confirmed TB (52% in smear-positive patients). Specificity was 89% (95%CI 81-94). Sensitivity was significantly higher in HIV-related TB (52% 95%CI 43-62 P < 0.001) compared to HIV-negative TB (21% 95%CI 9-37). Sensitivity in smear-negative patients was low (28% 95%CI 13-43) for combined HIV-positive and -negative patients. CONCLUSION: Our findings confirm greater sensitivity of urine LAM detection for HIV-related TB. However both sensitivity and specificity were suboptimal suggesting that this version cannot confirm or exclude TB in either HIV-infected or non-infected patients.