Diagnostic accuracy of a urine lipoarabinomannan strip-test for TB detection in HIV-infected hospitalised patients

Abstract

Lack of point-of-care tests for tuberculosis (TB) result in diagnostic delay, and increased mortality and healthcare-related costs. The urine Determine^TM TB-LAM point-of-care strip-test was evaluated in 335 prospectively-recruited hospitalised patients with suspected TB-HIV co-infection (group 1) and from 88 HIV-infected hospitalised patients with non-TB diagnoses (group 2). Cut-off point-specific analyses were performed using: 1) a microbiological reference standard (culture positive versus negative); and 2) a composite reference standard (exclusion of patients with clinical-TB from the culture-negative group). Using the microbiological reference and the manufacturer-recommended grade-1 cut-off point, LAM sensitivity and specificity was 66% (95% CI 57–74%). By contrast, using the composite reference sensitivity was 60% (95% CI 53–67%) and specificity improved to 96% (95% CI 89–100%) (p=0.001). The same pattern was seen when the grade-2 cut-off point was used (specificity 75% versus 96%; p=0.01). In group two patients specificity was poor using the grade-1 cut-off point, but improved significantly when the grade-2 cut-off point was used (90% versus 99%; p=0.009). The grade-2 cut-off point also offered superior inter-reader reliability (p=0.002). Sensitivity was highest in those with a CD4 Mycobacterium tuberculosis culture-positive patients. This preliminary study indicates that the LAM strip-test may be a potentially useful rapid rule-in test for TB in hospitalised patients with advanced immunosuppression. The grade 2, but not the manufacturer-recommended grade 1 cut-off point, offered superior rule-in utility and inter-reader reliability. Larger studies to evaluate cut-off points and diagnostic accuracy are urgently required.