Abstract
Rapid, reliable results can be given by molecular, direct detection and identification of the Mycobacterium tuberculosis (MTB/Mtb) complex from clinical samples. The Xpert MTB/RIF assay is an assay that has been availablefor more than a decade for identification of Mycobacterium tuberculosis and resistance to rifampicin. However, there is minimal evidence on its clinical usefulness in paucibacillary, non-respiratory samples. The Xpert MTB/RIF assay clinical utility index, its diagnostic characteristics and the number required to diagnose 2935 non-respiratory specimens submitted for routine mycobacterial work-up in a reference laboratory in an intermediate prevalence setting per specimen form were evaluated. The Xpert MTB/RIF assay showed a variable clinical utility index and number required to diagnose (NND) depending on the type of specimen, which was moderate in tissue biopsies (NND = 1.8) and excellent in pus and urine samples, compared to acid-fast microscopy and culture as a gold standard technique (NND = 1.1 and 1.2). Microscopy, on the other hand, consistently showed a weak to fair index of clinical usefulness in all specimen forms, with in NND of 2.3–12.5. The NND for detecting tuberculous infection in the cerebrospinal fluid by the Xpert MTB/RIF assay was noted to be 1.2, with a moderate clinical utility index of 0.8. The evidence presented indicates that the overall appropriate diagnostic utility of the Xpert MTB/RIF assay is clinically successful in most non-respiratory samples. To check the cost-effectiveness and prognostic effect of integrating this completely automated molecular-based assay into the routine testing algorithm for non-respiratory mycobacterial specimens, further data must be collected.