Objective: To evaluate the feasibility of the implementation of a commercial rapid molecular diagnostic test (Xpert MTB/RIF) for the routine diagnosis of smear-negative or extrapulmonary tuberculosis (TB) and its diagnostic accuracy, and to assess HIV prevalence in a real-life setting in Madagascar. This study was set in a tertiary care hospital in Madagascar.

Methods: A prospective cohort study was conducted of all consecutive cases with suspected smear-negative and/or extrapulmonary TB over a 2-year period. Cases were classified as proven, probable, or possible TB cases, or as having an alternative diagnosis.

Results: Of the 363 patients included, 183 (50.4%) had suspected smear-negative pulmonary TB and 180 (49.6%) had suspected extrapulmonary TB. For proven cases, the sensitivity, specificity, positive and negative predictive values of Xpert MTB/RIF were 82.4%, 98.8%, 98.3%, and 86.6%, respectively; for proven and probable cases grouped together, these values were 65%, 98.8%, 98.5%, and 64%, respectively. The diagnostic accuracy was slightly lower for extrapulmonary TB compared to smear-negative pulmonary TB. The prevalence of HIV infection was 12.1%, but almost half of these cases did not have TB (alternative diagnosis group).

Conclusions: The implementation of a rapid diagnosis programme for TB in a resource-poor setting is feasible. The performance of the Xpert-MTB/RIF was remarkable in this difficult-to-diagnose population. HIV prevalence in this study was much higher than the prevalence reported in the general population in Madagascar, in patients with TB and patients with conditions other than TB.