A previously proposed MGIT protocol was assessed focussing on its reliability to test the mainstay antituberculous drugs. Isoniazid (H) (0.1 mg/l) and rifampin (R) (1 mg/l) were assayed against 109 Mycobacterium tuberculosis isolates affecting patients at high risk of multidrug-resistant tuberculosis. All isolates were simultaneously tested on Löwenstein-Jensen medium by the proportion method considered the gold standard. MGIT readings were accomplished within 2 and 22 days after inoculation, at day 10, 93.4% of the tests were completed. Unsatisfactory to evaluate H activity, the assay misclassified 8.8% (5/57) H susceptible and 7.7% (4/52) H resistant isolates. Otherwise, it yielded correct results for all 60 R susceptible isolates and 93.9% (46/49) R resistant isolates. Properly backed-up by a conventional test and targeting high risk patients, the MGIT system proved to be a useful aid to anticipate most of resistance to R. Accuracy, cost and turnaround time were competitive compared with those of semi-automated culture-based or molecular methods.