The reliability of the BACTEC MGIT 960 system, an automated version of the Mycobacteria Growth Indicator Tube (MGIT), for antimicrobial susceptibility testing of Mycobacterium tuberculosis was evaluated on 78 clinical isolates. Rifampin (RMP), isoniazid (INH), streptomycin (SM), and ethambutol (EMB) were tested at the following concentrations: 1.0 microg/ml for RMP, 0.1 and 0.4 microg/ml for INH, 1.0 and 4.0 microg/ml for SM, and 5.0 and 7.5 microg/ml for EMB. Results were compared with those obtained by the BACTEC 460 TB radiometric system. Initially the reproducibility study showed 99.5% agreement on repeat testing with all the four drugs. With susceptibility testing of clinical isolates, excellent agreement between the two systems was found for all the drugs. A total of nine major errors were observed for only three isolates, resistant according to BACTEC MGIT 960 and susceptible according to BACTEC 460 TB, to SM (4.0 microg/ml), INH (0.1 microg/ml), and EMB (5.0 microg/ml) (one isolate) and to SM (1.0 microg/ml), INH (0.4 microg/ml), and EMB (5.0 microg/ml) (two isolates). When these isolates were tested by using the conventional proportion method on Löwenstein-Jensen medium, agreement with BACTEC MGIT 960 was found for five results and with BACTEC 460 TB for the remainder. The time to report results was 7.9 days by MGIT 960 and 7.3 days by BACTEC 460 TB, which was not found statistically significant (P > 0.05). In conclusion, the performance of BACTEC MGIT 960 was found similar to that of BACTEC 460 TB and this new system can be considered a good alternative to the radiometric method for routine susceptibility testing of M. tuberculosis.