In a multicenter study involving three reference centers for mycobacteria, the reliability of the Mycobacteria Growth Indicator Tube (MGIT) for rapid antimicrobial susceptibility testing (AST) of Mycobacterium tuberculosis was evaluated and compared to the radiometric method (BACTEC 460TB). Test cultures for which the results of the MGIT and BACTEC 460TB tests were discordant were checked by the conventional proportion method on solid medium. Four hundred forty-one isolates have been tested for susceptibility to isoniazid (INH), rifampin (RMP), ethambutol (EMB), and streptomycin (SM). Discrepant results were obtained for three isolates (0.7%) with INH (susceptible by MGIT, resistant by BACTEC 460TB), for four isolates (0.9%) with RMP (susceptible by MGIT, resistant by BACTEC 460TB), for six isolates (1.9%) with EMB (four susceptible by MGIT, resistant by BACTEC 460TB; two resistant by MGIT, susceptible by BACTEC 460TB), and for four isolates (0.9%) with SM (two susceptible by MGIT, resistant by BACTEC 460TB; two resistant by MGIT, susceptible by BACTEC 460TB). When cultures with discordant results were tested by the conventional proportion method, about half of the cultures yielded results similar to the BACTEC 460TB results, while the other half yielded results similar to the MGIT results. Turnaround times were 3 to 14 days (median, 8.8 days) for MGIT and 3 to 15 days (median, 7.8 days) for BACTEC 460TB. There was no statistically significant difference between the susceptibility testing results of the two methods (P > 0.05). These data demonstrate that the MGIT system is an accurate, nonradiometric alternative to the BACTEC 460TB method for rapid susceptibility testing of M. tuberculosis.