Abstract: Objective To evaluate the effects of the BACTEC MGIT 960 in detecting the resistance of clinical Mycobacterium tuberculosis isolates to the second-line anti-tuberculosis drugs. Methods The drug susceptibility test of the clinical isolates to four anti-tuberculosis drugs, capreomycin (CPM), kanamycin (KAN), ofloxacin (OFLX) and Ethionamide (ETH) was conducted by MGIT 960, and the results were compared with those obtained by L-J proportion method. Results A total of 111 clinical Mycobacterium tuberculosis isolates were detected. The consistency of the drug susceptibility tests by MGIT 960 and L-J proportion method was 95.5% for CPM (106/111), 93.7% for KAN (104/111), 97.3% for OFLX (108/111) and 85.6% for ETH (95/111). Compared with L-J proportion method, the sensitivity, specificity and accuracy of MGIT 960 were 80.0%, 97.0%, 95.5% for CPM, 92.3%, 93.9%, 93.7% for KAN, 95.3%, 98.5%, 97.3% for OFLX and 100%, 81.2%, 85.6% for ETH. The completion of the drug susceptibility test by MGIT 960 needed 8.3 days averagely, and that by L-J proportion method needed 26.2 days, the difference was statistical significant (P＜0.01). Conclusion The result of drug susceptibility test by MGIT 960 was highly consistent with that by L-J proportion method, but the former needs much shorter time than the latter. The application of MGIT 960 could facilitate the early diagnosis and treatment of drug-resistant tuberculosis.