Two commercial systems for the amplification and detection of Mycobacterium tuberculosis directly from respiratory samples were compared. The Roche Cobas Amplicor MTB Test and the Roche manual Amplicor MTB Test (Roche Diagnostic Systems, USA) were applied to 755 decontaminated respiratory specimens collected from 470 patients. Results were compared with those of acid-fast staining and culture. A total of 251 specimens were collected from 156 patients diagnosed with pulmonary tuberculosis, including 28 specimens corresponding to 13 patients that were receiving antituberculous treatment. Given the overall positivity rate of 33.2% (251/755), the sensitivity, specificity, and positive and negative predictive values were 92.4, 100, 100, and 96.5%, respectively, for the Cobas Amplicor MTB Test and 90.8, 100, 100, and 95.8%, respectively, for the Amplicor MTB Test. For 204 (81.3%) smear positive specimens and 47 (19.7%) smear negative specimens, the sensitivity values were 100 and 59.6%, respectively, for the Cobas Amplicor MTB Test and 100 and 51%, respectively, for the Amplicor MTB Test. There were no statistically significant differences in sensitivity or specificity between the two assays and culture (p>0.05). The overall results of both assays were concordant for 99.5% of the samples. It is concluded that although both nucleic acid amplification methods are rapid and specific for the detection of Mycobacterium tuberculosis complex in respiratory specimens, the Cobas Amplicor MTB Test appears to be slightly more sensitive than the Amplicor MTB Test when smear negative specimens are investigated.