To evaluate the experience of a clinical microbiology laboratory with a DNA amplification assay for routine detection of Mycobacterium tuberculosis, the Cobas Amplicor Mycobacterium tuberculosis (MTB) polymerase chain reaction (PCR) assay (Roche Diagnostics Systems, USA) was performed on 7,722 respiratory and 1,451 nonrespiratory specimens collected from 3,321 patients. The results were compared with those of culture in conventional Lowenstein-Jensen medium, culture in the MB/BacT system (Organon Teknika, France), and clinical investigations. A total of 240 of the 254 respiratory specimens culture positive for Mycobacterium tuberculosis were also positive in the PCR assay. Of the 7,300 culture-negative specimens, 45 (0.6%) were positive in the PCR. After detailed interpretation, the overall sensitivity, specificity, and positive and negative predictive values of the PCR assay were 84.5, 99.8, 94.1, and 99.4%, respectively, for respiratory specimens. The PCR assay was more sensitive for smear-positive respiratory specimens (97.1%) than for smear-negative respiratory specimens (48.6%). Of the 18 culture-positive (smear-negative) nonrespiratory specimens, 9 were positive in the PCR. None of the 1,384 culture-negative nonrespiratory specimens were positive in the PCR. The inhibition rates detected by the internal control of the test were 2.2% for respiratory specimens and 3.4% for nonrespiratory specimens. After resolving the discrepancies, the overall sensitivity, specificity, and positive and negative predictive values of the PCR assay were 82.5, 99.8, 94.3, and 99.4%, respectively, when compared to the results of diagnostic culture. In conclusion, the use of the Cobas Amplicor MTB-PCR assay might enable clinical microbiology laboratories with considerable previous experience in molecular biology testing to perform PCR and confirm tuberculosis infection immediately, leading to improved patient management.