A total of 722 respiratory and 86 nonrespiratory specimens obtained from 456 patients were tested for detection of Mycobacterium tuberculosis complex by a commercial polymerase chain reaction (PCR) kit (Amplicor, Roche Diagnostic Systems) and the results compared with those of microscopy and culture (solid and radiometric media). Respiratory and nonrespiratory specimens were analysed separately. Of the respiratory specimens, 54 were positive for Mycobacterium tuberculosis complex both in the PCR and in culture, five were positive in the PCR but negative in culture, and eight were positive in culture but negative in the PCR. Four cultures were positive for mycobacteria other than Mycobacterium tuberculosis; none of these gave a positive result in the commercial test. Resolution of discrepant results was performed by analysis of patients' clinical data. For respiratory specimens the sensitivity of the commercial test was 87.6%, the specificity 99.6%, the positive predictive value 96.6%, and the negative predictive value 98.7%. For nonrespiratory specimens the sensitivity was 60%, whereas the specificity ranged as high as 98.6%. For this group the positive predictive value was 85.7% and the negative predictive value 94.9%. When respiratory specimens are used, the commercial PCR test for detection of Mycobacterium tuberculosis complex, with its high sensitivity and specificity, is a good complementary diagnostic tool for rapid diagnosis of bronchopulmonary tuberculosis in a routine mycobacterial laboratory.