BACKGROUND: The PCR assay for the detection of M. tuberculosis has been used for 5 years in Chonnam National University Hospital. To evaluate the reliability of the PCR assay, the PCR results were compared with those of culture and clinical diagnosis. METHODS: This study analyzed the results of BACTEC culture and PCR for detection of M. tuberculosis between Jan. 1996 and Dec. 1998. A total of 7,430 specimens for the PCR, 16,163 specimens for the TB BACTEC culture and 4,810 specimens (3,167 patients) submitted for both PCR and BACTEC culture were analyzed for the clinical evaluation of PCR. RESULTS: When compared with BACTEC culture results, PCR had a sensitivity of 66.1% (127/192) and a specificity of 97.0% (3,631/3,740) for the detection of M. tuberculosis in respiratory specimens. The sensitivity, specificity, positive and negative predictive values for the PCR assay for the diagnosis of tuberculosis patients were 64.7%, 98.5%, 84.8%, and 95.6%, respectively; the values for BACTEC culture were 86.7%, 100%, 100%, and 98.3%, respectively. In addition, 39 out of 71 PCR positive cases which were BACTEC culture-negative had clinical data supporting the diagnosis of tuberculosis. The average detection times were 5 hours in PCR but 15.8 days in BACTEC culture. CONCLUSIONS: This study shows that PCR itself is not satisfactory enough to detect M. tuberculosis in clinical specimens. However, when it combines with BACTEC culture, they can be a powerful and rapid diagnostic tool to detect M. tuberculosis in clinical specimens.