Several studies have reported using methods based on polymerase chain reaction (PCR) to detect Mycobacterium tuberculosis in respiratory tract specimens. However, little is known about the actual clinical utility of PCR-based tests, and it is uncertain if PCR technology can be transferred to the clinical laboratory. To determine its utility, we evaluated a commercially developed PCR test system in a clinical laboratory using consecutive respiratory tract specimens. Microscopic examination of smears stained with acid-fast bacilli (AFB), culture, and a PCR-based test (Amplicor Mycobacterium tuberculosis assay; Roche Molecular Systems) were used to evaluate 535 consecutive sputum and bronchoalveolar lavage specimens from 227 patients. A clinical case definition of tuberculosis was used as the reference-standard to determine the utility of all diagnostic tests. For all specimens from patients with a new or a treatment-failure case of pulmonary tuberculosis, the positivity rate of PCR (58%) was similar to that of culture (56%) (p > 0.90) and substantially greater than microscopic examination of AFB-stained smears (22%) (p < 0.001). PCR and culture detected M. tuberculosis in 46 and 43%, respectively, of the specimens from patients who did not have AFB on microscopic examination of their respiratory tract specimens (p > 0.90). PCR had a false positive rate of 0.8%. In several instances, PCR detected M. tuberculosis when culture did not; and vice versa. The clinical utility of this PCR-based test is similar to that of culture for detecting M. tuberculosis in respiratory tract specimens.(ABSTRACT TRUNCATED AT 250 WORDS)