The performance of a nested PCR-based assay (the RAPID BAP-MTB; AsiaGen, Taichung, Taiwan) and the BD ProbeTec ET (DTB) system (Becton Dickinson, Sparks, Md.) for detection of Mycobacterium tuberculosis was evaluated with 600 consecutive clinical samples. These samples, including 552 respiratory specimens and 48 nonrespiratory specimens, were collected from 333 patients treated at National Taiwan University Hospital from September to October 2003. The results of both assays were compared to the gold standard of combined culture results and clinical diagnosis. The overall sensitivity and specificity of the RAPID BAP-MTB assay for respiratory specimens were 66.7% and 97.2%, respectively, and for the DTB assay they were 56.7% and 95.3%, respectively. The positive and negative predictive values for the RAPID BAP-MTB were 74.1% and 96.0%, respectively, and for the DTB assay they were 59.6% and 94.7%, respectively. For smear-negative samples, the sensitivity of the RAPID BAP-MTB and DTB assays was 57.1% and 40.5%, respectively. The RAPID BAP-MTB assay produced 14 false-positive results in 14 samples, including one of the six samples yielding Mycobacterium abscessus, one of the six samples yielding Mycobacterium avium intracellulare, one sample from a patient with a history of pulmonary tuberculosis with complete treatment, and three samples from three patients with a previous diagnosis of tuberculosis who were under treatment at the time of specimen collection. Among the 48 nonrespiratory specimens, the RAPID BAP-MTB assay was positive in one biopsy sample from a patient with lumbar tuberculous spondylitis and one pus sample from a patient with tuberculous cervical lymphadenopathy. Our results showed that the RAPID BAP-MTB assay is better than the DTB assay for both respiratory specimens and nonrespiratory specimens. The overall time for processing this assay is only 5 h. In addition, its diagnostic accuracy in smear-negative samples is as high as in smear-positive samples.