BACKGROUND: Mycobacterial culture is the gold standard test for diagnosing tuberculosis (TB), but it is time-consuming. Polymerase chain reaction (PCR) is a highly sensitive and specific method that can reduce the time required for diagnosis. The diagnostic efficacy of PCR differs, so this study determined the actual sensitivity of TB-PCR in tissue specimens. METHODS: We retrospectively reviewed 574 cases. The results of the nested PCR of the IS6110 gene, mycobacterial culture, TB-specific antigen-induced interferon-γ release assay (IGRA), acid-fast bacilli (AFB) staining, and histological findings were evaluated. RESULTS: The positivity rates were 17.6% for PCR, 3.3% for the AFB stain, 22.2% for mycobacterial culture, and 55.4% for IGRA. PCR had a low sensitivity (51.1%) and a high specificity (86.3%) based on the culture results of other studies. The sensitivity was higher (65.5%) in cases with necrotizing granuloma but showed the highest sensitivity (66.7%) in those with necrosis only. The concordance rate between the methods indicated that PCR was the best method compared to mycobacterial culture, and the concordance rate increased for the methods using positive result for PCR or histologic features. CONCLUSIONS: PCR of tissue specimens is a good alternative to detect tuberculosis, but it may not be as sensitive as previously suggested. Its reliability may also be influenced by some histological features. Our data showed a higher sensitivity when specimens contained necrosis, which indicated that only specimens with necrosis should be used for PCR to detect tuberculosis.