Background Little has been published on the performance of tuberculosis PCR with respect to the quality of tissue specimens. Laboratories often receive liquid samples from fine-needle aspirates with no visible tissue for testing. The sensitivity of tuberculosis (TB) PCR on these specimens is unknown.

Methods TB PCR was compared to culture or a combination of clinical and histopathological evidence of tuberculosis; a separate analysis excluded patients with current or previous treatment.

Results Sixty-five patients had 81 positive samples; 69 by PCR and 43 by culture. Excluding those on treatment, 51 of 57 (89%) were PCR positive versus 43 of 61 (70%) by culture. 44 samples had ‘no visible tissue’ noted. Five were PCR positive; only one was culture positive. At least two samples were falsely negative.

Conclusions Sensitivity of TB PCR is superior to culture on tissue. Five of seven TB cases with no visible tissue were PCR positive. The quality of the specimen deserves comment, as the two (5%) known false negatives are of concern.