faster alternative techniques are required to improve the diagnosis and control of pulmonary tuberculosis.

To evaluate the sample quality in the performance of PCR for diagnosis of pulmonary tuberculosis.

During one year, sputum samples were collected from 72 pulmonary tuberculosis patients and 12 non-tuberculosis controls, which were admitted to the Nereu Ramos hospital, Florianópolis city, Brazil. The samples were subjected to Ziehl-Neelsen-stained sputum smear microscopy and Lowestein-Jensen medium culture, which were defined as gold standard tests for mycobacteria, and polymerase chain reaction (PCR). Those samples that presented more than 40% of viable cells and less than 25% of epithelial cells were defined as high quality samples.

PCR showed sensitivity of 55.6%, specificity of 41.7%, positive predictive value of 85.1%, negative predictive value of 13.5%, and accuracy of 53.6%. High quality samples showed sensitivity of 72.4%, specificity of 50%, positive predictive value of 91.3%, negative predictive value of 20%, and accuracy of 69.7%. Low quality samples showed sensitivity of 44.2%, specificity of 37.5%, positive predictive value of 79.2%, negative predictive value of 11.1%, and accuracy of 43.1%.

use of high quality samples improved significantly the PCR performance, especially on their sensitivity and positive predictive values.