Quantitative-competitive polymerase chain reaction (QPCR) was performed on serial sputum samples from 22 consecutive cases of acid fast bacilli (AFB) smear-positive pulmonary tuberculosis. Of 94 specimens, 55, 72, and 83% were positive by culture, AFB smear, and QPCR, respectively. Of 52 culture-positive specimens, 6% were negative by PCR, and 13% were negative by AFB smear. Of 42 culture-negative specimens, AFB smear and QPCR were positive in 55 and 61%, respectively. AFB smear and QPCR results were strongly correlated (r = 0.75, p < 0.001), but each correlated less strongly with culture (r = 0.54, p < 0.005 for smear and r = 0.52, p < 0.005 for QPCR). When patients were classified by microbiologic response, responders tended to have less DNA in their sputum and shorter time to a negative PCR result compared to nonresponders. These data do not suggest a great advantage of QPCR over AFB smear for predicting culture results in patients with pulmonary tuberculosis.