Specific amplification of Mycobacterium tuberculosis DNA was investigated as an alternative to conventional microbiologic follow-up in 31 cases of smear- and culture-positive pulmonary tuberculosis. Strand displacement amplification (SDA) and the polymerase chain reaction (PCR) were applied to 438 sequential sputum specimens: 67 (15%) were positive by culture, 248 (57%) by SDA, and 231 (53%) by PCR (chi2=3.94, P=.05). Of 200 specimens collected >180 days after treatment started, none yielded positive cultures, while 50 (25%), representing 16 patients, were positive by both DNA assays. A weak correlation was demonstrated between DNA persistence in sputum and duration of culture positivity (r=0.45, P=.01), although no correlation was found with the radiographic extent of disease. The inability to distinguish live and dead organisms precludes DNA amplification from use in therapeutic monitoring. For this purpose, quantitative RNA assays are needed if such techniques are to supplant conventional microbiology.