The aim of this study was to determine the sensitivity of different methods--two commercial polymerase chain reaction (PCR) kits (a protocol of nested PCR and a protocol of amplification of the IS6110 insertion element), the radiometric Bactec system, the Septi-Chek AFB culture system, and culture in Löwenstein-Jensen (LJ) solid medium--for the detection of Mycobacterium tuberculosis. One hundred clinical samples from 51 patients with culture-positive tuberculosis (81 specimens) and 19 controls (19 specimens) were used. Eighty-nine percent of the samples were smear negative. In the 81 specimens obtained from patients with tuberculosis, the frequency of positivity was 66.6% for nested PCR, 63% for culture in liquid media, 38.3% for IS6110 assay, and 28.4% for culture in LJ medium. In 18 samples obtained by invasive procedures in patients with tuberculosis, mycobacterial DNA was detected by nested PCR in 83.3% (including all samples positive by culture on liquid media), by culture in liquid media in 77.7% by culture on LJ medium in 27.7%, and by the IS6110 assay in 11.1%. No false-positive results were obtained from the negative control specimens with any of the techniques tested. The sensitivity of the reamplification protocol appears to be superior to that of the IS6110 assay and similar to that of the Bactec system.