Direct detection of Mycobacterium tuberculosis by means of a commercial ligase chain reaction DNA amplification method (LCx M. tuberculosis; Abbott Diagnostics Division, Abbott Park, Ill.) was investigated with 511 (including 147 extrarespiratory) specimens collected from 358 patients. LCx results were compared with standard microbiological data, and conflicting cases were resolved according to the final clinical diagnosis. M. tuberculosis was detected in 45 of 358 subjects by means of the LCx test. The test was negative for all 30 specimens with mycobacteria other than M. tuberculosis. The sensitivity, specificity, and positive and negative predictive values for the LCx test, compared with culture results, were 93.90, 92.31, 70.00, and 98.75%, respectively; these values rose in resolved cases to 95.53, 99.25, 97.27, and 98.75%, respectively. With respiratory specimens, for which the LCx system is licensed, the sensitivity reached 98.97%. In patients with a final clinical diagnosis of tuberculosis the sensitivity of the LCx system was 89.36% compared to 82.98% for cultures and 78.72% for microscopy. We conclude that the LCx test is user friendly, rapid, fairly sensitive, and highly specific. It can also be effectively used on extrapulmonary specimens provided possible false-negative results are taken into account. However, the use of LCx test appears to be less appropriate for the monitoring of antituberculosis therapy, as the majority of samples from treated tuberculosis patients gave consistently positive results, despite the sterilization of cultures.