A prospective 2-month trial involving 617 respiratory tract specimens was conducted to compare sensitivity, specificity, and predictive values of the newly developed Gen-Probe Amplified Mycobacterium Tuberculosis Direct Test kit (AMTDT; Gen-Probe, Inc., San Diego, Calif.) for the rapid detection of Mycobacterium tuberculosis and of fluorescent acid-fast staining versus combined BACTEC 12B and solid-medium cultures as the "gold standard." A total of 590 specimens were culture and AMTDT negative. Twenty-one (3.4%) cultures yielded M. tuberculosis. Of these, 15 (71.4%) were detected by AMTDT, whereas 6 (28.6%) were missed. M. tuberculosis did not grow in six (28.6%) of AMTDT-positive specimens derived from three patients under treatment for tuberculosis. AMTDT exhibited a sensitivity, a specificity, a negative predictive value, and a positive predictive value of 71.4, 99, 99, and 71.4%, respectively. In comparison, the same values for fluorescent microscopy were 66.7, 98.3, 98.8, and 58.3%, respectively. AMTDT was easy to perform and highly specific. However, a screening test would require an improved sensitivity and, when feasible, the implementation of an internal amplification control.