We present data from a clinical trial study in which an automated version (Galileo) of a previously described Q-Beta replicase-amplified probe assay (J. S. Shah et al., J. Clin. Microbiol. 33:1435-1441, 1995) was used for the direct detection of Mycobacterium tuberculosis complex in sputum. The assay was designed to target specific regions of 23S rRNA found in M. tuberculosis, Mycobacterium bovis, Mycobacterium africanum, and Mycobacterium microti and had a sensitivity ranging from approximately <10 to 300 CFU. The assay was tested for cross-hybridization by using large numbers (e.g., 10(5)to 10(10) CFU/assay) of 133 other organisms commonly found in respiratory tract samples, including non-M. tuberculosis Mycobacterium spp., other bacteria, fungi, and viruses. All of these competitors tested negative by the assay. Automated assay results for 780 respiratory tract samples (sputum or bronchoalveolar lavage specimens) collected and tested at three trial sites in the United States) were compared with the results of culture and acid-fast microscopy. Aliquots of conventionally digested and decontaminated sputum pellets were heated at 100 degrees C and mechanically disrupted prior to hybridization and background reduction, amplification, and detection in a closed disposable test pack. Pertinent elements of individual patient histories relating to tuberculosis exposure, previous active disease, antituberculosis therapy status, etc., were considered in the resolution of discrepant results for 48 (assay false-positive) samples. Seventy-one of 90 (78.9%) culture-positive samples were positive when tested in the Galileo assay, while 7% of culture-negative samples were assay positive, corresponding to a sensitivity of 79% and a specificity of 93%. Following resolution of discrepant results by chart review, the sensitivity and specificity for the Q-Beta replicase amplification assay with the Galileo analyzer were 84 and 97%, respectively. A total of 69.2% of smear-negative (culture positive) samples were detected by the assay. Ten test packs at a time were automatically processed by the Galileo analyzer without operator intervention following loading of samples. The first result was reported in approximately 3 h, and the last result was available in 6.5 h. To our knowledge, this is the first report of a clinical study with a fully automated amplification probe hybridization assay for the detection of pathogens directly from a clinical specimen.