A us tra liaThe aim of this study was to evaluate the clinical use of a new culture system for the isolation of mycobacteria. Routine clinical specimens were cultured in the Mycobacteria Growth Indicator Tube, the radiometric Bactec 460 TB system and on Lowenstein Jensen (LJ) medium to compare recovery rates and times for detection of mycobacteria and contamination rates. MGIT was tested for its ability to support the growth of a wide range of mycobacterial species. Acid-fast bacilli (AFB) were detected on direct smears of 76 of 603 clinical specimens and mycobacteria were isolated by at least one method from 109 specimens; 93% of these were detected in the MGIT, 95% in the Bactec 460 TB system and 87% on L J medium. The MGIT, Bactec and L J media detected 92%, 97% and 95%, respectively, of 61 M. tuberculosis isolates and 94%, 94% and 77% of the 48 isolates belonging to the M. avium complex (MAC). The mean detection times in MGIT, Bactec and L J media for M. tuberculosis were 22, 14 and 27 days respectively, and for MAC were 14, 12, and 29 days, respectively. Growth of M. tuberculosis was detected in Bactec, within 4 weeks, in 93% of the 61 culture-positive specimens, compared with only 61% in MGIT and 66% on LJ. The number of MAC detected within 4 weeks was similar in Bactec and MGIT, but less in L J medium. Differences in sensitivity and time to detection of growth between media were greater for specimens in which AFB were not detected on direct smear than those on which AFB were seen. Contamination rates were similar in the three systems (3-4%). MGIT supported the growth of all 28 Mycobacterium spp. inoculated. MGIT has significant safety advantages and is less labour intensive than other methods, but the time to detection of M. tuberculosis, especially in smear-negative specimens, was longer in MGIT than in Bactec.