We evaluated the ability of the AccuProbe (Gen-Probe, San Diego, Calif.) to detect Mycobacterium gordonae and Mycobacterium avium complex directly in liquid medium flagged positive by the MB/BacT (Organon Teknika Corp., Durham, N.C.). Seventy-one bottles from clinical specimens containing M. gordonae and 34 containing M. avium, confirmed by culture, were tested by direct AccuProbe assay for both organisms after additional incubation for > or = 48 h and centrifugation at 4,500 x g for 15 min. Relative light unit (RLU) values were analyzed using the manufacturer's recommended cutoff of 30,000 RLU and a lower cutoff of 10,000 RLU. Using the 30,000 RLU cutoff, 55 of 71 (77.5%) specimens containing M. gordonae yielded positive results, whereas 28 of 34 (82.3%) M. avium complex specimens were correctly identified by direct probe. No specimens shown by culture to contain either M. gordonae or M. avium complex tested positive with the probe for the opposite organism (100% specificity). When the cutoff was lowered to 10,000 RLU, 67 of 71 M. gordonae (94.4%) and 32 of 34 M. avium complex (94.1%) specimens were correctly identified. This difference was significant for M. gordonae (P = 0.004) but not for M. avium complex (P = 0.26) compared to detection using the recommended RLU cutoff. Specificity was 100% for specimens containing M. gordonae that were tested with the M. avium complex probe using the 10,000 RLU cutoff, whereas specificity for specimens containing M. avium complex tested with the M. gordonae probe was 97%. Using a lower RLU cutoff for determining a positive result using the M. gordonae or M. avium complex probes when testing instrument-positive MB/BacT bottles directly will improve sensitivity without substantially compromising specificity.