To assess the relative yields in automated microbial detection systems of bacteria and yeasts isolated from the blood of adult patients with suspected sepsis, we compared the new VersaTREK system (VTI) (TREK Diagnostic Systems, Cleveland, OH) to the BacT/ALERT 3D system (3D) (bioMérieux, Inc., Durham, NC). Identical protocols were followed for the two systems. Paired aerobic (REDOX 1) and anaerobic (REDOX 2) VTI media were compared with standard aerobic (SA) and anaerobic (SN) 3D media; each of the four culture bottles was filled with 6 to 9 ml of blood. All bottles flagged positive by the instruments were subcultured to determine both true-positive (growth) and false-positive (no growth) cultures. Additionally, to assess false-negative bottles, terminal subcultures were done on all negative companion bottles to true-positive bottles. All isolates were identified by standard methods. All 4 bottles were adequately filled and yielded 413 clinically significant isolates in 5,389 (79%) of the 6,786 4-bottle sets obtained. Although no overall difference in yield or in time to detection was detected between the two systems, significantly more streptococci and enterococci as a group were detected by VTI. Moreover, significantly more microorganisms were detected by VTI for patients receiving antimicrobial therapy. The two systems were comparable (P, not significant) at detecting the 179 unimicrobial episodes of bacteremia seen. False-positive rates for aerobic and anaerobic bottles, respectively, were 1.6% and 0.9% for VTI and 0.7% and 0.8% for 3D. We conclude that the VTI and 3D systems are comparable for detection of bloodstream infections with bacteria or yeasts.