The performance characteristics of the E-test (AB Biodisk, Solna, Sweden), the ATB Staph, the Rapid ATB Staph, and the Vitek GPS-503 card (bioMérieux, La Balme Les Grottes, France) methods for the detection of oxacillin resistance in a collection of staphylococci with a high proportion of troublesome strains were evaluated. Sixty-four Staphylococcus aureus strains and 76 coagulase-negative staphylococcal strains were tested. All strains were mecA positive and were characterized by the oxacillin agar screen plate test; 75 (53.6%) were found to be heterogeneous by a large-inoculum oxacillin disk diffusion assay, and oxacillin MICs for 89 (63.6%) were < or = 32 microg/ml. Three (4.7%) S. aureus strains and 25 (32.9%) coagulase-negative strains were classified as susceptible by the E-test, as defined by the National Committee for Clinical Laboratory Standards (NCCLS) oxacillin breakpoint (MIC < or = 2 microg/ml). The ATB Staph method failed to detect oxacillin resistance in 7 (11%) S. aureus isolates and 32 (42.1%) coagulase-negative isolates. The MICs for all but six of these discrepant isolates were < or = 16 microg/ml. The Rapid ATB Staph method was tested against S. aureus strains only and yielded 15 (23.4%) false-susceptible results for strains for which the MICs were < or = 32 microg/ml. The Vitek system was the best-performing system, since it failed to detect oxacillin resistance in only 3 (4.7%) S. aureus strains and 15 (19.7%) coagulase-negative strains, the MICs for all of which were < or = 2 microg/ml. These data indicate that (i) the performance of the two ATB Staph systems can be limited when the prevalence of borderline-heteroresistant staphylococci is high and (ii) the unreliability of the E-test and the Vitek methods for detecting resistant coagulase-negative strains might be reduced by the potential revision of the oxacillin breakpoint currently recommended by the NCCLS.