OBJECTIVES: Laboratory detection of vancomycin-intermediate Staphylococcus aureus (VISA) and their heterogeneous VISA (hVISA) precursors is difficult. Thus, it is possible that vancomycin failures against supposedly vancomycin-susceptible S. aureus are due to undiagnosed VISA or hVISA. We tested this hypothesis in experimental endocarditis.¦METHODS: Rats with aortic valve infection due to the vancomycin-susceptible (MIC 2 mg/L), methicillin-resistant S. aureus M1V2 were treated for 2 days with doses of vancomycin that mimicked the pharmacokinetics seen in humans following intravenous administration of 1 g of the drug every 12 h. Half of the treated animals were killed 8 h after treatment arrest and half 3 days thereafter. Population analyses were done directly on vegetation homogenates or after one subculture in drug-free medium to mimic standard diagnostic procedures.¦RESULTS: Vancomycin cured 14 of 26 animals (54%; P<0.05 versus controls) after 2 days of treatment. When vegetation homogenates were plated directly on vancomycin-containing plates, 6 of 13 rats killed 8 h after treatment arrest had positive cultures, 1 of which harboured hVISA. Likewise, 6 of 13 rats killed 3 days thereafter had positive valve cultures, 5 of which harboured hVISA. However, one subculture of vegetations in drug-free broth was enough to revert all the hVISA phenotypes to the susceptible pattern of the parent. Thus, vancomycin selected for hVISA during therapy of experimental endocarditis due to vancomycin-susceptible S. aureus. These hVISA were associated with vancomycin failure. The hVISA phenotype persisted in vivo, even after vancomycin arrest, but was missed in vitro after a single passage of the vegetation homogenate on drug-free medium.¦CONCLUSIONS: hVISA might escape detection in clinical samples if they are subcultured before susceptibility tests.