The natural history and treatment of experimental endocarditis due to heterogeneous and homogeneous methicillin-resistant Staphylococcus epidermidis was investigated. Amoxicillin/clavulanate or vancomycin were administered for 3 days via a computerized pump to mimic human drug kinetics in animals. After challenge with the minimum inoculum producing 90% of infections (ID90), bacteria in the vegetations grew logarithmically for 16 h. Then, bacterial densities stabilized (at approximately 10(8) cfu/g) and growth rates sharply declined. Both regimens cured > or = 60% of endocarditis (due to heterogeneous or homogeneous bacteria) when started 12-16 h after infection, although the bacterial densities in the vegetations had increased by 20 times in between. In contrast, treatment started after 24 h failed in most animals, while bacterial densities had not increased any more. Thus, while both regimens were equivalent, the therapeutic outcome was best predicted by growth rates in the vegetations, not by bacterial densities. These observations highlight the importance of phenotypic tolerance developing in vivo.