Bacterial endocarditis is a difficult infection to cure, due to poor penetration of antibiotics into infected vegetations, altered metabolic state of bacteria within the lesion, and absence of adequate host-defense cellular response which could cooperate with antibiotic action. The contribution of animal models to a better understanding of the pathophysiology of the infection and to definition and improvement of therapeutic regimens of endocarditis in humans remains of great importance due to the difficulties encountered in clinical trials. The advantage of the experimental model is that besides the fact that it closely simulates the characteristics of the infection in humans, it provides clear endpoints which allow statistical comparisons among different therapeutic regimens: number of bacteria per gram of tissue, frequency of emergence of resistance, positivity of blood cultures, death vs. survival rates, and percentage of relapses after treatment has been stopped. All these parameters are more sensitive and more easy to study than in humans. The animal model has definitively established that bactericidal therapy is warranted and that in vitro susceptibility tests, especially those evaluating the killing rate, have a good predictive value on the therapeutic outcome. Two main aspects are discussed for their relevance to human therapy: (i) the kinetics of antibiotic diffusion into vegetations, with special reference to data obtained with autoradiography; and (ii) the specificity of some pharmacodynamic aspects of antibiotics in endocarditis, including the clinical consequences of these two parameters with respect to antibiotic dosing regimens and length of therapy. Animal models have also helped to define the importance of antibiotic dosing strategies to achieve in vivo synergism which appears essential to increase the rate of both bacteriological and clinical cure. Experiments have led to similar conclusions as those obtained with other models. The in vitro killing rate and the antibiotic levels obtained in situ are predictive of efficacy. Cautious extrapolation of experimental data to humans appears warranted.
