Use of pharmacodynamic indices to predict efficacy of combination therapy in vivo.

Abstract

Although combination therapy with antimicrobial agents is often used, no available method explains or predicts the efficacies of these combinations satisfactorily. Since the efficacies of antimicrobial agents can be described by pharmacodynamic indices (PDIs), such as area under the concentration-time curve (AUC), peak level, and the time that the concentration is above the MIC (time>MIC), it was hypothesized that the same PDIs would be valid in explaining efficacy during combination therapy. Twenty-four-hour efficacy data (numbers of CFU) for Pseudomonas aeruginosa in a neutropenic mouse thigh model were determined for various combination regimens: ticarcillin-tobramycin (n = 41 different regimens), ceftazidime-netilmicin (n = 60), ciprofloxacin-ceftazidime (n = 59), netilmicin-ciprofloxacin (n = 38) and for each of these agents given singly. Multiple regression analysis was used to determine the importance of various PDIs (time>MIC, time>0.25 x the MIC, time>4 x the MIC, peak level, AUC, AUC/MIC, and their logarithmically transformed values) during monotherapy and combination therapy. The PDIs that best explained the efficacies of single-agent regimens were time>0.25 x the MIC for beta-lactams and log AUC/MIC for ciprofloxacin and the aminoglycosides. For the combination regimens, regression analysis showed that efficacy could best be explained by the combination of the two PDIs that each best explained the response for the respective agents given singly. A regression model for the efficacy of combination therapy was developed by use of a linear combination of the regression models of the PDI with the highest R(2) for each agent given singly. The model values for the single-agent therapies were then used in that equation, and the predicted values that were obtained were compared with the experimental values. The responses of the combination regimens could best be predicted by the sum of the responses of the single-agent regimens as functions of their respective PDIs (e.g., time>0.25 x the MIC for ticarcillin and log AUC/MIC for tobramycin). The relationship between the predicted response and the observed response for the combination regimens may be useful for determination of the presence of synergism. We conclude that the PDIs for the individual drugs used in this study are class dependent and predictive of outcome not only when the drugs are given as single agents but also when they are given in combination. When given in combination, there appears to be a degree of synergism independent of the dosing regimen applied.