PK/PD Indices of Antibiotics Predicted by Mechanism-Based PKPD Models
Elisabet I. Nielsen (1,2), Otto Cars (3), Lena E. Friberg (1)
(1) Department of Pharmaceutical Biosciences, Uppsala University, Uppsala, Sweden; (2) Hospital Pharmacy, University Hospital, Uppsala, Sweden; (3) Department of Medical Sciences, section of Infectious Diseases, Uppsala University, Uppsala, Sweden
Objectives: Currently used dosing regimens of antibiotics are generally based on summary endpoints. One of three PK/PD indices (Cmax/MIC, AUC/MIC, or T>MIC) is selected for each drug and the magnitude of the index required for efficacy is assessed and assumed to be consistent across patient populations. Mechanism-based PKPD models, considering the time-course of efficacy, may provide for a better characterization of the exposure-response relationship. The aim was to evaluate the capability of PKPD models developed based on in vitro data (1, 2) in identifying the currently used PK/PD indices and their magnitudes reported for efficacy. A secondary aim was to challenge the methodology used in the establishment of the PK/PD indices, and thereby their consistency across study conditions.
Methods: Six drugs (benzylpenicillin, cefuroxime, erythromycin, gentamicin, moxifloxacin, and vancomycin) representing a broad selection of PK and PD characteristics were investigated. For each drug, a dose fractionation study was simulated, using a wide range of total daily doses given as intermittent doses or as constant drug exposure. The time-course of the drug concentration (PK) and the bacterial response to drug exposure (PKPD) were predicted. Non-linear least squares regression analyses determined the PK/PD index being most predictive of the effect.
Results: In accordance to previous findings, T>MIC was the PK/PD index best correlated to the PD endpoint for benzylpenicillin and cefuroxime, while AUC/MIC was the best predictor of the effect for the remaining four antibacterial drugs. The estimated magnitudes of the PK/PD index required for efficacy corresponded well with those earlier observed clinically. However, the choice and magnitude of the PK/PD index were both shown to be sensitive to study design, the PK profile of the drug and the uncertainty in MICs.
Conclusions: The currently used PK/PD indices were successfully selected based on predictions from mechanism-based PKPD models for all six drugs investigated. This supports that PKPD models based on in vitro time-kill curve studies can be predictive of antibacterial effects observed in vivo. As the selection and magnitude of the PK/PD indices were sensitive to study design and not always consistent between patient subpopulations with different PK, the PK/PD indices can extrapolate poorly across patient sub-populations and mechanism-based models may be of more predictive value in development of dosing regimens.
 Nielsen EI, et al. Antimicrob Agents Chemother. 2007 Jan;51(1):128-36.
 Mohamed AF, et al. PAGE 19. Abstr 1876 [www.page-meeting.org/?abstract=1876]