Monte Carlo simulations based on phase 1 studies predict target attainment of ceftobiprole in nosocomial pneumonia patients: a validation study.

Monte Carlo simulation (MCS) of antimicrobial dosage regimens during drug development to derive predicted target attainment values is frequently used to choose the optimal dose for the treatment of patients in phase 2 and 3 studies. A criticism is that pharmacokinetic (PK) parameter estimates and variability in healthy volunteers are smaller than those in patients. In this study, the initial estimates of exposure from MCS were compared with actual exposure data in patients treated with ceftobiprole in a phase 3 nosocomial-pneumonia (NP) study (NTC00210964). Results of MCS using population PK data from ceftobiprole derived from 12 healthy volunteers were used (J. W. Mouton, A. Schmitt-Hoffmann, S. Shapiro, N. Nashed, N. C. Punt, Antimicrob. Agents Chemother. 48:1713-1718, 2004). Actual individual exposures in patients were derived after building a population pharmacokinetic model and were used to calculate the individual exposure to ceftobiprole (the percentage of time the unbound concentration exceeds the MIC [percent fT > MIC]) for a range of MIC values. For the ranges of percent fT > MIC used to determine the dosage schedule in the phase 3 NP study, the MCS using data from a single phase 1 study in healthy volunteers accurately predicted the actual clinical exposure to ceftobiprole. The difference at 50% fT > MIC at an MIC of 4 mg/liter was 3.5% for PK-sampled patients. For higher values of percent fT > MIC and MICs, the MCS slightly underestimated the target attainment, probably due to extreme values in the PK profile distribution used in the simulations. The probability of target attainment based on MCS in healthy volunteers adequately predicted the actual exposures in a patient population, including severely ill patients.