**DOI:**10.1128/AAC.01181-06

## ABSTRACT

Ceftobiprole is a broad-spectrum cephalosporin with activity against methicillin-resistant *Staphylococcus aureus* (MRSA) that is undergoing phase III trials for the treatment of complicated skin and skin structure infections and nosocomial pneumonia. The objectives were to describe the pharmacodynamic profiles of ceftobiprole given at 500 mg intravenously (i.v.) every 8 h (q8h) (2-h infusion) and 500 mg i.v. every 12 h (q12h) (1-h infusion) to determine the overall probability of target attainment (PTA) by weighting for the expected distributions of renal function in the populations of interests, to determine the PTA against representative pathogens encountered in clinical trials, and to determine the optimal renal dose adjustment for ceftobiprole at 500 mg i.v. q8h (2-h infusion). Data for a total of 150 subjects in phase I/II trials were analyzed by using the population pharmacokinetic modeling program BigNPOD (nonparametric optimal design). Monte Carlo simulation was performed with the ADAPT II program to estimate the PTA at which the free drug concentrations exceed the MIC for 30 to 60% of the dosing interval (30 to 60% *fT* > MIC). For ceftobiprole at 500 mg i.v. q12h, the probabilities of achieving 30% and 50% *fT* > MIC exceeded 90% for MICs ≤2 mg/liter and ≤1 mg/liter, respectively, For ceftobiprole at 500 mg i.v. q8h, the probabilities of achieving 40 and 60% *fT* > MIC exceeded 90% for MICs ≤4 mg/liter and ≤2 mg/liter, respectively. For ceftobiprole at both 500 mg i.v. q12h and 500 mg i.v. q8h, the probability of achieving a nearly bactericidal effect (50% *fT* > MIC) exceeded 90% for methicillin-susceptible *S. aureus* and MRSA. For gram-negative pathogens, the PTA for achieving a nearly maximal bactericidal effect (60% *fT* > MIC) for ceftobiprole at 500 mg i.v. q8h exceeded 90% for non-AmpC-producing gram-negative organisms. Ceftobiprole at 500 mg i.v. q12h, for patients who had a creatinine clearance rate of ≤50 ml/min, was identified as the most appropriate treatment regimen for patients who require renal dose adjustment for mild to moderate renal impairment.

Methicillin-resistant *Staphylococcus aureus* (MRSA) has recently emerged as a serious problem in the community (13, 18, 21, 23, 25-27), and it already represents 40 to 50% of staphylococcal isolates in many health care institutions (2). Given the magnitude of the problem, new agents with activity against MRSA are a welcome addition to the physician's therapeutic armamentarium. While new agents such as linezolid, new glycopeptides, and daptomycin have become available (5, 31), clinicians have long experience and trust in β-lactam antibiotics as therapeutic agents for life-threatening infections, and these drugs also have a broad margin of safety.

MRSA isolates are endowed with a *mecA* cassette that allows the strain to express a new penicillin-binding protein (PBP 2a or PBP 2′). This protein has a markedly diminished affinity for acylating most β-lactam drugs, which is the genesis of its clinical resistance to such agents (17). Ceftobiprole, an investigational cephalosporin currently in phase III trials for complicated skin and skin structure infections (cSSSIs) and nosocomial pneumonia (NP), represents a major advance among the β-lactam antibiotics, in that its structure allows it to bind to PBP 2a and thus kill MRSA (4). Indeed, a considerable amount of preclinical data demonstrate that ceftobiprole acts in a bactericidal fashion in preclinical infection models, both in vitro and in vivo (3, 6, 9, 10, 12, 15, 16, 19, 33, 34). The drug possesses excellent activity against MRSA and other resistant gram-positive pathogens, such as penicillin-resistant *Streptococcus pneumoniae* (3, 6, 9, 12, 16, 19). Ceftobiprole also possesses activity against the common gram-negative pathogens often seen in the health care setting, making it a unique broad-spectrum agent with potential for early, empirical use (3, 9, 15, 34). Data from the laboratory of Andes and Craig demonstrated that pharmacodynamics of ceftobiprole are similar to those of other compounds within the cephalosporin class (1). For cephalosporins, pharmacodynamic studies have demonstrated that a bacteriostatic effect is achieved when free drug concentrations exceed the MIC for 30% of the dosing interval (30% *fT* > MIC) for staphylococci and 40% *fT* > MIC for gram-negative bacilli. Likewise, nearly maximal organism killing is achieved at 50% and 60% *fT* > MIC for staphylococci and gram-negative bacilli, respectively (1, 7, 11).

As ceftobiprole was entering phase III clinical trials, we believed that it was critical to examine the population pharmacokinetics of this agent in a large set of subjects and to perform a Monte Carlo simulation from this analysis to identify the probability of target attainment (PTA) for the candidate doses that might be used in these trials. We were most interested in examining the impact of altering the infusion time on PTA for the NP dosing regimen. Furthermore, as patients with renal functional impairment would be entering these trials, we wished to use the estimated creatinine clearance (CL_{CR}) as a covariate in the population pharmacokinetic analysis, so that we could examine the impacts of different degrees of renal functional impairment on the PTA. As we were examining the impact of renal functional impairment, we believed that it would add insight to examine the distributions of renal function encountered in two previous registration studies for the indications that will ultimately be sought with ceftobiprole, cSSSIs and NP (14, 32). These distributions would allow us to take an expectation, so that the ultimate PTA values would take into account pharmacokinetics, the MIC distribution, the effect of renal functional impairment, and the likely distribution of levels of renal function that would be encountered in the populations of interest. Finally, we wished to examine different strategies of dose adjustment for renal dysfunction and to explore the impact on the PTA as well as the degree of exposure just prior to the point of dose adjustment. The aim of the latter exercise was to find a CR_{CL} breakpoint and dose adjustment that would leave the PTA substantially unaltered and still not result in profound accumulation.

## MATERIALS AND METHODS

Subject population.A total of 150 subjects (29, 30) participated in the population pharmacokinetic analysis. Of these, 29 were patients who had participated in phase II trials of ceftobiprole. The CL_{CR}s for all subjects were estimated by the method of Cockcroft and Gault, and estimates of the glomerular filtration rate (GFR) ranged from 7 ml/min to 169 ml/min.

Drug assay.Ceftobiprole in serum was assayed by a validated liquid chromatography/tandem mass spectrometry assay. The assay was linear over the range of 20 to 6,400 ng/ml. The sensitivity of the assay with plasma was 20 ng/ml. The interassay precisions of the assay were 7% at 60 ng/ml and 6.6% at 4,800 ng/ml.

Population model.Because ceftobiprole was administered as a prodrug, we used a three-compartment model, in which the prodrug was infused at a constant rate into a hydrolysis compartment. An assumption of 100% hydrolysis was made. A first-order hydrolysis constant connected the hydrolysis compartment to the central compartment. The rest of the model was a standard two-compartment open model with first-order elimination from the central compartment and first-order intercompartmental transfer rate constants. Ceftobiprole clearance was made proportional to the estimated CL_{CR} as follows: (slope × estimated CL_{CR}) + clearance intercept.

Population pharmacokinetic analysis.All patients and all samples were analyzed simultaneously by using the population pharmacokinetic modeling program BigNPOD (nonparametric optimal design) of Leary et al. (20). The inverse of the estimated assay variance was used as the first estimate for weighting in the pharmacokinetic modeling. Weighting was based on the assumption that the total observed variance was proportional to assay variance, which was determined on a between-day basis. The analysis was performed with adaptive γ, a scalar which multiplies the polynomial described above and is optimized with each cycle to produce the best approximation to the homoscedastic assumption. Bayesian estimates were obtained for each subject by using the “population-of-one” utility in the BigNPOD program. The fit of the model to the data was assessed by examining an observed-predicted plot after the Bayesian step. The mean weighted error (predicted value − observed value) served as the measure of bias. The bias-adjusted mean weighted squared error served as the measure of precision.

Monte Carlo simulation.Monte Carlo simulation was performed for 9,999 subjects by using the ADAPT II package of D'Argenio and Schumitzky (8). The following regimens were evaluated: ceftobiprole at 500 mg intravenously (i.v.) every 8 h (q8h) (0.5-, 1-, and 2-h infusions) and ceftobiprole at 500 mg i.v. every 12 h (q12h) (1-h infusion). Both normal and log-normal distributions were evaluated. Distributional choice was made by examining the fidelity with which the original parameter values and their dispersions were recapitulated by the simulations. As subjects with various degrees of renal functional impairment were included in the data set, the estimated CL_{CR} was frozen at values of 20 ml/min, 40 ml/min, 60 ml/min, 80 ml/min, 100 ml/min, and 120 ml/min for the simulations; and PTAs were determined for MICs between 0.25 mg/liter and 8 mg/liter. The target values for analysis were 30%, 40%, 50%, and 60% *fT* > MIC. For staphylococci, 30% and 50% *fT* > MIC were the targets for stasis and nearly maximal killing, respectively. For gram-negative organisms, 40% and 60% *fT* > MIC were the targets for stasis and nearly maximal killing, respectively. Although Andes and Craig (1) reported that only ∼30% *fT* > MIC was required for stasis against gram-negative bacilli, we used 40% *fT* > MIC as the bacteriostatic target, based on the results of previous pharmacodynamic studies of cephalosporins and on the large standard deviation and 95% confidence interval (10 to 46%) surrounding the mean *fT* > MIC estimate in previous studies (1, 7, 11). Protein binding values and the consequent free drug fractions were provided by the sponsor (Johnson and Johnson Pharmaceutical Research and Development). Ceftobiprole was 16% bound and 84% free over the relevant concentration range.

The overall PTA was calculated by weighting for the expected distributions of renal function in the populations of interests. Weighting was accomplished by using the distribution of CL_{CR} observed in two previous registration studies for the indications that will ultimately be sought with ceftobiprole, cSSTIs and NP (14, 32). Specifically, the overall PTA at each MIC was calculated by summing the product of the PTA for each CL_{CR} by the fraction of patients in the clinical trial with that respective CL_{CR}. For example, the PTA at a given MIC with a CL_{CR} of 20 ml/min was multiplied by the fraction of patients in the trial having CL_{CR}s near that value (14). The same was true for estimated values of 40 ml/min, etc. All the values were then added to obtain a weighted value for the PTA at that MIC. We believed that this method would put the data into proper perspective and would thus have the greatest clinical value. Consistent with the recommendations of the CLSI, a PTA of 90% was considered acceptable, since the overwhelming majority of patients will achieve sufficient antibiotic exposures at this threshold.

Clinical trial data for cSSSIs and NP.The distribution of the CL_{CR}s from the 750-mg levofloxacin registration trials for cSSTI and NP indications were provided by the sponsor (Johnson and Johnson Pharmaceutical Research and Development) (14, 32). We used the demographics from the 750-mg levofloxacin registration trials as a measure of the CL_{CR} distributions because they were modern studies of substantial patient size that involved the same patient populations (patients with NP and cSSSIs) as the ceftobiprole phase III trials. Both arms of these studies were used to estimate the CL_{CR} distributions.

Microbiology data.Ceftobiprole MIC data from a set of United States and European gram-negative isolates, obtained from Focus BioInova 2004 surveillance studies, were used as a measure of the MIC distribution and frequency for the gram-negative organism-specific PTA analysis. Organisms and ceftobiprole MICs for MRSA and methicillin-susceptible *S. aureus* (MSSA) strains were identified from the ceftobiprole phase III cSSSI registration trial and were supplied by the sponsor (Johnson and Johnson Pharmaceutical Research and Development).

Renal dose adjustment analysis.Two dose adjustment strategies for ceftobripole at 500 mg i.v. q8h (2-h infusion) were explored for patients with renal functional impairment. In one strategy, the dosing interval and infusion time were maintained (q8h and 2 h, respectively) and the dose was halved (250 mg) at an estimated CL_{CR} of 30 ml/min. In the second strategy, the dose and infusion time were maintained at 500 mg over 2 h, but the interval was lengthened from q8h to q12h at an estimated CL_{CR} of 50 ml/min. We examined these two CL_{CR} thresholds because they are consistent with the renal dose adjustments for other β-lactams according to CL_{CR}s. PTAs for 30 to 60% *fT* > MIC were examined for each candidate regimen at the CL_{CR} dose adjustment threshold with the adjusted dose. The exposure ratio by the area under the concentration-time curve (AUC) for 24 h at steady state (AUC_{24SS}) was also calculated for each candidate regimen. Specifically, the exposure ratio was the AUC_{24SS} for ceftobiprole at 500 mg i.v. q8h (2-h infusion) when the CL_{CR} was fixed at the candidate renal dose adjustment threshold (30 and 50 ml/min) relative to the AUC_{24SS} distribution when the CL_{CR} was 100 ml/min. While an explicit concentration-toxicity relationship currently does not currently exist, the target maximal exposure was twofold or less, since studies with volunteers have demonstrated the tolerability of ceftobiprole at 1 g q8h.

## RESULTS

Subject population.The demographics of the 150 subjects are presented in Table 1. The patients participating in phase II clinical trials were older than the volunteer population, as would be expected. Interestingly, the subset of patients with infections secondary to injection drug use had estimated CL_{CR}s that tended to be higher than those for the other subjects in this analysis.

Population pharmacokinetic analysis.The population pharmacokinetic parameters are shown in Table 2. The value of the first-order hydrolysis rate constant is large and is concordant with the speed with which prodrug concentrations disappear. The volume of the central compartment is concordant with that presented in an earlier publication that examined the pharmacokinetics of this compound in volunteers. The clearance of ceftobiprole has an intercept term of 2.35 liter/h and a slope term of 0.51, indicating that for a patient with an estimated CL_{CR} of 4.8 liters/h (80 ml/min), the clearance would be 2.45 liters/h, giving an overall clearance of 4.8 liters/h: ceftobiprole clearance = (slope × estimated CL_{CR}) + clearance intercept.

The fit of the model to the data was good. After the Bayesian step, the observed-predicted plot showed a regression line of observed = (1.003 × predicted) + 0.627 (*r*^{2} = 0.947; *P* ≪ 0.001).

PTAs. (i) Ceftobiprole NP and cSSSI dosing schedules.The results of the Monte Carlo simulation with 9,999 simulated subjects for ceftobiprole at 500 mg i.v. q8h (NP dosing schedule) are displayed in Table 3. There was a mild to moderate effect of increasing the infusion time on the PTA. Specifically, there was a minor effect of the observed PTA when the infusion time was extended from 0.5 to 1 h. There was, however, a more noticeable improvement in the observed PTA when the infusion time was extended to 2 h. In addition, the alteration of the infusion time over the range of 0.5 h to 2.0 h provided a minimal impact on the PTA when the MIC was ≤1 mg/liter but had a greater impact when the MICs were ≥2 mg/liter.

The results of the Monte Carlo simulation for ceftobiprole administered at a 500-mg dose q12h as a 1-h intravenous infusion (cSSSI dosing schedule) are displayed in Table 4. The probability of achieving a static effect (30% *fT* > MIC) against *S. aureus* was ≥90% for MICs ≤2 mg/liter for all CL_{CR} values. The probability of achieving a nearly maximal bactericidal effect (50% *fT* > MIC) was >85% for MICs ≤1 mg/liter for all CL_{CR}s.

The overall PTAs for ceftobiprole at 500 mg i.v. q8h (2-h infusion) and ceftobiprole at 500 mg i.v. q12h (1-h infusion) are shown in Fig. 1A and B, respectively. It should be noted that the overall PTA shown in Fig. 1 at each MIC has been weighted by the probability that a patient has a specific estimated CL_{CR}, as determined from the previous registration trials for levofloxacin at 750 mg (14, 32). In the Monte Carlo simulation analysis of ceftobiprole at 500 mg i.v. q8h (2-h infusion), the probabilities of achieving 40 and 60% *fT* > MIC exceeded 90% for MICs ≤4 mg/liter and ≤2 mg/liter, respectively. For ceftobiprole at 500 mg i.v. q12h, the probabilities of achieving 30% and 50% *fT* > MIC exceeded 90% for MICs ≤2 mg/liter and ≤1 mg/liter, respectively.

(ii) Organism-specific analysis.The results of the PTAs for MRSA and MSSA are shown in Table 5. The MIC range for 170 MRSA isolates was 0.5 to 2.0 mg/liter, and the MIC range for 291 MSSA isolates was 0.12 mg/liter to 1.0 mg/liter. The MIC_{50} and MIC_{90} for MRSA were 0.5 mg/liter and 1 mg/liter, respectively. For MSSA, the MIC_{50} and MIC_{90} were 0.25 mg/liter and 0.5 mg/liter, respectively. Ceftobiprole at 500 mg q8h (2-h infusion) provided excellent PTAs for a nearly maximal bactericidal effect (50% *f*T > MIC) for essentially the whole distribution of MSSA and MRSA isolates. For the cSSSI regimen (ceftobripole at 500 mg i.v. q12h), the probability of achieving a maximal bactericidal effect (50% *fT* > MIC) was >90% for both MSSA and MRSA isolates.

The results of PTA analysis for gram-negative pathogens are shown in Table 6. Among the gram-negative pathogens used in the analysis, 166 isolates were AmpC beta-lactamase producers (*Citrobacter* spp., *Enterobacter* spp., *Serratia marcescens*), with a MIC_{90} of 16 mg/liter. Among the non-AmpC-producing, non-*Pseudomonas aeruginosa* isolates (*Escherichia coli*, *Klebsiella* spp., *Proteus mirabilis*), there were 206 isolates for which the MIC was ≤0.25 mg/liter, and these encompassed more than 90% of the isolates. We also examined a distribution of 407 isolates of *Pseudomonas aeruginosa*. For this pathogen, the MIC_{50} was 4.0 mg/liter and the MIC_{90} was 32 mg/liter. The overall probability that ceftobiprole at 500 mg i.v. q8h would achieve a nearly bactericidal effect (60% *fT* > MIC) exceeded 90% for non-AmpC-producing gram-negative bacilli. For AmpC-producing gram-negative bacilli, the overall probability that ceftobiprole would achieve a static effect (40% *fT* > MIC) was 89.8%. The probability that ceftobiprole would achieve a static effect against *Pseudomonas aeruginosa* was 71.1%.

Renal dose adjustment analysis.The results of the renal dose adjustment analysis are shown in Fig. 2 and 3. In Fig. 2A, the PTA for a 2-h infusion of 250 mg q8h is displayed for a CL_{CR} of 30 ml/min. The probability of achieving 60% *fT* > MIC exceeded 90% for an organism with an MIC of ≤2 mg/liter. Figure 2B shows the distribution of exposure ratios (AUC_{24SS} fixed at a CL_{CR} of 30 ml/min relative to the AUC_{24SS} distribution fixed at a CL_{CR} of 100 ml/min). Waiting to adjust the dose until the CL_{CR} attained 30 ml/min resulted in some subjects experiencing an exposure ratio greater than 3, although the majority of subjects had an exposure ratio of less than 2.

The PTA for a dose and schedule of 500 mg q12h is displayed in Fig. 3A. The probability of achieving 50% *fT* > MIC exceeded 90% for MICs ≤2 mg/liter and was greater than 87% for the nearly maximal killing target of 60% *fT* > MIC at this MIC. The distribution of exposure ratios is displayed in Fig. 3B and demonstrates that all subjects had an exposure ratio of less than a factor of 2.

## DISCUSSION

Ceftobiprole is an innovative anti-MRSA broad-spectrum cephalosporin currently undergoing phase III trials for the treatment of cSSSIs and NP. As ceftobiprole was entering phase III clinical trials, we performed a large population pharmacokinetic analysis (150 subjects) and Monte Carlo simulation to identify the PTA for candidate doses that were under consideration for these trials. For the NP dosing scheme (ceftobiprole at 500 mg i.v. q8h), we examined the influence of altering the infusion time on PTA. At the dose and schedule of 500 mg i.v. q8h, the alteration of the infusion time over the range of 0.5 h to 2.0 h provided a minimal impact on the probability of target attainment when the MIC was ≤1 mg/liter but had a greater impact at MICs ≥2 mg/liter (Table 3). These observed PTA differences, however, are less than those seen previously with prolonged infusion times for carbapenems because the targets are different for these agents (22).

Based on the results of the Monte Carlo simulation, ceftobiprole administered at 500 mg i.v. q8h as a 2-h infusion was selected for the phase III clinical trial of the treatment for NP. Examination of the results of the overall PTA analysis for ceftobiprole at 500 mg i.v. q8h as a 2-h infusion (Fig. 1A) demonstrated that at 2 mg/liter, PTAs exceeded 98% for the stasis target for *Staphylococcus aureus* (30% *fT* > MIC) and for gram-negative bacilli (40% *fT* > MIC). For 50% and 60% *fT* > MIC, the PTA exceeded 93% at an MIC of 2 mg/liter, indicating that nearly maximal killing of MRSA and gram-negative bacilli will be achieved for the vast majority of patients. This is best put into perspective by considering the MIC distribution. For MRSA, the MIC_{90} is 2.0 mg/liter, suggesting that ceftobiprole will provide robust activity and will be suitable for patients who are infected with MRSA. When one examines its performance against gram-negative bacilli (Table 6), it is clear that the nearly maximal effect (60% *fT*>MIC) against such organisms as *Escherichia coli*, *Klebsiella pneumoniae*, and *Proteus mirabilis* will be achieved in excess of 94% of the time. Ceftobiprole demonstrates a PTA of ∼90% for stasis and almost 88% for a nearly maximal effect against such organisms as *Enterobacter cloacae*, *Enterobacter aerogenes*, *Citrobacter* species, and *S. marcescens* that produce a chromosomally encoded AmpC β-lactamase. When these AmpC β-lactamase-producing organisms are present, care should be exercised and clinical results need to be examined closely, particularly in circumstances where the bacterial burden is very high.

Ceftobiprole at 500 mg i.v. q12h as a 1-h infusion was selected for the clinical trial of the treatment of cSSSIs. Similar to the PTA analysis for the treatment of NP and as detailed in Materials and Methods, it should be emphasized that the overall PTAs discussed herein include a curve specifically for subjects with different CL_{CR}s. The impact of different estimated CL_{CR}s on the PTA can clearly be seen in Table 4. Depending on the estimated CL_{CR}, the probability of attaining stasis is shown to go from greater than 80% to in excess of 90%. When one weights the probabilities by the probability of observing the values of the estimated CL_{CR}, the overall probabilities of achieving 30% *fT* > MIC and 50% *fT* > MIC exceeded 90% for MICs ≤2 mg/liter and ≤1 mg/liter, respectively (Fig. 1B). Furthermore, the probability of achieving a nearly maximal effect exceeded 90% for both MSSA and MRSA (Table 5).

While it is critical to have high PTA activity, the ideal antimicrobial agent should be effective while minimizing the probability of concentration-related toxicity. Clearly, as renal clearance is a major component of total clearance, as CL_{CR} declines, the drug concentrations rise. To prevent undue drug exposure, we used the population pharmacokinetic analysis to examine two different dose/schedule adjustment schemes for ceftobiprole.

When the dose and schedule were unaltered down to a GFR of 30 ml/min, a dose reduction to 250 mg i.v. q8h (2-h infusion) at this CL_{CR} clearly provides PTAs that are excellent (Fig. 2A). The price for this is that while the median exposure ratio (relative to the distribution for patients with a GFR of 100 ml/min) was 1.68, the 90th, 95th, and 99th percentiles of the exposure ratios were 2.34, 2.53, and 2.84, respectively, with some subjects in this 9,999-subject simulation having an exposure ratio that exceeded 3 (Fig. 2B). While an explicit concentration-toxicity relationship currently does not exist, we believed that it would be prudent to keep the maximal exposure ratio down to less than 2 because studies with volunteers have demonstrated the tolerability of ceftobiprole at 1 g q8h. To date, higher doses have not been evaluated.

A second dose/schedule alteration scheme was examined, in which the regimen was changed to ceftobiprole at 500 mg i.v. q12h at a CL_{CR} of 50 ml/min. The PTA exceeded 90% for a target of 50% *fT* > MIC out to an MIC of 2 mg/liter and exceeded 87% for the nearly maximal killing target of 60% *fT* > MIC at this MIC, indicating that the change to q12h still maintained the PTAs at this level of renal function (CL_{CR}, 50 ml/min). Of interest, the AUC_{24SS} values did not increase much and the greatest exposure ratio (relative to the distribution seen at a GFR of 100 ml/min) was less than 2 (Fig. 3B). Thus, ceftobiprole at 500 mg q12h with an estimated CL_{CR} of 50 ml/min maintained an excellent PTA for important pathogens while limiting the exposure ratio to less than 2. As data for patients in a phase II trial were included in the analysis, we can have a significant degree of confidence that the doses and schedules detailed here will provide excellent activity against the target pathogens while minimizing the risks due to concentration-dependent toxicities if an appropriate dose/schedule alteration scheme is employed (i.e., switch to 500 mg q12h at an estimated CL_{CR} of 50 ml/min). A limitation that should be recognized is that very few subjects with estimated GFRs less than 20 ml/min were studied. Still remaining to be examined are patients with greater degrees of renal functional impairment as well as patients on hemodialysis/hemofiltration, who would likely be encountered in an intensive care unit with serious infections.

Mouton and colleagues (24) published an earlier work on ceftobiprole dose selection using Monte Carlo simulations that had an outcome somewhat different from ours. They recommended a dose of 500 mg q12h as the standard regimen (24). It is important to understand the genesis of these two different evaluations. In the study of Mouton et al. (24), a small number (*n* = 12) of volunteer subjects was evaluated. More importantly, the patients were closely controlled for their vital statistics, with the result that the standard deviation of drug clearance was 10%. Studies with volunteers are often deemed the most conservative type of evaluation that can be considered for a new drug, because volunteers are generally young and healthy and are likely to have the highest drug clearances with the shortest drug half-lives. However, when one performs a Monte Carlo simulation, the measure of central tendency (high drug clearance, short half-lives) is only part of the story. Because we are explicitly creating a distribution, it is important to understand the measure of dispersion. For drug clearance, Mouton and colleagues (24) found a mean value of 5.111 liters/h and a standard deviation of 0.518 liter/h. The upper bound of the 95% confidence interval for clearance by using these values is 6.126 liters/h, meaning that it is impossible for the Monte Carlo simulation to identify more than 5% of simulated subjects whose values would differ from the mean value by more than 1 liter/h. This leads to a truncation of the ability to test the validity of the dose. Indeed, in their evaluation, there was a 100% target attainment probability at an MIC of 2 mg/liter for 40% *fT* > MIC at a dose and schedule of 500 mg q12h.

In our evaluation, the standard deviations of the intercept term for clearance and the slope term for clearance represented 62% and 84% of the mean, respectively. This is because of the greater number of subjects evaluated (*n* = 150) and the fact that a number of patients from phase II clinical trials were included. Of interest, 22 of these patients were i.v. drug users and their clearances were actually toward the top end of the distribution. This observation of high clearances of antibiotics in infected drug users was previously noted by Rybak et al. (28).

In summary, we have performed an extensive evaluation of the pharmacokinetics of ceftobiprole and have used this information to perform a Monte Carlo simulation to identify the dose and the schedule of drug that will provide high levels of activity against target pathogens. For the first time, we have included previous clinical trial experience (the distribution of estimated CL_{CR} values in patients with cSSTIs and NP [14, 32]) within a population model in which drug clearance is made proportional to the estimated CL_{CR} in order to evaluate the impact of renal functional impairment on the PTA. Also for the first time, we have used this technique to examine the range of exposures resulting from dose/schedule alteration with renal function impairment. We have demonstrated that PTAs would not be adversely affected by alterations that also limit drug exposure, thus providing therapy with a high probability of antimicrobial activity together with a minimal probability of concentration-related toxicities. All of these findings need to be validated in the clinical trial arena.

## ACKNOWLEDGMENTS

This study was supported by a grant from Johnson and Johnson Pharmaceutical Research and Development.

G.L.D. is a consultant to Johnson and Johnson Pharmaceutical Research and Development. R.P. is an employee of Basilea Pharmaceutica. J.B.K., B.P.M., H.C.K., K.B., and G.N. are employees of J&J Pharmaceuticals. No other potential conflicts of interest exist for the other authors.

## FOOTNOTES

- Received 21 September 2006.
- Returned for modification 31 October 2006.
- Accepted 17 March 2007.
↵▿ Published ahead of print on 26 March 2007.

- American Society for Microbiology