Pharmacokinetic-Pharmacodynamic Evaluation of Ertapenem for Patients with Hospital-Acquired or Ventilator-Associated Bacterial Pneumonia

Ertapenem provides activity against many pathogens commonly associated with hospital-acquired and ventilator-associated bacterial pneumoniae (HABP and VABP, respectively), including methicillin-susceptible Staphylococcus aureus and numerous Gram-negative pathogens with one major gap in coverage, Pseudomonas aeruginosa. Pharmacokinetic-pharmacodynamic (PK-PD) target attainment analyses were conducted to evaluate ertapenem against the most prevalent Enterobacteriaceae causing HABP/VABP.

ABSTRACT

Ertapenem provides activity against many pathogens commonly associated with hospital-acquired and ventilator-associated bacterial pneumoniae (HABP and VABP, respectively), including methicillin-susceptible Staphylococcus aureus and numerous Gram-negative pathogens with one major gap in coverage, Pseudomonas aeruginosa. Pharmacokinetic-pharmacodynamic (PK-PD) target attainment analyses were conducted to evaluate ertapenem against the most prevalent Enterobacteriaceae causing HABP/VABP. The objective of these analyses was to provide dose selection support for and demonstrate the appropriateness of ertapenem to empirically treat patients with HABP/VABP when administered with murepavadin, a novel targeted antimicrobial exhibiting a highly specific spectrum of activity against P. aeruginosa. A previously developed population pharmacokinetic model, a total-drug epithelial lining fluid (ELF) to free-drug serum penetration ratio, contemporary in vitro surveillance data for ertapenem against Enterobacteriaceae, and percentage of the dosing interval for which drug concentrations exceed the MIC value (%T>MIC) targets associated with efficacy were used to conduct Monte Carlo simulations for five ertapenem regimens administered over short or prolonged durations of infusion. Overall total-drug ELF percent probabilities of PK-PD target attainment based on a %T>MIC target of 35% among simulated patients with HABP/VABP arising from Enterobacteriaceae based on pathogen prevalence data for nosocomial pneumonia ranged from 89.1 to 92.7% for all five ertapenem regimens evaluated. Total-drug ELF percent probabilities of PK-PD target attainment ranged from 99.8 to 100%, 97.9 to 100%, 10.6 to 74.1%, and 0 to 1.50% at MIC values of 0.06, 0.12, 1, and 4 μg/ml, respectively (MIC₉₀ values for Escherichia coli, Serratia marcescens, Enterobacter species, and Klebsiella pneumoniae, respectively). Results of these analyses provide support for the evaluation of ertapenem in combination with murepavadin for the treatment of patients with HABP/VABP.

INTRODUCTION

Nosocomial pneumonia, defined herein as hospital-acquired and ventilator-associated bacterial pneumonia (HABP and VABP, respectively), has been identified as a leading cause of death attributed to health care-associated infections and a major contributor to increased health care spending in the case of VABP (1–3). Pseudomonas aeruginosa is a predominant causative pathogen of these infections, with an estimated worldwide prevalence of 22.4% among patients with HABP and 26.6% among those with VABP (4). Moreover, the prevalence of multidrug-resistant (MDR) P. aeruginosa isolates among these patients in the United States and worldwide have been reported to be as high as 10 and 30.5%, respectively (5, 6). And while some data indicate stability or a slow decline in the prevalence of MDR P. aeruginosa isolates within the United States (7, 8), the current pervasiveness of these organisms poses a great threat to patients. These observations illustrate not only the need for new antimicrobials to treat patients with HABP and VABP but also, more specifically, the need for antimicrobials which exhibit novel mechanisms of action for combating the threat posed by MDR P. aeruginosa.

Murepavadin (POL7080) is one such novel investigational agent. This peptide antimicrobial is primarily active against P. aeruginosa and has no significant effect against other tested species, including other Pseudomonas species, with the exception of Pseudomonas stutzeri and Pseudomonas mendocina (9). Murepavadin elicits its antibacterial activity by binding to lipopolysaccharide transport protein D (LptD), a common outer membrane protein in Gram-negative bacteria (10). LptD functions in conjunction with LptE to transport lipopolysaccharide from the periplasm to the outer leaflet of the outer cell membrane (11). Murepavadin’s narrow spectrum of activity is suspected to be due in part to its selective binding to P. aeruginosa LptD, which exhibits a peculiarly large periplasmic domain (10, 11). Given this spectrum of activity, murepavadin is being developed by Polyphor, Ltd., for the treatment of patients with HABP and VABP due to P. aeruginosa and is in phase III of clinical development (ClinicalTrials registration numbers NCT03582007 and NCT03409679) (9).

Murepavadin’s very narrow spectrum of activity complicates the design of clinical studies intended to evaluate efficacy. HABP and VABP are serious infections, often presenting as monomicrobial or polymicrobial diseases due to various Gram-positive and/or Gram-negative organisms (12). Therefore, when empirical therapy is provided (i.e., pathogen and MIC are unknown), murepavadin must be coadministered with one or more antimicrobials to cover other common pathogens. However, all recommended therapies for HABP and VABP (excluding those specifically intended for methicillin-resistant Staphylococcus aureus [MRSA]) which could be considered potential agents to coadminister with murepavadin also exhibit activity against P. aeruginosa. Thus, the evaluation of efficacious outcomes due to murepavadin would be confounded if it is paired with such an agent in a clinical trial. Ertapenem, a carbapenem which lacks activity against P. aeruginosa but covers most other common causative pathogens of HABP and VABP, excluding MRSA, represents an ideal agent to evaluate as part of combination therapy with murepavadin for the treatment of patients with HABP/VABP. However, due to its lack of P. aeruginosa coverage, ertapenem has not been studied for the treatment of HABP/VABP.

Herein, we describe pharmacokinetic-pharmacodynamic (PK-PD) target attainment analyses that were undertaken to evaluate ertapenem regimens against the most prevalent Enterobacteriaceae associated with HABP and VABP. Given that ertapenem was not developed for the treatment of patients with HABP or VABP, the dosing of this agent may not be optimized for this use. Therefore, the objective of these analyses was to provide dose selection support for and evaluate the appropriateness of ertapenem to empirically treat patients with HABP and VABP arising from common Enterobacteriaceae when the drug is administered with murepavadin.

RESULTS

Pharmacokinetic analyses.

In order to evaluate the relationship between ertapenem clearance and normalized creatinine clearance (CLcrn), ertapenem PK data in patients with various degrees of renal function, described by Mistry et al., were assessed (13). Ertapenem clearance values were plotted against the mean CLcrn for each renal group, and a linear function was fit to the data, as displayed in Fig. S2 in the supplemental material. Given that patients with nosocomial pneumonia frequently have renal impairment, the fitted model parameter estimates for the relationship between clearance and CLcrn were incorporated into the ertapenem population PK model. The final equation describing ertapenem clearance for a typical subject is presented in equation 1:

$$\text{Clearance}=[1.71+0.01143\times (\text{CLcrn}-110)]\times {(\text{weight}÷95.9)}^{0.278}$$ (1)

To confirm that the ertapenem population PK model, which was developed using serum concentrations, could accurately predict plasma concentrations, an external validation was performed. As demonstrated by Fig. S3, the observed total-drug plasma concentrations agreed well with the simulated total-drug serum concentration-time profile at all evaluated time points, thus indicating that the model could accurately simulate expected plasma concentrations in addition to serum concentrations.

The ertapenem total-drug ELF/free­drug serum penetration ratio was computed using data described by Boselli et al. (14). The area under the concentration-time curve (AUC) values for 24-h total-drug ELF and median 24-h free-drug serum using pooled data across all patients (n = 15) were 68.2 and 200 mg·h/liter, respectively. The computed ELF penetration ratio was 34%.

Human protein binding analyses.

Figure S1 shows the exponential function, as described by equation 1, for the relationship between the mean percentage of the fraction of unbound drug and total-drug ertapenem plasma concentration derived from the human ertapenem protein binding data of Majumdar et al. (15). The estimated coefficient for the percentage of the fraction unbound and the exponential coefficient were 4.28% and 0.00446 liters/mg, respectively. Both were estimated with excellent precision, as indicated by the low standard errors of the means (%SEM = 1.41 and 1.67, respectively). The r² for the relationship was >0.99, indicating that the model fit the data well and explained 99% of the variability in the free fraction. The free fraction was estimated in the model with equation 2:

$$\text{Free fraction}(\%)=4.28\hspace{0.17em}(e^{0\text{.00446}\times \text{total-drug plasma concentration}})$$ (2)

PK-PD target attainment analyses based on total-drug ELF exposures.

Table 1 shows the percent probabilities of PK-PD target attainment by MIC and overall on day 1 based on the percentage of the dosing interval that total-drug ELF concentrations exceed the MIC (total-drug ELF %T>MIC) target of 35% among simulated patients administered ertapenem 1- and 2-g regimens. Percent probabilities of PK-PD target attainment at the MIC₉₀ values for Klebsiella pneumoniae, Escherichia coli, Enterobacter species, and Serratia marcescens (>2, 0.06, 1, and 0.12 μg/ml, respectively) ranged from 0 to 1.50%, 99.8 to 100%, 10.6 to 74.1%, and 97.9 to 100%, respectively, among simulated patients administered ertapenem 1- and 2-g regimens.

TABLE 1.

Percent probabilities of PK-PD target attainment by MIC and overall on day 1 based on the total-drug ELF %T>MIC target of 35% among simulated patients after administration of 1-g and 2-g ertapenem regimens

^aBased on the MIC distributions described in Table 3, 0.06, 0.12, 1, and >2 μg/ml were the MIC₉₀ values for E. coli, S. marcescens, Enterobacter spp., and K. pneumoniae, respectively.

^bShaded cells indicate percent probabilities of PK-PD target attainment of ≥90%. Values for individual genera represent the overall percent probability of PK-PD target attainment weighted over the ertapenem MIC distribution for each pathogen. Values for Enterobacteriaceae represent the overall percent probability of PK-PD target attainment for simulated patients with nosocomial pneumonia arising from Enterobacteriaceae based on the prevalence data for the four specified Enterobacteriaceae among patients with nosocomial pneumonia described in Table 4 and the associated ertapenem MIC distributions for these pathogens described in Table 3.

When percent probabilities of PK-PD target attainment were based on the distributions of MIC values for each pathogen, overall percent probabilities of PK-PD target attainment for K. pneumoniae, E. coli, Enterobacter spp., S. marcescens, and Enterobacteriaceae ranged from 83.7 to 86.5%, 98.5 to 99.3%, 84.8 to 93.2%, 96.8 to 98.5%, and 89.1 to 92.7%, respectively. The data provided in Table 1 overlaid upon the MIC distributions for K. pneumoniae, E. coli, Enterobacter spp., and S. marcescens isolates are shown in Fig. 1.

FIG 1.

Percent probabilities of PK-PD target attainment by MIC value on day 1 based on the total-drug ELF %T>MIC target of 35% among simulated patients after administration of ertapenem 1- and 2-g regimens, overlaid on the K. pneumoniae, E. coli, Enterobacter spp., and S. marcescens MIC distributions.

Results based on the %T>MIC targets of 40 and 45% are presented in Tables S1 and S2, respectively. For the majority of assessments, the MIC values at which percent probabilities of PK-PD target attainment of ≥90% were achieved were the same as those for the above-described assessments based on a %T>MIC target of 35%.

Overall percent probabilities of PK-PD target attainment by total-drug ELF %T>MIC target based on K. pneumoniae, E. coli, Enterobacter spp., and S. marcescens MIC distributions for non-carbapenem-resistant Enterobacteriaceae (non-CRE) isolates are shown in Table S3. Across all scenarios evaluated, overall percent probabilities of PK-PD target attainment improved 0.3 to 10.3% when identical scenarios were compared based on MIC distributions for the non-CRE subsets versus those for all isolates. Improvements in overall percent probabilities of PK-PD target attainment were greatest for K. pneumoniae. This finding was expected, given that the MIC distribution for this pathogen was most influenced by the exclusion of CRE isolates.

PK-PD target attainment analyses based on free-drug plasma exposures.

Table 2 shows the percent probabilities of PK-PD target attainment by MIC value and overall on day 1 based on the free-drug plasma %T>MIC target of 35% among simulated patients administered the ertapenem 1- and 2-g regimens. Percent probabilities of PK-PD target attainment at the MIC₉₀ values for K. pneumoniae, Enterobacter spp., and S. marcescens of >2, 1, and 0.12 μg/ml, respectively, ranged from 1.65 to 55.4%, 77.4 to 99.6%, and 99.8% to 100%, respectively, among simulated patients administered ertapenem 1- and 2-g regimens. For E. coli, percent probabilities of PK-PD target attainment at the MIC₉₀ value of 0.06 μg/ml were 100% for all regimens.

TABLE 2.

Percent probabilities of PK-PD target attainment by MIC and overall on day 1 based on the free-drug plasma %T>MIC target of 35% among simulated patients after administration of 1-g and 2-g ertapenem regimens

^aBased on the MIC distributions described in Table 3, 0.06, 0.12, 1, and >2 μg/ml were the MIC₉₀ values for E. coli, S. marcescens, Enterobacter spp., and K. pneumoniae, respectively.

^bShaded cells indicate percent probabilities of PK-PD target attainment of ≥90%. Values for individual genera represent the overall percent probability of PK-PD target attainment weighted over the ertapenem MIC distribution for each pathogen. Values for Enterobacteriaceae represent the overall percent probability of PK-PD target attainment for simulated patients with nosocomial pneumonia arising from Enterobacteriaceae based on the prevalence data for the four specified Enterobacteriaceae among patients with nosocomial pneumonia described in Table 4 and the associated ertapenem MIC distributions for these pathogens described in Table 3.

When percent probabilities of PK-PD target attainment were based on the distributions of MIC values for each pathogen, overall percent probabilities of PK-PD target attainment for K. pneumoniae, E. coli, Enterobacter spp., S. marcescens, and Enterobacteriaceae ranged from 86.7 to 94.3%, 99.3 to 99.8%, 93.3 to 98.0%, 98.4 to 99.4%, and 92.8 to 97.2%, respectively. The data provided in Table 2 overlaid upon the MIC distributions for K. pneumoniae, E. coli, Enterobacter spp., and S. marcescens isolates are shown in Fig. 2.

FIG 2.

Percent probabilities of PK-PD target attainment by MIC value on day 1 based on the free-drug plasma %T>MIC target of 35% among simulated patients after administration of ertapenem 1- and 2-g regimens, overlaid on the K. pneumoniae, E. coli, Enterobacter spp., and S. marcescens MIC distributions.

Results based on the %T>MIC targets of 40 and 45% are presented in Tables S4 and S5, respectively. For the majority of assessments, the MIC values at which percent probabilities of PK-PD target attainment of ≥90% were achieved were the same as those for the above-described assessments based on a %T>MIC target of 35%.

Overall percent probabilities of PK-PD target attainment by the free-drug plasma %T>MIC target based on K. pneumoniae, E. coli, Enterobacter spp., and S. marcescens MIC distributions for non-CRE isolates are shown in Table S6. Across all scenarios evaluated, overall percent probabilities of PK-PD target attainment increased by 0.1 to 10.2% when identical scenarios were compared based on MIC distributions for the non-CRE subsets versus those for all isolates. Improvements in overall percent probabilities of PK-PD target attainment were greatest for K. pneumoniae. This finding was expected, given that the MIC distribution for this pathogen was most influenced by the exclusion of CRE isolates.

DISCUSSION

The objective of these analyses was to evaluate ertapenem regimens to empirically treat patients with HABP and VABP arising from common Enterobacteriaceae. These analyses were carried out to assess the appropriateness of ertapenem as part of a combination therapy with murepavadin for future clinical trials in patients with HABP/VABP. As described herein, PK-PD target attainment analyses were conducted to evaluate regimens administering ertapenem as 1 and 2 g intravenously (i.v.) every 12 or 24 h (q12h or q24h, respectively) and over short or prolonged durations of infusion (0.5 and 3 h, respectively) for the treatment of patients with HABP/VABP. These analyses were carried out using a previously developed population PK model (16), nonclinical PK-PD targets for efficacy (17, 18), in vitro surveillance data, and Monte Carlo simulation.

Among the %T>MIC targets evaluated, emphasis was placed on PK-PD target attainment results based on a 1-log₁₀ CFU reduction endpoint for which the %T>MIC target was 35%. This bacterial reduction endpoint is considered an appropriate threshold upon which to base dose selection decisions for patients with HABP or VABP (19, 20). Results based on free-drug plasma exposures demonstrated percent probabilities of PK-PD target attainment of ≥90% at MIC values that were one to two dilutions higher than those based on total-drug ELF exposures. Given the estimated ertapenem ELF penetration ratio of 34% described herein, this finding was not surprising. However, as described below, emphasis was placed on the PK-PD target attainment results based on total-drug ELF exposures, given that ELF represents the effect site for patients with HABP/VABP.

Percent probabilities of PK-PD target attainment based on the total-drug ELF %T>MIC target of 35% were ≥97.9% at the MIC₉₀ values for both E. coli and S. marcescens (0.06 and 0.12 μg/ml, respectively) for all five ertapenem regimens. At the MIC₉₀ value for Enterobacter spp. (1 μg/ml), percent probabilities of PK-PD target attainment were substantially more favorable following administration of the 2-g/day regimens than the 1-g/day regimens (64.3 to 74.1% versus 10.6 to 12.6%). The results of these analyses were poor at the MIC₉₀ value for K. pneumoniae (≥2 μg/ml) regardless of the regimen evaluated (0 to 1.50%).

Overall percent probabilities of PK-PD target attainment based on the total-drug ELF %T>MIC target of 35% were high for all four Enterobacteriaceae regardless of the ertapenem regimen evaluated (≥83.7%). However, the most favorable results were achieved for simulated patients after administration of ertapenem regimens of 1 g q12h over 30 min and 2 g q24h over 3 h (86.4 to 99.3% and 86.5 to 99.3%, respectively), followed by administration of 2 g q24h over 30 min (86.1 to 99.2%).

When considered in the context of the prevalence data for the above-described four Enterobacteriaceae among patients with nosocomial pneumonia and the associated ertapenem MIC distributions for these pathogens, overall percent probabilities of PK-PD target attainment were ≥89.1% regardless of the regimen evaluated. These results are important, given that they account for the likelihood of encountering a given pathogen and its probable MIC value, thus making these results more reflective of a clinical scenario in which therapy is empirically selected (i.e., pathogen and MIC are unknown). Again, although high percent probabilities of PK-PD target attainment were achieved across all regimens, results were most favorable following administration of the 2-g/day regimens. Percent probabilities of PK-PD target attainment were 92.7% for both the 1 g q12h over 30 min and the 2 g q24h over 3 h regimens and 92.2% for the 2 g q24h over 30 min regimen.

The percent probabilities of PK-PD target attainment achieved for K. pneumoniae at clinically relevant MIC values were poor relative to those for the other three Enterobacteriaceae evaluated. This observation is due in large part to the fact that the K. pneumoniae susceptibility distribution utilized was bimodal with a second peak at the MIC dilution of >2 μg/ml (i.e., the MIC₉₀), which accounts for 12.4% (n = 373) of all isolates observed. Among these isolates, 82.6% were also resistant to meropenem (data not shown), based on the Clinical and Laboratory Standards Institute (CLSI) breakpoint of ≥4 μg/ml (21).

It is also important to consider ertapenem’s activity relative to that of standard-of-care regimens. Meropenem is considered to be a standard-of-care agent for HABP/VABP when empirical coverage for MRSA is already provided or not indicated (12). In vitro surveillance data described herein, which demonstrated similar percentages of Enterobacteriaceae susceptible to ertapenem and meropenem (94.9 and 96.9%, respectively) (data not shown) based on CLSI breakpoints (21), suggest that ertapenem plus murepavadin and meropenem have similar spectrums of activity. These findings are important, given that the intent of these analyses was to determine an appropriate ertapenem regimen to be used in combination with murepavadin, an investigational highly specific outer membrane protein-targeting antibiotic, for the treatment of patients with HABP/VABP in future clinical studies in which anti-pseudomonal-beta-lactams such as meropenem will be evaluated as comparator agents. Of note, if MRSA were suspected in these patients, an additional agent would be administered to provide MRSA coverage.

The above-described PK-PD target attainment results were also interpreted in the context of an MIC distribution that one would expect to encounter in a more controlled setting, such as a clinical trial, a setting in which patients with HABP/VABP due to CRE are excluded from study. In such an instance, the MIC values for K. pneumoniae, E. coli, Enterobacter spp., and S. marcescens would be expected to be lower, as evidenced by the MIC₉₀ values for the non-CRE K. pneumoniae and Enterobacter spp. subsets described in Table S7 in the supplemental material (0.25 μg/ml and 0.5 μg/ml, respectively). Moreover, if non-CRE isolates can be assumed, overall percent probabilities of PK-PD target attainment would be up to 10.2% higher across all comparisons of identical scenarios based on the MIC distributions for non-CRE isolates relative to those for all Enterobacteriaceae (Table S3).

An important limitation of these analyses was the lack of optimal preclinical PK-PD data for ertapenem and other carbapenems. As a result, %T>MIC targets for carbapenems were derived from neutropenic murine thigh infection model studies instead of neutropenic murine lung infection model studies. Although previous studies have demonstrated concordance between carbapenem free-drug plasma %T>MIC targets for a given endpoint derived from murine thigh and lung infection models (22–25), murine ELF PK data for carbapenems were not available, thus precluding the opportunity to estimate murine ELF penetration. Consequently, by assuming that carbapenem PK-PD targets for efficacy derived from murine thigh and lung infection models were similar, we assumed murine ELF penetration to be 100%. However, total-drug ELF %T>MIC targets for carbapenems may in fact be lower if the point estimate for murine ELF penetration is similar to that based on data from humans for ertapenem (34%) or doripenem (23 to 31%) (26). Therefore, given the data available, the assessment of the above-described total-drug ELF %T>MIC targets to support dose selection was considered to be a conservative approach.

Another limitation of these analyses was the use of a population PK model which included the use of an ad hoc relationship to characterize the relationship between ertapenem clearance and CLcrn. To date, only three ertapenem population PK models which include formally selected covariates have been published (16, 27, 28). Among these publications, one does not report final population PK model parameter estimates (27), and another describes the development of a model which was based on a limited set of burn patients (n = 8) (28), the PK for whom are substantially different from those of nonburn patients (29). Consequently, the population PK model developed by Lakota et al. (16) was selected for the analyses described herein, and the aforementioned relationship between ertapenem clearance and CLcrn described by Mistry et al. was added to this model, given the importance of CLCrn to explain ertapenem clearance (13).

These analyses included a number of strengths not presented in previous evaluations of ertapenem PK-PD target attainment analyses. Aside from the work described herein, to our knowledge, only three other groups previously utilized nonlinear functions to characterize ertapenem protein binding when conducting ertapenem PK-PD target attainment analyses (30–32). Additional strengths of these analyses were the utilization of a population PK model which included formally selected covariates and the evaluation of effect-site exposures for patients with HABP/VABP (i.e., ELF AUC values), two considerations which appear to be absent in the available literature describing ertapenem PK-PD target attainment analyses (30–37). The former consideration allowed for the evaluation of simulated exposures which accounted for the impact of body size and renal function on ertapenem PK. The latter consideration, which was based on noncompartmental analyses performed using data from Boselli et al. to obtain the total-drug ELF to free-drug plasma AUC penetration ratio (14), allowed for the generation of total-drug ELF concentration-time profiles. However, despite the measures taken to account for effect-site exposures and important covariate relationships, refinement of the population PK model using plasma and ELF concentrations collected from infected patients will be important to obtain a robust model and confirm inferences based on the simulated data described herein.

In conclusion, results of these analyses demonstrated high overall percent probabilities of total-drug ELF and free-drug plasma PK-PD target attainment for K. pneumoniae, E. coli, Enterobacter spp., and S. marcescens among simulated patients administered ertapenem 1- and 2-g regimens. Such data provide support for the evaluation of ertapenem in combination with murepavadin, an investigational agent with in vitro activity against P. aeruginosa, for the treatment of patients with HABP/VABP in future clinical trials.

MATERIALS AND METHODS

A previously developed ertapenem population PK model, nonclinical PK-PD targets for efficacy, in vitro surveillance data, and Monte Carlo simulation were utilized to evaluate percent probabilities of PK-PD target attainment by MIC and overall. The results of these analyses were interpreted in the context of K. pneumoniae, E. coli, Enterobacter spp., and S. marcescens MIC distributions for ertapenem.

Population pharmacokinetic model.

Development of the population PK model utilized to characterize the disposition of ertapenem is described extensively elsewhere (16). In brief, the model was developed using serum data from normal-weight, obese, and morbidly obese healthy volunteers. Ertapenem serum concentrations were best described using a linear three-compartment model with total body weight as a covariate on clearance and body surface area (BSA) as a covariate on central volume.

Although ertapenem has been shown to be primarily eliminated through renal excretion (13, 27, 38), the population PK model described above did not contain a relationship between clearance and CLcrn. This is likely due to the lack of subjects with renal impairment in the analysis data set utilized to develop this model, as reflected by the minimum CLcrn value of 83 ml/min/1.73 m² among the subjects evaluated. Therefore, a relationship between ertapenem clearance and CLcrn was developed using data from Mistry et al. (13). Based on this data set, geometric mean ertapenem clearance values were available for subjects without renal impairment and with mild, moderate, and severe impairment. Ertapenem clearance values were plotted against the mean CLcrn value for each group, and a linear function was fit to the data. The derived relationship was incorporated into the ertapenem population PK model.

Additionally, in order to account for the increased variability in ertapenem PK expected in critically ill patients, the between-subject variability on the model parameter estimates clearance, central volume, and volume of peripheral compartment 1, which were 9.57, 7.92, and 7.92%, respectively, were increased to 30% (16, 27).

Nonclinical PK-PD targets for efficacy.

For carbapenems, the PK-PD index most predictive of efficacy has previously been shown to be the %T>MIC (39). Nonclinical PK-PD targets for efficacy that were used for evaluation were based on studies carried out for ertapenem and another carbapenem, doripenem, against Gram-negative pathogens, including Enterobacteriaceae isolates, using a neutropenic murine thigh infection model (17, 18). These data demonstrated that free-drug plasma %T>MIC targets associated with 1- and 2-log₁₀ CFU reductions from baseline were approximately 35 and 45%, respectively. Therefore, free-drug plasma %T>MIC targets of 35, 40, and 45% were evaluated.

Ertapenem in vitro activity.

Evaluation of the prevalence of bacterial pathogens among patients with nosocomial pneumonia worldwide was used to determine the Enterobacteriaceae of interest (4). These data demonstrated that K. pneumoniae, E. coli, Enterobacter spp., and S. marcescens were the four most prevalent Enterobacteriaceae within this subset of patients.

Table 3 shows the ertapenem MIC distributions for K. pneumoniae, E. coli, Enterobacter spp., and S. marcescens based on isolates collected worldwide that were used to interpret the PK-PD target attainment results and calculate overall percent probabilities of PK-PD target attainment. These isolates were part of a collection of 13,726 Enterobacteriaceae isolates collected worldwide (93 medical centers located in 33 countries and 9 U.S. census divisions) by JMI Laboratories, Inc., from the 2017 SENTRY Antimicrobial Surveillance Program (data on file, Polyphor, Ltd.). As described previously (40), the SENTRY Antimicrobial Surveillance Program collects and processes bacterial isolates causing a variety of infection types in a large number of medical centers worldwide. Isolates are consecutively collected and centrally processed for viability, purity, and bacterial identification and susceptibility tested by the reference broth microdilution method against numerous antimicrobial agents.

TABLE 3.

Ertapenem MIC distributions for K. pneumoniae, E. coli, Enterobacter spp., and S. marcescens based on isolates collected worldwide

^aAll isolates were collected by JMI Laboratories during 2017.

^bShaded cells represent data for the MIC values up to and including the MIC₉₀ values.

For the PK-PD target attainment analyses described herein, MIC data for a total of 3,009, 6,043, 1,535, and 620 K. pneumoniae, E. coli, Enterobacter spp., and S. marcescens isolates tested, respectively, were evaluated. The MIC₉₀ values among the K. pneumoniae, E. coli, Enterobacter spp., and S. marcescens isolates were >2, 0.06, 1, and 0.12 μg/ml, respectively. These pathogens were the four most prevalent among the Enterobacteriaceae isolates collected from patients hospitalized due to nosocomial pneumonia regardless of stratification by patients with hospital-acquired or ventilator-associated bacterial pneumonia (data not shown). Using a subset of the above-described data from patients hospitalized due to nosocomial pneumonia, the relative prevalences of K. pneumoniae, E. coli, Enterobacter spp., and S. marcescens among patients with Enterobacteriaceae isolates were determined and are shown in Table 4.

TABLE 4.

Prevalence of Enterobacteriaceae among the subset of isolates collected from hospitalized patients with nosocomial pneumonia worldwide

Pathogen	No. of isolatesa	% of total no. of isolates (n = 1,272)	% of selected Enterobacteriaceae isolates (n = 994)b
K. pneumoniae	381	30.0	38.3
E. coli	228	17.9	22.9
Enterobacter spp.	251	19.7	25.3
S. marcescens	134	10.5	13.5
Other Enterobacteriaceae	278	21.9

^aAll isolates were collected by JMI Laboratories during 2017.

^bDefined as K. pneumoniae, E. coli, Enterobacter spp., and S. marcescens.

Among the Enterobacteriaceae presented in Table 3, a total of the 2,686, 6,022, 1,493, and 611 K. pneumoniae, E. coli, Enterobacter spp., and S. marcescens isolates, respectively, were determined to be non-CRE. An isolate was defined as CRE if it displayed an imipenem, doripenem, and/or meropenem MIC value of >2 μg/ml based on interpretive criteria for in vitro testing described by the Clinical and Laboratory Standards Institute (21). The ertapenem MIC distributions for the above-described non-CRE K. pneumoniae, E. coli, Enterobacter spp., and S. marcescens isolates are shown in Table S7 in the supplemental material. The relative prevalences of these non-CRE isolates among patients hospitalized due to nosocomial pneumonia are shown in Table S8.

Ertapenem human plasma protein binding.

As reported by Majumdar et al. (15), the plasma protein binding of ertapenem in humans is nonlinear. Therefore, unweighted, nonlinear, least-squares regression was performed in R statistical software, version 3.3.1 (41), to fit a function to the digitized ertapenem protein binding data set of Majumdar et al., which included ertapenem plasma concentrations that ranged from 0.6 to 283 μg/ml (15).

Generation of ertapenem exposures in simulated patients.

A simulated patient population was generated using R, version 3.3.1 (41), the package mrgsolve (42), and demographic data from patients with nosocomial pneumonia (43). A bootstrap technique was used to randomly assign combinations of weight, BSA, and CLcrn for each simulated patient from a demographic data set collected from 101 patients with nosocomial pneumonia (43). The original data set of 123 patients with nosocomial pneumonia was restricted to a subset of 101 patients with CLcrn values of ≤125 ml/min/1.73 m² to prevent extrapolating the relationship between ertapenem clearance and CLcrn beyond the point at which it would likely continue to be characterized using a simple linear relationship. Using the above-described population PK model, total-drug serum concentration-time profiles from 0 to 24 h were generated for each of these simulated patients after administration of five i.v. ertapenem regimens: 1 and 2 g q24h over 0.5 and 3 h and 1 g q12h over 0.5 h.

Given that the preclinical %T>MIC targets and human protein binding relationship described herein were derived based on ertapenem plasma concentrations, simulation of plasma concentrations (as opposed to serum concentrations) was desired. However, given that the above-described population PK model was developed based on total-drug serum concentrations, it was important to confirm that this model could reliably predict ertapenem plasma concentrations in addition to serum concentrations. Therefore, an external validation was performed using average ertapenem plasma data (38). Average observed ertapenem total-drug plasma concentrations over 24 h following a single ertapenem 1-g i.v. dose infused over 30 min were overlaid on the median simulated total-drug serum concentration-time profile following administration of the same regimen to 2,000 subjects, each weighing 70 kg with a BSA of 1.73 m². Concordance between simulated serum concentrations and observed plasma concentrations was evaluated.

Simulated total-drug serum concentrations from 0 to 24 h were converted to free-drug plasma concentrations through the application of the nonlinear relationship describing ertapenem protein binding in human plasma. Using the resulting free-drug plasma concentration-time profiles and the computed total-drug ELF/free­drug serum penetration ratio (methods described below), total-drug ELF concentration-time profiles were generated for each simulated patient.

The total-drug ELF/free­drug serum penetration ratio was computed using data described by Boselli et al. (14). Free-drug serum ertapenem concentrations were available from 15 VABP patients at 1, 12, and 24 h following the start of a 1-h infusion of 1 g of ertapenem. A 24-h free-drug ertapenem serum AUC was computed for each patient using noncompartmental techniques. Given that only a single total-drug ELF PK concentration was available for each patient, the median total-drug ELF concentrations at 1, 12, and 24 h following the start of the infusion were utilized to compute 24-h total-drug ertapenem ELF AUC. The 24-h total-drug ELF AUC was divided by the median 24-h free-drug serum AUC to compute the total-drug ELF/free­drug serum penetration ratio.

PK-PD target attainment analyses.

Using the above-described total-drug ELF and free-drug plasma ertapenem concentration-time profiles, %T>MIC was determined for each simulated patient on day 1. Ertapenem %T>MIC values were calculated for each simulated patient by counting the total number of total-drug ELF or free-drug plasma concentrations that were above a given MIC value, multiplying this number by the time interval between simulated concentrations (0.2 h), and then dividing this product by 24 h. Total-drug ELF and free-drug plasma %T>MIC was determined for fixed MIC values ranging from 0.008 to 4 μg/ml. Resulting total-drug ELF and free-drug plasma %T>MIC values were assessed to determine the percent probability of attaining the total-drug ELF and free-drug plasma %T>MIC targets previously described by MIC value.

Overall percent probabilities of PK-PD target attainment were determined for ertapenem against K. pneumoniae, E. coli, Enterobacter spp., and S. marcescens using the applicable MIC distribution for each pathogen. For a given %T>MIC target and MIC distribution, this was determined as the sum of the products of the percent probability of PK-PD target attainment at a given MIC value and the proportion of isolates in the collection with a corresponding MIC value. The MIC distributions for K. pneumoniae, E. coli, Enterobacter spp., and S. marcescens that were utilized are presented in Table 3.

Percent probabilities of PK-PD target attainment for simulated patients with HABP/VABP arising from Enterobacteriaceae were calculated for each %T>MIC target and ertapenem regimen using the prevalence data for four Enterobacteriaceae among patients with nosocomial pneumonia described in Table 4 and the associated ertapenem MIC distributions for these pathogens described in Table 3.

The above-described approach for calculating overall percent probabilities of PK­PD target attainment was carried out using the in vitro surveillance data for all Enterobacteriaceae and the subset of non-CRE isolates.