Population Pharmacokinetic Analysis of Cefiderocol, a Parenteral Siderophore Cephalosporin, in Healthy Subjects, Subjects with Various Degrees of Renal Function, and Patients with Complicated Urinary Tract Infection or Acute Uncomplicated Pyelonephritis

ABSTRACT Cefiderocol, a novel parenteral siderophore cephalosporin, exhibits potent efficacy against most Gram-negative bacteria, including carbapenem-resistant strains. The aim of this study was to perform a population pharmacokinetic (PK) analysis based on plasma cefiderocol concentrations in healthy subjects, subjects with various degrees of renal function, and patients with complicated urinary tract infection (cUTI) or acute uncomplicated pyelonephritis (AUP) caused by Gram-negative pathogens and to calculate the fraction of the time during the dosing interval where the free drug concentration in plasma exceeds the MIC (fTMIC). Population PK models were developed with three renal function markers, body surface area-adjusted estimated glomerular filtration rate (eGFR), absolute eGFR, and creatinine clearance, on the basis of 2,571 plasma concentrations from 91 subjects without infection and 238 patients with infection. The population PK models with each renal function marker adequately described the plasma cefiderocol concentrations. Clear relationships of total clearance (CL) to all renal function markers were observed. Body weight and disease status (with or without infection) were also significant covariates. The CL in patients with infection was 26% higher than that in subjects without infection. The fTMIC values were more than 75% in all patients (and were 100% in most patients), suggesting that a sufficient exposure to cefiderocol was provided by the tested dose regimens (2 g every 8 h as the standard dose regimen) for the treatment of cUTI or AUP caused by Gram-negative pathogens.

C efiderocol (product code S-649266) is a new injectable cephalosporin with a catechol group on the side chain at the C-3 position of the cephalosporin core that was initially identified by Shionogi & Co., Ltd., Osaka, Japan, and that exerts its antibacterial activity by inhibiting the synthesis of bacterial cell walls. Cefiderocol exhibits potent efficacy in vitro and in vivo against most Gram-negative bacteria, including carbapenem-resistant strains of Enterobacteriaceae, Pseudomonas aeruginosa, and Acinetobacter baumannii (1)(2)(3)(4). Cefiderocol is being developed for the treatment of carbapenem-resistant Gram-negative bacterial infections, including nosocomial pneumonia, bloodstream infections, and complicated urinary tract infection (cUTI).
For cefiderocol, which exhibits bactericidal activity dependent on the duration of action, the pharmacokinetic (PK)/pharmacodynamic (PD) index most closely correlated with efficacy is the fraction of the time during the dosing interval where the free drug concentration in plasma exceeds the MIC (fT MIC ) (5-7), as has been described with other cephalosporins (8,9). In vivo animal infection models demonstrated a bacteriostatic effect at an fT MIC of 40% to 70% and a bactericidal effect (Ն1-log reduction) at an fT MIC of 55% to 80% against carbapenem-resistant strains of Enterobacteriaceae, P. aeruginosa, and A. baumannii (5)(6)(7). Cefiderocol PK are linear over the range of 100 to 2,000 mg (10). Cefiderocol is mainly excreted unchanged via the kidneys (as 60% to 70% of the dose in subjects with normal renal function), and thus, the clearance of cefiderocol is dependent on renal function (11). The in vitro plasma protein binding of cefiderocol was 57.8%. The population PK model was previously developed on the basis of the concentration data of cefiderocol in healthy subjects and subjects with various degrees of renal function (12). The developed model well described the plasma concentration data.
The aim of this study was to perform a population PK analysis based on the plasma cefiderocol concentrations in healthy subjects, subjects with various degrees of renal function, and patients with cUTI or acute uncomplicated pyelonephritis (AUP) caused by Gram-negative pathogens in a phase 2 study of cefiderocol for the treatment of cUTI (13) and to calculate the fT MIC . A summary of the study designs is shown in Table S1 in the supplemental material. The population PK models were developed using three renal function markers: (i) the estimated glomerular filtration rate (eGFR), which was calculated by the modification of diet in renal disease (MDRD) equation (14) or an equation reported by Matsuo et al. for Japanese subjects (15) (the body surface area-adjusted eGFR [eGFRadj]); (ii) the eGFR converted by multiplying by the individual's body surface area and dividing by 1.73 m 2 (absolute eGFR [eGFRabs]); and (iii) creatinine clearance (CL CR ), which was calculated by the Cockcroft-Gault equation (16). These three renal function markers were assessed separately in the population PK analysis because they have been used as renal function markers (14,(17)(18)(19) and the selection of renal function markers might affect the prediction of cefiderocol PK. fT MIC was calculated on the basis of simulated steady-state plasma cefiderocol concentrations and the MICs of Gram-negative uropathogens detected in the cUTI study.

RESULTS
A total of 2,571 plasma cefiderocol concentrations obtained from 329 subjects were used for developing the population PK models. Data for samples stored under unstable conditions, samples with anomalous concentrations, or samples with concentrations below the limit of quantification (BLQ) (n ϭ 264 concentrations) (see Materials and Methods) were excluded from the analysis.
The subject characteristics are shown in Table 1. The parameter estimates are provided in Table 2. A 3-compartment model was used as a structural PK model since the population mean parameters were estimated appropriately and were similar to those estimated in the previous population PK analyses (12). The proportional error model was selected, since the relative standard error (RSE; in percent) of the PK parameter estimates obtained using the combination error model were large (4.1% to 1,721%), which suggested that the model was not robust.
The correlations of the renal function markers are presented in Fig. S1. The correlations for eGFRadj and eGFRabs were comparable, while they were slightly lower than the correlation for CL CR . Clear relationships of CL to eGFRadj, eGFRabs, and CL CR were observed using the base model and are shown in Fig. 1 for CL CR and Fig. S2 for eGFRadj and eGFRabs. Each renal function marker was a significant covariate for CL, and each was incorporated using the power model, which provided an objective function value (OBJ) similar to or less than that from the piecewise linear model and which had a smaller number of estimable parameters than the other model. The effects of body weight on CL and the volume of distribution in the central and peripheral compartments (V 1 and V 2 , respectively) and the effect of disease status on V 1 were significant in the final model with eGFRadj. The effects of body weight on V 1 and V 2 and the effect of disease status on CL and V 1 were significant in the final models with eGFRabs and CL CR . The final model with CL CR demonstrated that CL and V 1 in patients with infection were 26% and 36% higher, respectively, than those in subjects without infection. The  The number of estimable parameters was the same among the final models developed with eGFRadj, eGFRabs, and CL CR ; and the typical parameter values, the IIV for each parameter, and the intraindividual variability were comparable among the final models (Table 2). Of the three final models, the OBJ of the model with CL CR was the lowest, and thus, the calculation of post hoc PK parameters and fT MIC was performed by using the final model with CL CR . The goodness-of-fit (GOF) plot for the final model with CL CR is presented in Fig. S3. As shown in Fig. 2, the visual predictive check (VPC) indicated that the median predicted concentration profiles by disease status and renal function group well captured the observed data with a lack of bias. The prediction intervals in subjects without infection were relatively wide compared to the distribution of the observed data. Figure 3 shows box plots for individual post hoc CL values with empirical Bayesian estimation by renal function group (augmented and normal renal function; mild, moderate, and severe renal impairment; and end-stage renal disease [ESRD]). The CL of cefiderocol decreased with decreasing renal function. Figure 4 shows box plots for individual post hoc V 1 values for patients with infection by body weight group (Ͻ55, 55 to Ͻ70, 70 to Ͻ90, or 90 to 138 kg). V 1 was slightly dependent on body weight, which is consistent with the fact that body weight was a significant covariate on V 1 in the population PK analysis. The maximum concentration (C max ) and daily area under the concentration-time curve (AUC) for patients with infection, calculated using individual post hoc PK parameters, are summarized by dose regimen in Table 3. The daily AUC was similar among the dose groups.
A summary of the MIC distribution for each pathogen is shown in Table S3. The calculated fT MIC values based on the simulated steady-state plasma cefiderocol concentrations and the MIC of Gram-negative uropathogens detected in the cUTI study were more than 75% in all patients (and were 100% in most patients).

DISCUSSION
We separately developed three population PK models based on plasma cefiderocol concentration data for subjects with or without infection by using different renal function markers (eGFRadj, eGFRabs, and CL CR ). All models developed with the different renal function markers adequately described the plasma cefiderocol concentration data. These results suggest that any renal function marker could be used to adjust the cefiderocol dose. eGFRadj and eGFRabs were similar, while they were slightly lower than that of CL CR (see Fig. S1 in the supplemental material), which is consistent with well-known findings (14,17). The MDRD equation is recognized as providing estimates of the GFR more accurate than those provided by the Cockcroft-Gault equation (18). On the other hand, there have been reports that the CL CR estimated by the Cockcroft-Gault equation is closer to the measured creatinine clearance than the estimated GFR calculated with other equations, including the MDRD equation, for estimation of an augmented renal function (measured CL CR Ͼ 130 ml/min) in critically ill patients (19). CL CR was the best predictor of cefiderocol PK on the basis of the OBJ by the use of NONMEM software. However, the difference in the predictive performance among the models with each renal function marker would not be clinically significant. Therefore, it was suggested that any of these renal function markers can be used for dose adjustment and simulations based on renal function markers.
Body weight was a statistically significant covariate on CL, V 1 , and V 2 in the final model with eGFRadj and V 1 and V 2 in the final models with eGFRabs and CL CR . Body weight was selected as a covariate on CL for the model with eGFRadj but not the models with eGFRabs and CL CR . This is probably because eGFRabs and CL CR could accommodate the effect of body scale for describing the cefiderocol PK but eGFRadj could not. The post hoc analyses suggested that V 1 was slightly dependent on body weight. However, the individual V 1 values overlapped among the body weight groups, as shown in Fig. 4, and ratios of the median values of V 1 relative to the typical value of V 1 for infected patients (11.1 liters) were close to 1, with the values of the ratios being 0.85 for individuals weighing Ͻ55 kg, 0.87 for individuals weighing 55 to Ͻ70 kg, 0.96 for individuals weighing 70 to Ͻ90 kg, and 1.27 for individuals weighing Ն90 kg, suggesting that the effect of body weight on cefiderocol PK would not be clinically significant. Renal function groups defined by CL CR were as follows: augmented renal function, CL CR Ն 120 ml/min; normal renal function or mild renal impairment, CL CR ϭ 60 to Ͻ120 ml/min; moderate or severe renal impairment or end-stage renal disease (ESRD), CL CR ϭ 5 to Ͻ60 ml/min. Noninfected, subjects without infection; Infected, patients with infection; Time, time after the previous dose; solid lines, observed median; dashed lines, observed 2.5th and 97.5th percentiles; dark gray shaded areas, model-predicted 95% confidence interval of the median; light gray shaded areas, model-predicted 95% confidence intervals of the 2.5th and 97.5th percentiles. was used to estimate individual parameters. Renal function groups defined by CL CR were as follows: augmented renal function: CL CR Ն 120 ml/min; normal renal function, CL CR ϭ 90 to Ͻ120 ml/min; mild renal impairment, CL CR ϭ 60 to Ͻ90 ml/min; moderate renal impairment, CL CR ϭ 30 to Ͻ60 ml/min; severe renal impairment, CL CR ϭ 15 to Ͻ30 ml/min; end-stage renal disease (ESRD), CL CR ϭ 5 to Ͻ15 ml/min. Thick center lines, medians; top and bottom lines of the boxes, first and third quartiles (interquartile range), respectively; whiskers, the most extreme data within 1.5ϫ the interquartile range; circles, outliers beyond 1.5ϫ the interquartile range.
Cefiderocol Population PK in cUTI or AUP Patients Antimicrobial Agents and Chemotherapy The disease status (with or without infection) was a significant covariate on V 1 in the final model with eGFRadj and CL and V 1 in the final models with eGFRabs and CL CR . The final model with CL CR suggested that the values of CL and V 1 in patients with infection were 26% and 36% higher, respectively, than those in subjects without infection. These results were consistent with the report for ceftolozane, a parenteral cephalosporin, in patients with cUTI (in which the values of both clearance and the volume of distribution were 21% higher in subjects with infection than subjects without infection) (20). The IIV for patients with infection was higher than that for subjects without infection, as shown in Fig. 2, which is probably because the plasma concentrations from the patients were limited (3 points per patient) and the IIV could not be calculated adequately.
The fT MIC values were more than 75% in all patients (and were 100% in most patients), suggesting that the level of cefiderocol exposure obtained with the dose regimen used in the cUTI study (Table S2) would be sufficient for the treatment of cUTI and AUP caused by Gram-negative uropathogens (median MIC, 0.06 g/ml; MIC range, 0.004 to 8 g/ml; MIC 90 , 1 g/ml). This sufficient exposure was expected from the results of Monte Carlo simulations, which indicated, using the PK model for healthy subjects, that a dose of 2 g every 8 h (q8h) with a 1-h infusion provided a high probability of attainment of a target of an fT MIC of 75% against organisms with MICs up to 4 g/ml (12). In addition, the mean urine cefiderocol concentrations for 8 patients  in the cUTI study were 2,710 g/ml (range, 953 to 5,520 g/ml) at 2 h after the start of infusion and 1,520 g/ml (range, 336 to 4,220 g/ml) at 6 h after the start of infusion. Urine cefiderocol concentrations were also high relative to the MIC values detected in the cUTI study. As the protein-unbound fraction was not obtained in the phase 2 study of cefiderocol for the treatment of cUTI, individual fT MIC values were calculated on the basis of the free concentrations in plasma using a fixed value for the unbound fraction of 0.422. The effect of the fixed unbound fraction would be minimal for the calculation of fT MIC because the plasma unbound fraction was similar between the various renal function groups (11). As shown in Fig. 3c, the clearance of cefiderocol in the 23 cUTI or AUP patients with augmented renal function (CL CR Ն 120 ml/min) was higher than that in the patients with normal renal function (CL CR ϭ 90 to Ͻ120 ml/min). Creatinine clearance was not measured in this study. Although the use of a measured creatinine clearance may be more appropriate to define augmented renal function, the use of an equation-derived value, such as CL CR estimated by the Cockcroft-Gault equation, would be clinically practical. Monte Carlo simulations suggested that a more frequent dose (every 6 h) had a benefit for subjects with augmented renal function to attain a sufficient fT MIC (12). The target patient population for cefiderocol includes critically ill patients infected with multidrug-resistant strains, which would be less susceptible than the uropathogens collected from the cUTI study. Augmented renal function is often observed in critically ill patients. Therefore, shortening of the cefiderocol dosing interval would be recommended for patients with augmented renal function to obtain enough exposure against organisms for which the MIC is higher.
In summary, our models developed with eGFRadj, eGFRabs, or CL CR described well the PK of cefiderocol. Clear relationships of CL to renal function markers were observed, as expected. It was revealed that the exposure to cefiderocol in patients with infection would be modestly lower than that in subjects without infection. A cefiderocol exposure sufficient for the treatment of cUTI and AUP caused by Gram-negative uropathogens was provided by the tested dose regimens (2 g q8h as the standard dose regimen).

MATERIALS AND METHODS
Data. Plasma cefiderocol concentration data from two phase 1 studies (10-12) and one phase 2 cUTI study (13) were used for the modeling (see Table S1 in the supplemental material). The population PK models were previously developed on the basis of plasma and urine cefiderocol concentration data for 54 healthy subjects and plasma concentration data for 37 subjects with various degrees of renal function (12).
In this study, the plasma concentration data for cUTI and AUP patients in the cUTI study (13) were included to develop the population PK models. The cUTI study was a multinational, double-blind, randomized study to assess the efficacy and safety of cefiderocol in hospitalized adults with cUTI with or without pyelonephritis or AUP caused by Gram-negative pathogens in comparison with intravenous imipenem-cilastatin (IMP-CS). The patients received 2 g as a 1-h intravenous infusion three times daily at 8-h intervals (q8h) for 7 or 14 days. The dose of cefiderocol was reduced on the basis of renal function and body weight, as shown in Table S2, consistent with the dosing instructions for IMP-CS, in order to maintain the blinding to the 2 treatments.
Blood samples for PK testing were not collected from 7 patients mainly due to patient withdrawal from the study. Blood samples for PK testing from 1 patient were not analyzed because the conditions used to store the sample did not meet the criteria required to maintain stability. Three concentrations from 1 patient were not used for analysis because the sampling times were unidentified. A total of 264 samples obtained after cefiderocol administration, which were mainly from healthy subjects, had concentrations that were BLQ and were excluded from the analyses. Plasma concentrations from 1 patient were entirely excluded from the analysis because they were all BLQ. A total of 156 blood samples from 52 patients were delayed in their delivery to the laboratory where the buffer was added to the samples for stabilization and had been stored at Ϫ20°C for more than 7 days after the samples were drawn but prior to the addition of the buffer; therefore, they were excluded from the analyses since cefiderocol is not thought to be stable in blood samples on the basis of stability data. Eight plasma concentrations obtained in the renal impairment study and the cUTI study were excluded from the analysis because they were considered anomalous, as the concentrations were much higher than the typical plasma concentrations, with the C max being 153 g/ml following a 2-g dose. After exclusion of these data, a total of 2,571 plasma cefiderocol concentrations from 329 subjects were used for the development of the population PK models. MIC data for 195 pathogens from 189 patients were used for calculation of the fT MIC . Subject characteristics obtained at the baseline were used (Table 1).

SUPPLEMENTAL MATERIAL
Supplemental material for this article may be found at https://doi.org/10.1128/AAC .01391-17.