Ceftolozane-Tazobactam Population Pharmacokinetics and Dose Selection for Further Clinical Evaluation in Pediatric Patients with Complicated Urinary Tract or Complicated Intra-abdominal Infections

Ceftolozane-tazobactam, a combination of the novel antipseudomonal cephalosporin ceftolozane and the well-established extended-spectrum β-lactamase inhibitor tazobactam, is approved for treating complicated urinary tract infections (cUTI) and complicated intra-abdominal infections (cIAI) in adults. To determine doses likely to be safe and efficacious in phase 2 pediatric trials for the same indications, single-dose ceftolozane-tazobactam plasma pharmacokinetic data from a recently completed phase 1 trial in pediatric patients (birth to <18 years old) with proven/suspected Gram-negative bacterial infections, along with pharmacokinetic data from 12 adult studies, were integrated into a population pharmacokinetic (popPK) analysis.

ABSTRACT

Ceftolozane-tazobactam, a combination of the novel antipseudomonal cephalosporin ceftolozane and the well-established extended-spectrum β-lactamase inhibitor tazobactam, is approved for treating complicated urinary tract infections (cUTI) and complicated intra-abdominal infections (cIAI) in adults. To determine doses likely to be safe and efficacious in phase 2 pediatric trials for the same indications, single-dose ceftolozane-tazobactam plasma pharmacokinetic data from a recently completed phase 1 trial in pediatric patients (birth to <18 years old) with proven/suspected Gram-negative bacterial infections, along with pharmacokinetic data from 12 adult studies, were integrated into a population pharmacokinetic (popPK) analysis. Two-compartment linear models with first-order elimination described the concentration-time profiles of ceftolozane and tazobactam in pediatric patients well. Renal function and body weight were identified to be significant predictors of ceftolozane-tazobactam pharmacokinetics. Renal function, as measured by the estimated glomerular filtration rate (eGFR), significantly affected the clearance of both ceftolozane and tazobactam. Body weight significantly affected clearance and the distribution volume, also of both ceftolozane and tazobactam. Patients with infections had a 32.3% lower tazobactam clearance than healthy volunteers. Using the final popPK models, simulations of various dosing regimens were conducted to assess each regimen’s plasma exposure and the probability of pharmacokinetic/pharmacodynamic target attainment. Based on these simulations, the following doses are recommended for further clinical evaluation in phase 2 pediatric trials for cUTI and cIAI (in patients with an eGFR of ≥50 ml/min/1.73 m² only): for children ≥12 years old, 1.5 g ceftolozane-tazobactam (1 g ceftolozane with 0.5 g tazobactam), and for neonates/very young infants, infants, and children <12 years old, 20/10 mg/kg of body weight ceftolozane-tazobactam, both via a 1-h intravenous infusion every 8 h.

TEXT

Ceftolozane-tazobactam is a combination antibacterial agent consisting of the novel antipseudomonal cephalosporin ceftolozane and the established extended-spectrum β-lactamase inhibitor tazobactam (1). Ceftolozane-tazobactam has potent in vitro activity against Gram-negative pathogens, including many strains of carbapenem- or multidrug-resistant Pseudomonas aeruginosa and of Enterobacteriaceae (new taxonomy, Enterobacterales) producing common extended-spectrum β-lactamases (ESBL); this in vitro activity has also been confirmed against isolates obtained from pediatric patients (2–7). Ceftolozane-tazobactam, administered every 8 h as a 1-h intravenous (i.v.) infusion at a dose of 1.5 g (ceftolozane at 1 g and tazobactam at 0.5 g), is currently approved for treating complicated urinary tract infections (cUTI) and complicated intra-abdominal infections (cIAI) in adults (8). A phase 3 trial in patients with ventilated nosocomial pneumonia was recently completed (ClinicalTrials.gov registration number NCT02070757).

The pharmacokinetics (PK) of ceftolozane and tazobactam have been investigated in healthy adult volunteers, adults with renal impairment, and adult patients with cUTI and cIAI (9–12). Exposures of both ceftolozane and tazobactam increase in proportion to dose, and their PK are unaffected by each other. Ceftolozane has low protein binding (16 to 21%) and a mean steady-state volume of distribution of 13.5 liters and is renally eliminated (>95% as parent drug) primarily through glomerular filtration with a 3.12-h half-life (1, 8, 10). Tazobactam has a mean steady-state volume of distribution of 18.2 liters and a similarly low protein binding (30%) and is also eliminated by renal excretion (but through a combination of active tubular secretion and glomerular filtration) (10), mostly as the parent drug (20% as the pharmacologically inactive metabolite tazobactam M1), with a 1.03-h half-life (8). Ceftolozane-tazobactam dosing must be adjusted in patients with a creatinine clearance (CL_CR) of ≤50 ml/min (8). Like other β-lactams, the PK/pharmacodynamic (PD) target that best correlates with ceftolozane efficacy is the amount of time (as a proportion of the total dosing interval) that the free drug concentration remains above the MIC, expressed as the %fT>MIC. Data from a neutropenic mouse thigh infection model showed that the %fT>MIC ranged from 26.7% to 35.3% for a 1-log kill (13). Therefore, a %fT>MIC of 30% (of an 8-h dosing interval) for an MIC of 4 μg/ml (the Clinical and Laboratory Standards Institute susceptibility breakpoint for P. aeruginosa [14]) can be assumed to be a suitable PK/PD target for ceftolozane, based on existing regulatory PK/PD guidance for the development of antibacterial treatments for potentially life-threatening infections (15). Tazobactam is a competitive, essentially irreversible inhibitor of specific serine β-lactamases, including common ESBLs (16–19). Since tazobactam does not have intrinsic antibacterial activity, an MIC cannot be determined. Instead, the threshold concentration (C_T) needed to effectively neutralize susceptible bacterial β-lactamases can be used for evaluating target attainment. Data from a mouse model support free tazobactam exposures remaining above a C_T of 1 μg/ml for 20% of the 8-h dosing interval, expressed as a %fT>C_T of 20%, as a suitable PK/PD target for tazobactam (20).

Limited data are available regarding ceftolozane-tazobactam PK in children and adolescents. This was recently explored in a phase 1 study evaluating the PK, safety, and tolerability of single i.v. ceftolozane-tazobactam doses in pediatric patients (birth to <18 years old) with proven/suspected Gram-negative bacterial infections (9). Patients were enrolled in one of six age groups, including term and preterm neonates/very young infants (postnatal age of 7 days to <3 months old), and ceftolozane-tazobactam doses were age adjusted. Plasma exposures of ceftolozane-tazobactam were estimated for each age group using noncompartmental methods and were generally found to be comparable to those previously reported for adults (10). In addition, ceftolozane-tazobactam was well tolerated, and no safety concerns were identified (9).

The next step in the clinical development program of ceftolozane-tazobactam will be to conduct phase 2 trials in children and adolescents. In order to propose ceftolozane-tazobactam dosing regimens likely to be safe and efficacious in phase 2 pediatric trials for cUTI and cIAI, we integrated the ceftolozane-tazobactam plasma PK data from this recently completed phase 1 study into a population pharmacokinetic (popPK) analysis. We subsequently used this model to simulate various age-adjusted dosing regimens and to assess each regimen’s resulting drug exposures (to ensure that they did not exceed those previously observed in adults) and probability of attaining ceftolozane-tazobactam PK/PD targets.

RESULTS

PopPK analysis.

The popPK analysis showed that two-compartment linear models with first-order elimination best described the concentration-time profiles of both ceftolozane and tazobactam. The base model structures were consistent with those of previously published ceftolozane-tazobactam popPK models in adults (21, 22).

In our final ceftolozane model, body weight was a significant covariate on clearance (CL), volume of distribution for the central compartment (V_c), and volume of distribution for the peripheral compartment (V_p), while the estimated glomerular filtration rate (eGFR) was a significant covariate on CL only. The effect of body weight on the PK parameters was included using allometric scaling, with estimated exponents of 0.764, 1.124, and 0.484 for CL, V_c, and V_p, respectively. A visual predictive check (VPC) for ceftolozane in pediatric patients (Fig. 1) demonstrated that the model appeared to be generally unbiased for the entire 8-h dosing interval: there was good agreement between the observed and predicted medians and the observed and predicted 95th percentiles, with only a slight degree of underprediction at the 5th percentile. As is evident in Fig. 2, the goodness-of-fit plots created for the pediatric population showed that the residuals randomly distributed around zero, and the model-predicted concentrations were in good agreement with the observed concentrations for the pediatric population; the same was also seen with goodness-of-fit plots for adults (data not shown). All model-derived popPK parameters for ceftolozane (Table 1) were estimated with good precision (standard error of the mean [SEM] expressed as a percentage, <27%), except for the interindividual variability (IIV) of V_p, which was estimated with a 46.2% SEM.

FIG 1.

Visual predictive check for pediatric data from the final population PK model for ceftolozane (A) and tazobactam (B). The observed 5th percentile, median, and 95th percentile are shown in red. The predicted 5th percentile, median, and 95th percentile are shown in blue; red-shaded areas are the 95% confidence intervals around predicted medians, and blue-shaded areas are the 95% confidence intervals around predicted 5th and 95th percentiles.

FIG 2.

Goodness-of-fit plots for pediatric data, by age cohort, from the ceftolozane final population PK model.

TABLE 1.

Parameter estimates and standard errors from the final population PK model for ceftolozane^c

Parameter	Final parameter estimate				IIV/RV
	Typical value	% SEM	Lower bound of CI	Upper bound of CI	Magnitude	% SEM
CL (systemic clearance, liters/h)	5.882	1.567	5.701	6.063	34.61 %CV	15.60
CL (body wt on CL, power)	0.7637	8.401	0.6380	0.8895
CL (MDRD equation eGFR on CL, power)	0.7036	8.442	0.5872	0.8200
Vc (vol of distribution for central compartment, liters)	10.64	2.507	10.11	11.16	44.26 %CV	24.65
Vc (body wt on Vc, power)	1.124	6.926	0.9718	1.277
Q (intercompartmental clearance, liters/h)	2.545a	10.62	2.015	3.074	NE	NA
Vp (vol of distribution for peripheral compartment, liters)	4.227a	4.466	3.857	4.596	16.33 %CV	46.20
Vp (body wt on Vp, power)	0.4840	14.27	0.3487	0.6193
Cov (IIV in Vc, IIV in CL)b	0.07624	22.32	0.04289	0.1096	NA	NA
Residual variability
Proportional	0.02373	14.52	0.01698	0.03049	15.41 %CV	NA
Additive	0.008780	26.56	0.004210	0.01335	0.09370 SD	NA

^aThese parameter estimates were found to be highly correlated (r² = 0.9313).

^bThe calculated correlation coefficient (r) of the off-diagonal omegas was 0.4977, with r² being equal to 0.2477 for Cov (IIV in V_c, IIV in CL).

^cThe minimum value of the objective function was 18,799.471. CI, confidence interval; CL, systemic clearance; Cov, covariance; %CV, coefficient of variation expressed as a percentage; eGFR, estimated glomerular filtration rate; IIV, interindividual variability; MDRD, modification of diet in renal disease; NA, not applicable; NE, not estimated; Q, intercompartmental clearance; RV, residual variability; SEM, standard error of the mean expressed as a percentage; V_c, volume of distribution for the central compartment; V_p, volume of distribution for the peripheral compartment.

In the final tazobactam model, body weight was a significant covariate on CL, V_c, and V_p, and eGFR and the presence of infection were significant covariates on CL. Body weight was included in the final model using allometric exponents of 0.652, 0.737, and 0.830 for CL, V_c, and V_p, respectively. Additionally, body weight was also included as a covariate on intercompartmental clearance (Q), using allometric scaling and a fixed exponent of 0.75. The tazobactam VPC plot (Fig. 1) showed the model having good agreement between observed and predicted medians and observed and predicted 5th to 95th percentiles over the course of the dosing interval. The tazobactam CL for a patient with any infection was 67.7% of the CL for a participant with no infection, assuming identical body weight and eGFR. The tazobactam goodness-of-fit plots created for the pediatric (Fig. 3) and adult (data not shown) populations suggested that the model fit the observed data well, without any obvious trends in the residuals as a function of concentration or time. Most model-derived popPK parameters for tazobactam (Table 2) were estimated with good precision (<25% SEM).

FIG 3.

Goodness-of-fit plots for pediatric data, by age cohort, from the tazobactam final population PK model.

TABLE 2.

Parameter estimates and standard errors from the final population PK model for tazobactam^d

Parameter	Final parameter estimate				IIV/RV
	Typical value	% SEM	Lower bound of CI	Upper bound of CI	Magnitude	% SEM
CL (systemic clearance, liters/h)	20.8a	5.81	18.5	23.2	52.9 %CV	24.7
CL (body wt on CL, power)	0.652	9.83	0.527	0.778
CL (MDRD eGFR on CL, power)	0.733	8.92	0.604	0.861
CL (any infection on CL, proportion)	0.677	14.8	0.481	0.872
Vc (vol of distribution for central compartment, liters)	12.9a	6.18	11.4	14.5	43.9 %CVb	54.7
Vc (body wt on Vc, power)	0.737	8.43	0.615	0.858
Q (intercompartmental clearance, liters/h)	4.06	6.03	3.58	4.54	NE	NA
Q (body wt on Q, power)	0.75 (fixed)	NA	NA	NA
Vp (vol of distribution for peripheral compartment, liters)	5.06	5.99	4.46	5.65	25.9 %CV	59.7
Vp (body wt on Vp, power)	0.830	8.04	0.699	0.961
Cov (IIV in Vc, IIV in CL)c	0.203b	48.9	0.00846	0.398	NA	NA
Proportional residual variability	0.0555	13.6	0.0407	0.0703	23.6 %CV	NA

^aThese parameter estimates were found to be highly correlated (r² = 0.898).

^bThese parameter estimates were found to be highly correlated (r² = 0.861).

^cThe calculated correlation coefficient (r) of the off-diagonal omegas was 0.875, with r² being equal to 0.766 for Cov (IIV in V_c, IIV in CL).

^dThe minimum value of the objective function was 4,359.41. CI, confidence interval; CL, systemic clearance; Cov, covariance; %CV, coefficient of variation expressed as a percentage; eGFR, estimated glomerular filtration rate; IIV, interindividual variability; MDRD, modification of diet in renal disease; NA, not applicable; NE, not estimated; Q, intercompartmental clearance; RV, residual variability; SEM, standard error of the mean expressed as a percentage; V_c, volume of distribution for the central compartment; V_p, volume of distribution for the peripheral compartment.

Simulations.

Simulated ceftolozane and tazobactam PK parameters in adults at the approved 1.5-g ceftolozane-tazobactam dose are summarized in Tables 3 and 4, respectively. Based on pediatric simulation results (see Table S1 in the supplemental material), the ceftolozane-tazobactam dosing regimens that best met the prespecified PK criterion for the steady-state, unbound area under the concentration-time curve for the 8-h dosing interval (AUC_0–8) and the maximum concentration (C_max) were the adult regimen of 1/0.5 g for patients 12 to <18 years old and 20/10 mg/kg of body weight for patients from birth (postnatal age of 7 days, >32 weeks of gestation) to <12 years old (Tables 3 and 4). In terms of the prespecified PK/PD criterion, these same dosing regimens achieved a ceftolozane probability of target attainment (PTA) of ≥90% for an MIC of 4 μg/ml (Fig. 4) and a tazobactam PTA of ≥90% for a tazobactam threshold concentration of 1 μg/ml (Fig. 5).

TABLE 3.

Ceftolozane exposure and target attainment summary for simulated adult and simulated pediatric patients across age groups, based on recommended age-adjusted doses^c

Parameter and statistic	Value for the following patients:
	Adult patients (n = 1,000)	Patients in pediatric age groups
		12 to <18 yr (1,000 mg, n = 1,200)	≥7 to <12 yr (20 mg/kg, n = 1,000)	≥2 to <7 yr (20 mg/kg, n = 1,000)	≥3 mo to <2 yr (20 mg/kg n = 1,000)	Birtha to <3 mo (20 mg/kg, n = 1,200)
Age (yr)
Mean (SD)	53.57 (19.09)	15.00 (1.74)	9.52 (1.45)	4.51 (1.45)	0.80 (0.44)	0.13 (0.07)
Median	56.00	14.99	9.53	4.53	0.71	0.12
5th, 95th percentile	22.0, 81.0	12.3, 17.7	7.3, 11.8	2.2, 6.7	0.3, 1.7	0.0, 0.2
Steady-state AUC0–8 (μg/ml · h)
Mean (SD)	213.36 (171.91)	172.11 (78.76)	176.08 (69.92)	153.13 (60.32)	162.04 (63.15)	157.36 (65.19)
Median	176.00	160.00	161.00	141.00	149.00	147.00
5th, 95th percentile	95.3, 388.0	73.2, 319.0	90.3, 315.0	77.0, 268.0	81.2, 276.5	71.3, 282.5
Steady-state Cmax (μg/ml)
Mean (SD)	69.66 (45.09)	86.04 (36.25)	92.68 (29.84)	82.37 (23.96)	74.37 (19.33)	62.06 (15.18)
Median	56.10	80.35	86.90	79.35	72.40	60.15
5th, 95th percentile	32.9, 154.0	37.8, 154.0	53.1, 150.0	47.8, 125.0	45.9, 109.0	39.9, 89.4
% of patients with first-dose target attainmentb	100	99.0	99.5	99.2	100	99.7

^aBirth was defined as a postnatal age of 7 days and >32 weeks of gestation.

^bThe ceftolozane PK/PD target was a %fT>MIC of 30% for an MIC value of 4 μg/ml.

^cAUC_0–8, area under the concentration-time curve for the 8-h dosing interval; C_max, peak concentration.

TABLE 4.

Tazobactam exposure and target attainment summary for simulated adult and simulated pediatric patients across age groups, based on recommended age-adjusted doses^c

Parameter and statistic	Value for the following patients:
	Adult patients (n = 1,000)	Patients in pediatric age groups
		12 to <18 yr (500 mg, n = 1,200)	≥7 to <12 yr (10 mg/kg, n = 1,000)	≥2 to <7 yr (10 mg/kg, n = 1,000)	≥3 mo to <2 yr (10 mg/kg, n = 1,000)	Birtha to <3 mo (10 mg/kg, n = 1,200)
Age (yr)
Mean (SD)	46.37 (18.72)	15.00 (1.74)	9.52 (1.45)	4.51 (1.45)	0.80 (0.44)	0.13 (0.07)
Median	44.00	14.99	9.53	4.53	0.71	0.12
5th, 95th percentile	20.0, 78.0	12.3, 17.7	7.3, 11.8	2.2, 6.7	0.3, 1.7	0.0, 0.2
Steady-state AUC0–8 (μg/ml · h)
Mean (SD)	80.71 (148.87)	37.12 (23.29)	38.02 (24.49)	30.22 (19.18)	29.87 (19.05)	26.77 (16.50)
Median	29.80	31.20	32.30	26.10	25.00	22.30
5th, 95th percentile	15.4, 379.0	12.4, 83.1	12.3, 84.5	10.2, 62.9	10.0, 63.5	8.4, 57.3
Steady-state Cmax (μg/ml)
Mean (SD)	30.30 (34.73)	27.92 (14.51)	29.19 (14.69)	24.04 (11.78)	21.69 (10.54)	18.37 (8.81)
Median	17.70	24.75	26.45	21.80	19.50	16.80
5th, 95th percentile	10.3, 108.0	10.7, 54.6	11.7, 57.7	9.6, 46.1	8.5, 42.2	7.6, 35.2
% of patients with first-dose target attainmentb	100	97.4	97.8	95.0	95.8	93.0

^aBirth was defined as a postnatal age of 7 days and >32 weeks of gestation.

^bThe tazobactam PK/PD target was a %fT>C_T of 20% for a C_T value of 1 μg/ml.

^cAUC_0–8, area under the concentration-time curve for the 8-h dosing interval; C_max, peak concentration.

FIG 4.

Ceftolozane first-dose target attainment (%fT>MIC of 30% for an MIC of 4 μg/ml, represented by the vertical line) for adult patients and simulated pediatric patients receiving the recommended dosing regimens. PTA, probability of target attainment.

FIG 5.

Tazobactam first-dose target attainment (%fT>C_T of 20%, represented by the vertical line) for adult patients and simulated pediatric patients receiving the recommended dosing regimens.

DISCUSSION

This paper describes the development of a ceftolozane-tazobactam popPK model utilizing PK data obtained from clinical trials in adults (healthy volunteers, volunteers with renal impairment, and patients with cUTI or cIAI) and children (with proven/suspected Gram-negative bacterial infections from a postnatal age of 7 days to <18 years old). This model was then used for PK/PD simulations to determine ceftolozane-tazobactam dosing regimens estimated to be safe and effective for further clinical evaluation in pediatric trials. Of note, these regimens will require further safety and efficacy study in the setting of pediatric phase 2 and 3 trials before they can be recommended for clinical use. This approach to pediatric dose selection is in line with the accepted clinical development process for pediatric indications. First, intensive PK sampling is performed in a pediatric phase 1 trial to characterize a drug’s PK profile in children; this step was completed previously (9). Second, a popPK-based dose selection process is carried out using a model incorporating all available data (including data from adults) in order to estimate appropriate doses for phase 2 trials; this step is represented by the current work. Third, the known pediatric PK profile is supplemented with sparse PK sampling in phase 2 trials, which may subsequently influence the final dose decisions for phase 3 studies in children; such phase 2 trials were recently initiated. The advantage of this popPK-based approach is that the need for intensive PK sampling in children, a vulnerable patient population, is minimized while still being able to effectively determine appropriate pediatric dosing recommendations.

Our final popPK model was a two-compartment linear model with first-order elimination including body weight, eGFR, and/or the presence of infection as significant covariates on CL, V_c, and/or V_p for ceftolozane and tazobactam. Goodness-of-fit plots and VPCs demonstrated that the pediatric concentration profiles for ceftolozane and tazobactam, including central tendency and associated variability in the pediatric data, were well represented by this final model. Since ceftolozane-tazobactam demonstrated a favorable safety profile in phase 2 and 3 studies conducted in adults, the known adult plasma exposures of ceftolozane and tazobactam offered a robust benchmark to assess the simulated pediatric dosing regimens. Simulated ceftolozane-tazobactam dose regimens of a fixed 1/0.5 g for patients ≥12 years old and a weight-based 20/10 mg/kg for patients <12 years old met both prespecified dose selection criteria. First, the 95th percentiles of simulated pediatric steady-state exposures (AUC_0–8 and C_max) of ceftolozane and of tazobactam remained below the corresponding 95th percentiles of the simulated adult exposures; the intention behind this first, PK-based dose selection criterion was to maintain the distribution of exposures in children within the corresponding distribution in adults, thus ensuring that the resulting pediatric exposures fall within existing clinical experience. Second, the PTA for both ceftolozane and tazobactam PK/PD targets was high with the first dose; it is thus anticipated that these regimens will be efficacious in pediatric patients.

The structure and parameter estimates of our final popPK model were very similar to those of a previously published adult model (21). One difference was that the previous model included infection as a significant predictor of both the CL and V_c of ceftolozane (21). Although we evaluated infection status during covariate selection, it was not a statistically significant predictor of ceftolozane PK and therefore not included in our final ceftolozane model. There are three potential explanations for this difference. First, since healthy volunteers are generally never enrolled in pediatric clinical trials evaluating disease treatments, all patients in the pediatric phase 1 study (n = 37) had infections. This may have acted as a partial confounder, since adult data were obtained from both healthy volunteers and patients. Second, the pediatric phase 1 trial had a small sample size, which could have impacted the results. Third, infection was defined differently in the adult trials (i.e., either cIAI or cUTI) and the pediatric phase 1 trial (i.e., patients received antibacterial therapy for any proven/suspected Gram-negative bacterial infection or for perioperative prophylaxis). Importantly, no tazobactam data were available for adult patients with cUTI. It was therefore not possible to investigate the specific effect of cUTI on tazobactam CL, and we assumed that the effect of any infection was the same. Inclusion of renal function as a covariate on CL for ceftolozane and tazobactam was as expected, however, given that both drugs are primarily renally excreted. By including body weight and eGFR as covariates, additional correction for pediatric patients through the use of other ontogeny factors (such as renal maturation) was no longer needed, and by using separate eGFR ranges for each pediatric age group in the simulations, renal function maturation was incorporated.

The final model was subsequently used in simulations intended to determine the best age-adjusted dosing regimen(s) in pediatric patients. The age groups selected for these simulations reflected the source data from the pediatric phase 1 trial, which enrolled patients into the same groupings. For each age group, we simulated various regimens. In adolescents (i.e., individuals ≥12 years old), we opted to simulate fixed dosing regimens only, due to this group’s overlapping characteristics with adults (e.g., body weight distributions). Since adults receive a fixed dose, this overlap suggested that there was no need to explore weight-based dosing, which is more challenging to correctly administer in the clinical setting, in adolescents. In children ages ≥7 to <12 years, both fixed and weight-based dosing regimens were simulated, to understand how the body weight distribution in this age group might impact exposure distributions. Based on those results, we concluded that a switch to simulating only weight-based dosing regimens was appropriate for children <12 years of age.

All evaluated doses met the PK/PD criterion for ceftolozane, and only the lowest ceftolozane-tazobactam dose (i.e., 10/5 mg/kg) in the youngest age group did not meet the PK/PD criterion for tazobactam. On the other hand, most evaluated regimens failed to meet the prespecified PK criterion, since they resulted in steady-state exposures higher than the corresponding values in the simulated adult population. Whenever more than one simulated dosing regimen met all selection criteria in a specific age group, we strove to select the regimen with the lowest dose and shortest infusion time as the recommended, optimized dose; keeping dosing regimens similar between age groups was another important consideration. Given that all simulated regimens resulted in a PTA well in excess of 90% for both ceftolozane (>99%) and tazobactam (93% to 98%), there is no apparent benefit in selecting higher doses and/or utilizing prolonged infusions. Selecting the lowest effective dose with a 1-h infusion time provides the best balance between target attainment and minimizing the use of health care resources. In addition, by recommending the lowest practical total number of age-adjusted dosing regimens and only 1-h infusions (as is standard in adults), we hope to simplify the administration of ceftolozane-tazobactam in pediatric clinical trials. With the recommended pediatric doses, the predicted C_max distribution (i.e., within the 5th and 95th percentiles) overlapped that of simulated adults, even though the mean simulated ceftolozane C_max was up to one-third greater in pediatric patients ≥2 years old than in adults, with the greatest difference being in the ≥7- to <12-year-olds. In the younger age groups, variability in plasma concentrations is minimized through the use of a weight-based rather than a fixed dose. We also recommend that weight-based doses in patients <12 years old not exceed the currently approved adult dose (i.e., 1.5 g for cUTI and cIAI), because of the extensive clinical experience at this dose in adults. Since the pediatric source data for the popPK model only included plasma concentrations from patients with an eGFR of ≥50 ml/min/1.73 m², our dose recommendations do not apply to patients with a lower level of renal function.

Our analyses were associated with several limitations. First, since rich concentration data for ceftolozane and tazobactam were available only from healthy adult volunteers (and some patients from the pediatric phase 1 trial), the structural form of the PK models was largely driven by these more informative phase 1 trial data, rather than the sparse PK sampling from clinical trials in patients. Second, the amount of pediatric data compared to the amount of adult data was relatively small. We addressed this by exploring model diagnostics in the pediatric population alone (Fig. 1 and 3), to ensure that no systematic bias affected the results for the pediatric population. Alternative approaches, such as Bayesian analysis, are also sometimes used to limit the influence of adult data in similar analyses. Third, only very limited data from pediatric patients with renal insufficiency were available for inclusion in the popPK model, a noteworthy point, given that both ceftolozane and tazobactam are renally excreted and that ceftolozane-tazobactam doses in adults need to be adjusted for those with impaired renal function. Fourth, since data from the underlying pediatric phase 1 study only included patients with a gestational age of >32 weeks and a postnatal age of >7 days, our simulations were limited to these gestational and postnatal ages. Additional phase 1 work is ongoing to characterize the impact of these two age-related factors on ceftolozane-tazobactam PK in neonates. Despite these limitations, our modeling and simulation analyses provide strong support for optimized pediatric dose selection.

In conclusion, the popPK models developed for ceftolozane and tazobactam well characterized the ceftolozane-tazobactam PK in adults and pediatric patients from birth to <18 years of age. Dosing simulations suggested that for pediatric patients with an eGFR of ≥50 ml/min/1.73 m², the optimum ceftolozane-tazobactam doses for further clinical evaluation in phase 2 trials are (i) 1.5 g ceftolozane-tazobactam for patients ≥12 years old and (ii) 20/10 mg/kg ceftolozane-tazobactam for patients <12 years old, both administered via a 1-h infusion every 8 h. The total amount of weight-based dosing should not exceed the adult 1.5-g dose currently approved for cUTI and cIAI. The efficacy and safety of those recommended dose regimens are currently being assessed in phase 2 trials enrolling pediatric patients with cUTI (ClinicalTrials.gov registration number NCT03230838) and cIAI (ClinicalTrials.gov registration number NCT03217136).

MATERIALS AND METHODS

Study population.

Plasma concentration-time data, along with study- and patient-specific information, from 13 clinical trials involving ceftolozane alone or ceftolozane-tazobactam were pooled; these trials included 7 studies in healthy adult volunteers (10, 23–28), 5 in adults with infections and/or renal impairment (11, 12, 29), and 1 in pediatric patients (9). For ceftolozane, data from 452 participants and 5,699 observed concentrations, including 31 pediatric patients contributing 123 concentrations, were used. For tazobactam, data from 318 participants and 3,148 concentrations, including 30 pediatric patients contributing 102 concentrations, were used. For adults, rich concentration data were available only from healthy volunteers in phase 1 studies, while adult patients with cIAI or cUTI were sparsely sampled. In the phase 1 pediatric trial, neonates/very young infants (postnatal age of 7 days to <3 months old; n = 6 included in the present analysis) underwent sparse PK sampling, while patients ≥3 months (n = 23 included in the present analysis) underwent rich PK sampling; detailed sampling schedules were published previously (9).

Dosing regimens.

The dose and the frequency of ceftolozane and/or tazobactam administration varied across the included clinical trials. In adults, ceftolozane was administered at 250 to 3,000 mg and tazobactam was administered at 250 to 1,500 mg either alone or in combination. In 4 studies, ceftolozane and/or tazobactam was administered as a single dose. In 9 studies (including 3 studies that also had single-dose cohorts), ceftolozane and/or tazobactam was administered as multiple doses every 6 h, every 8 h, every 12 h, or once daily. In the phase 1 pediatric study, age-based, single ceftolozane-tazobactam doses ranging from 12/6 mg/kg of body weight to 30/15 mg/kg of body weight (not to exceed 1.5 g) or the adult 1.5-g dose was evaluated in the following age groups: ≥12 to <18 years, ≥7 to <12 years, ≥2 to <7 years, ≥3 months to <2 years, and a postnatal age of 7 days (>32 weeks of gestation) to <3 months. Data from a 6th pediatric cohort, i.e., individuals with a postnatal age of 7 days (≤32 weeks of gestation) to <3 months old, were not available in time for inclusion in the present analysis. In all studies, ceftolozane and/or tazobactam was administered as a 60-min i.v. infusion.

Bioanalytical methods.

Plasma ceftolozane and tazobactam total drug concentrations were measured using a validated liquid chromatography-tandem mass spectrometry (LC-MS/MS) method as described previously (9, 10, 24). The ceftolozane lower limits of quantitation (LLOQs) for the assay were 0.10 μg/ml for 5 studies and 0.25 μg/ml for the remaining 8 studies, while the tazobactam LLOQ was 0.10 μg/ml for the 10 studies in which tazobactam was administered.

PopPK methods.

All exploratory data analyses were performed using SAS (version 9.4) software (SAS Institute, Chesterbrook, PA, USA), KIWI (version 1.6) software (Cognigen Corporation, Buffalo, NY, USA), and Perl-speaks-NONMEM (version 4.4.0) software (30). PopPK modeling was performed using the computer program NONMEM (version 7, level 3.0; ICON Development Solution, North Wales, PA, USA). Model goodness of fit was evaluated based on successful model convergence, reasonable parameter estimates with adequate precision, typical goodness-of-fit diagnostic plots, and other diagnostics as suitable for a particular model.

Base structural model development.

Base structural PK model development used all available data, including complete data from all adult clinical studies and most data from the phase 1 pediatric study (i.e., except for the cohort with a postnatal age of 7 days [≤32 weeks of gestation] to <3 months old, which was the only pediatric cohort without available data at that time). As the first step, based on a preliminary examination of the concentration-time data and on the previously published popPK models for ceftolozane and tazobactam (21, 22), two-compartment models with first-order elimination were assessed for appropriateness in describing the pooled ceftolozane and tazobactam PK data. The number of compartments tested was guided by the number of apparent linear phases exhibited in semilogarithmic concentration-time plots. Because all source PK data were collected following i.v. dosing, clearance and volume parameters represented absolute values.

Covariate analysis.

The following covariates were evaluated: age, sex, race, body weight, eGFR, and presence of infection. For ceftolozane, presence of infection was categorized as healthy, cUTI, cIAI, or other infections (i.e., unspecified presumed/documented Gram-negative bacterial infections in pediatric patients). For tazobactam, adult data were available only for patients with cIAI, and the presence of infection was thus categorized as healthy versus any infection. For our analyses, eGFR for adult participants was calculated by the modification of diet in renal disease (MDRD) equation (31), whereas for pediatric patients, eGFR was calculated using the modified Schwartz formula (32). As per the source study protocols, the estimated CL_CR for adult participants was calculated using the Cockcroft-Gault formula (33).

All covariates were assumed to have remained constant throughout each study period. During covariate evaluation, the effect of body weight on CL, V_c, Q, and V_p was assessed by incorporating allometric scaling by body weight (relative to a 70-kg individual) using an estimated exponent, as described in the following equation: CL = TVCL × (WTKG/70)^θCL, where CL is the clearance for an individual, TVCL is the typical value of clearance, WTKG is the participant body weight (in kilograms), and θCL is an exponent for the allometry function (for CL).

Similar equations were used for V_c, V_p, and Q; in each equation, theta (θ) values were initially attempted to be fitted to the data. The effects of age on CL, V_c, V_p, and Q and of eGFR on CL and Q were evaluated using a power model with a reference age of 38 years and a reference eGFR of 100 ml/min/m². The effects of categorical covariates (sex, race, and presence of infection) on PK parameters were evaluated using a proportional model. Because each individual race category other than white represented <10% of the overall population, race was categorized as white versus nonwhite.

The significance of covariate relationships was evaluated using a forward inclusion/backward elimination process. For the forward inclusion process, covariates were deemed significant if they contributed a change in the minimum objective function value (OFV) of ≥6.64 (α = 0.01, 1 degree of freedom [df] for the χ² distribution) and resulted in a decrease in IIV of ≥5% in the parameter of interest. Backward elimination was then conducted, whereby a covariate was retained if it resulted in a change in the OFV of ≥10.83 (α = 0.001, 1 df for the χ² distribution) when removed from the model.

Model refinement.

The reduced multivariable model (i.e., the model with only the significant covariates included) was evaluated for any remaining biases in IIV and proportional residual variability (RV) error models. This evaluation included the possible addition of new IIV terms to other parameters in the model, reevaluation of the appropriateness of the functional form for each IIV term and the RV model, and assessment of possible correlations between η variables. To assess the possibility of correlations, the individual estimates of the η variable for each parameter with an interindividual error term were plotted against the individual η estimates for all other parameters included in the model (η biplots). If correlations between the η variables were observed for any of the parameters, efforts were made to estimate the corresponding covariance between the interindividual error terms in the model. Covariance terms were tested only between η variables that applied to the same population. All IIV and RV models were evaluated for bias, and alternative error models were evaluated as necessary.

Diagnostic plots of the unexplained IIV in the parameters versus all covariates were evaluated to detect potential inadequacies or biases in the covariate models and to ensure that no trends that may have indicated a potential relationship that had not been sufficiently described by the model remained. The model was then checked for possible simplifications of covariate equations, such as power functions that could be reduced to linear functions or significant discrete group covariates that could be redefined using fewer groups or parameters.

Model evaluation.

Assuming that uncertainty in the final model parameters was small relative to other sources of variability, the adequacy of the final model was evaluated by a simulation-based VPC method. In NONMEM, the final model was used to simulate 1,000 replicates of the analysis data set. The 5th percentile, median, and 95th percentile of the distributions of simulated and observed concentrations were then calculated. These percentiles were subsequently plotted versus time, with the original observed data set and/or percentiles based on the observed data being overlaid, in order to visually assess the concordance between the model-based simulated and the observed data. In addition, the percentages of observed data falling below or above the simulation-based prediction interval were calculated.

Simulations.

Using the final popPK models, simulations were performed to determine the expected range of ceftolozane and tazobactam exposures in pediatric patients resulting from various, preselected ceftolozane-tazobactam dosing regimens. The following ceftolozane-tazobactam dosing regimens were evaluated by age cohort: (i) for individuals ages ≥12 to <18 years, 1/0.5 g, 1.5/0.75 g, and 2/1 g as 1-h infusions; (ii) for individuals ages ≥7 to <12 years, 20/10 mg/kg, 30/15 mg/kg, 40/20 mg/kg, 60/30 mg/kg, 1/0.5 g, 1.5/0.75 g, and 2/1 g as 1-h infusions and 20/10 mg/kg as a 2-h infusion; (iii) for individuals ages ≥2 to <7 years and ≥3 months to <2 years, 20/10, 30/15, 40/20, or 60/30 mg/kg as a 1-h infusion; and (iv) for individuals from birth (postnatal age of 7 days, >32 weeks of gestation) to <3 months of age, 10/5, 20/10, 30/15, or 40/20 mg/kg as 1-h infusions and 20/10 mg/kg as a 2-h infusion. Dose regimens were simulated as administered every 8 h for 3 full days (for a total of 9 doses, with the last dose being administered at 64 h), which was assumed to produce steady-state exposures. Concentration values were simulated every 5 min during infusion and every 15 min postinfusion for the first and last infusions.

A data set of virtual pediatric patients was used as the basis for these simulations (see Table S2 in the supplemental material). The virtual population was generated using available data for weight distributions by sex and age subgroup from the Centers for Disease Control and Prevention chart of weight for age for patients aged 0 to 3 months and from the National Health and Nutrition Examination Survey (NHANES) database for patients aged 3 months to 18 years (34, 35). For the age groups of ≥12 to <18 years and birth (postnatal age of 7 days, >32 weeks of gestation) to <3 months old, 1,200 virtual patients were simulated per group. For the age groups of ≥7 to <12 years, ≥2 to <7 years, and ≥3 months to <2 years, 1,000 virtual patients were simulated per group. eGFR distributions for this virtual population were derived from data distributions reported in NHANES III for patients ages ≥2 to 18 years (36) and reported by Schwartz and Furth (37) and Heilbron et al. (38) for younger patients (Table S3). For each virtual patient, eGFR values were assigned based on age and sex using the normal distribution and the associated mean and standard deviation; renal function maturation was therefore indirectly incorporated into the simulations, because separate eGFR ranges were used for each age group. However, the eGFR distributions were truncated by requiring that eGFR be no less than 50 ml/min/1.73 m² for each patient to match the eligibility criteria for planned pediatric phase 2 clinical trials in patients with cUTI and cIAI. A detailed covariate summary for simulated pediatric patients is presented in Table S4. Because ceftolozane-tazobactam is approved and associated with a favorable safety profile in adults with cIAI and cUTI, we compared simulated ceftolozane-tazobactam exposures in virtual pediatric patients with adult exposures. To determine comparative exposures, an additional simulation was conducted for 1,000 adult patients; this simulation was generated by bootstrapping from adult patients enrolled in two ceftolozane-tazobactam phase 2 trials (12, 30). Empirical Bayesian estimates for adult patients from the final model were used as the basis of the simulation.

The simulated exposure measures for ceftolozane and tazobactam were summarized for the first dose and at steady state (defined as the last dose on day 3). Three parameter values were determined for each simulated patient and summarized for each age group and dosing regimen: (i) C_max and (ii) AUC_0–8 for ceftolozane and tazobactam, based on total drug concentration for the dosing interval, and (iii) the percentage of time in a dosing interval that the patient’s free drug concentrations met the respective PK/PD targets for ceftolozane and for tazobactam. Based on previously published preclinical data, a %fT>MIC of 30% for an MIC of 4 μg/ml was selected as the PK/PD target for ceftolozane and a %fT>C_T of 20% was selected as the PK/PD target for tazobactam (13, 14, 20). AUC values were computed from the simulated concentration data for each patient profile using the linear trapezoidal method. The free drug concentration was calculated by assuming protein binding of 21% and 30% for ceftolozane and tazobactam, respectively (8). PTA, defined as the percentage of simulated patients who met the respective PK/PD targets for free ceftolozane or free tazobactam, was estimated for each age group and dosing regimen.

There were two prespecified criteria for pediatric dose selection. First, a PK criterion acted as a maximum exposure bound; specifically, the 95th percentile of either ceftolozane or tazobactam exposures (i.e., AUC_0–8 and C_max) could not exceed the corresponding 95th percentiles of adult exposures, which would indicate that the simulated pediatric PK distributions fell within the distributions simulated for adults at the approved dosing regimen. Second, a PK/PD criterion, i.e., a PTA of ≥90% for the PK/PD targets of both ceftolozane and tazobactam after the first dose, acted as a potential efficacy bound.