Population Pharmacokinetics and Dosing Optimization of Ceftazidime in Infants

ABSTRACT

Ceftazidime, a third-generation cephalosporin, can be used for the treatment of adults and children with infections due to susceptible bacteria. To date, the pediatric pharmacokinetic data are limited in infants, and therefore we aimed to evaluate the population pharmacokinetics of ceftazidime in infants and to define the appropriate dose to optimize ceftazidime treatment. Blood samples were collected from children treated with ceftazidime, and concentrations of the drug were quantified by high-performance liquid chromatography with UV detection (HPLC-UV). A population pharmacokinetic analysis was performed using NONMEM software (version 7.2.0). Fifty-one infants (age range, 0.1 to 2.0 years) were included. Sparse pharmacokinetic samples (n = 90) were available for analysis. A one-compartment model with first-order elimination showed the best fit with the data. A covariate analysis identified that body weight and creatinine clearance (CL_CR) were significant covariates influencing ceftazidime clearance. Monte Carlo simulation demonstrated that the currently used dosing regimen of 50 mg/kg twice daily was associated with a high risk of underdosing in infants. In order to reach the target of 70% of the time that the free antimicrobial drug concentration exceeds the MIC (fT_>MIC), 25 mg/kg every 8 h (q8h) and 50 mg/kg q8h were required for MICs of 4 and 8 mg/liter, respectively. The population pharmacokinetic characteristics of ceftazidime were evaluated in infants. An evidence-based dosing regimen was established based on simulation.

INTRODUCTION

Ceftazidime, a third-generation cephalosporin, is one of the most used antibiotics to treat bacterial infections in children, including neonates. It is active against Gram-positive coccus bacteria such as Streptococcus and common Gram-negative organisms, including Escherichia coli, Haemophilus influenzae, Klebsiella pneumoniae, Moraxella catarrhalis, Proteus mirabilis, Proteus vulgaris, and Providencia stuartii (1, 2). It is usually administered by the parenteral route (intravenously or intramuscularly), is widely distributed to body tissues and fluids, and exhibits low protein binding (about 10%) (2, 3). No ceftazidime metabolites have been identified, and the drug is eliminated almost entirely (80 to 90%) by glomerular filtration (4, 5). Its broad spectrum, as well as its overall low toxicity profile, has made ceftazidime a popular choice for both targeted and empirical therapy in pediatric clinical practice (6).

Despite the extensive use of ceftazidime in practice, pharmacokinetic data for this drug have been scarce in infants (7, 8). Ceftazidime pharmacokinetic parameters have been published for neonates (9, 10) and children (8, 11) and have shown a significant impact of renal maturation on the drug's clearance. Obviously, the lack of pharmacokinetic data in infants may cause improper dosage, increasing the risks of failure of antimicrobial therapy or the emergence of antibiotic resistance in this subgroup of pediatric patients (12). Thus, the objectives of the present work were to evaluate the pharmacokinetics of ceftazidime in infants by using a population approach and to establish an evidence-based dosing regimen of ceftazidime in this vulnerable population.

RESULTS

Study population.

Fifty-one patients from September 2015 to December 2016 were included in the study. All the patients fulfilled the inclusion and exclusion criteria. All of them received 50 mg/kg ceftazidime every 12 h (q12h). No patients discontinued the ceftazidime treatment due to adverse events. The mean (standard deviation [SD]) body weight of the 51 patients at the time of the study was 8.0 (2.4) kg. A summary of patient characteristics is given in Table 1.

TABLE 1.

Baseline characteristics of 51 infants

Characteristic	Total(s)	Mean (SD)	Median (range)a
No. of patients	51
Gender (no. of males, no. of females)	28, 23
No. of samples	90
Age (yrs)		0.7 (0.6)	0.5 (0.1–2.0)
Current wt (kg)		8.0 (2.4)	7.5 (3.0–13.0)
Serum creatinine concn (μmol/liter)		24 (8)	24 (16–44)
Creatinine clearance (ml/min)		122 (30)	124 (65–181)
Ceftazidime dose (mg/dose)		388 (108)	370 (150–580)
Ceftazidime dose (mg/kg/dose)		49 (4)	50 (38–60)

^aRange, minimum value–maximum value.

Model building.

For population modeling, 90 ceftazidime concentrations were available. The ceftazidime concentrations of pharmacokinetic samples ranged from below the limit of quantification (LOQ) to 190 μg/ml; 12 concentrations were said to be below the LOQ. In statistical analyses, these values were deleted and replaced with a constant value of 0.25 μg/ml, half the LOQ. The concentration versus time profile and the concentration with a logarithmic scale versus time profile are shown in Fig. 1A and B.

FIG 1.

Ceftazidime concentration (A) and ceftazidime concentration with a logarithmic scale (B) versus time.

A one-compartment model with first-order elimination fitted the data. The model was parameterized concerning the volume of distribution (V) and clearance (CL) of ceftazidime. Interindividual variability was best described by an exponential model and was then estimated for V and CL. A combined additive and proportional model best described residual variability.

Covariate analysis.

The allometric size approach was used by incorporating a priori the body weight into the base model (allometric coefficients of 0.75 for CL and 1 for V), because this model with fixed allometric coefficients was more fit than that with unfixed coefficients, which caused the significant drop in the objective function value (OFV) of 23 points. Creatinine clearance was identified as the most important covariate for CL, associated with a decline in the OFV of 30.3 points. None of other tested covariates caused further significant improvement of the model.

The median (range) values of estimated weight-normalized CL and V at steady state were 0.17 (0.10 to 0.24) liters/h/kg and 0.40 (0.34 to 0.46) liters/kg, respectively. The area under the concentration-time curve from 0 to 12 h (AUC_0–12) at steady state (AUC_0–12 = dose/CL) for the evaluated dose regimen ranged from 184 to 499 mg · h/liter. Ceftazidime CL increased allometrically with body weight and creatinine clearance in infants. The relationships between ceftazidime CL with body weight and creatinine clearance are shown in Fig. 2A and B.

FIG 2.

The relationship between ceftazidime CL and body weight (A) or creatinine clearance (B).

Model evaluation.

Model diagnostics showed acceptable goodness-of-fit for the final model of ceftazidime. As shown in Fig. 3A and B, predictions are unbiased. In the diagnostic plots of conditional weighted residuals (CWRES) versus time and population prediction (PRED), no trends were observed (Fig. 3C and D). Also, the median parameter estimates resulting from the bootstrap procedure closely agreed with the respective values from the final population model, indicating that the final model is stable and can redetermine the estimates of population pharmacokinetic parameters (Table 2). The normalized prediction distribution errors (NPDE) are presented in Fig. 3E and F. NPDE distribution and the histogram followed the normal distribution N(0, 1) and density, indicating a good fit of the model to the individual data. The mean and variance of NPDE were 0.126 (Wilcoxon's signed-rank test, P = 0.276) and 1.07 (Fisher's variance test, 0.614), respectively.

FIG 3.

Model evaluation for ceftazidime. (A) Population predicted (PRED) versus observed (DV) concentrations; (B) individual predicted (IPRED) versus DV concentrations; (C) conditional weighted residuals (CWRES) versus time; (D) CWRES versus PRED; (E) QQ plot of the distribution of the normalized prediction distribution errors (NPDE) versus the theoretical N(0, 1) distribution; (F) histogram of the distribution of the NPDE, with the density of the standard Gaussian distribution overlaid.

TABLE 2.

Population pharmacokinetic parameters of ceftazidime and bootstrap results

Parametera	Full data set		Bootstrap results
	Final estimate	RSE (%)b	Median	5th–95th
CL (liters/h)
CL = θ1 × (BW/7.5)0.75 × FCRCL × exp(η1) Θ1	1.30	5.3	1.29	1.19–1.41
FCRCL = (CRCL/124)θ3 Θ3	0.82	14.4	0.84	0.63–1.01
V = θ2 × (BW/7.5) × exp(η2) Θ2	2.97	6.1	2.95	2.68–3.21
Interindividual variability (%)
CL	17.0	35.9	16.4	9.6–21.2
V	12.6	67.9	12.7	4.9–18.6
Residual variability (%)
ERR(1)	38.2	25.4	36.9	28.7–44.7
ERR(2)	16.0	87.9	15.6	9.2–22.2

^aV, volume of distribution; CL, clearance; CRCL, creatinine clearance; BW, body weight; ERR, error. In our population, 7.5 kg is the median body weight (day of the study), and 124 ml/min is the median creatinine clearance.

^bRSE, relative standard error.

Dosing regimen evaluation and optimization.

The target attainment (>70% of patients with a concentration exceeding the MIC during 70% of the dosing interval) rates as functions of the simulated dose for standard MIC susceptibility breakpoints of 4 and 8 mg/liter are shown in Fig. 4A and B. When the dosing regimen prescribed in the study (50 mg/kg q12h) was used, 40.5% and 12.8% of infants achieved the pharmacodynamic target (70% of the time that the free antimicrobial drug concentration exceeds the MIC [fT_>MIC]) for MICs of 4 mg/liter and 8 mg/liter, respectively, showing that this was an underdose. When the dosing interval was shortened to 8 h, the pharmacodynamic target (70% fT_>MIC) could be achieved in 67.8% of patients given ceftazidime with a MIC of 4 mg/liter at 25 mg/kg q8h and 67.8% of patients given ceftazidime with a MIC of 8 mg/liter at 50 mg/kg q8h.

FIG 4.

Probability of target (70% fT_>MIC) attainment for ceftazidime with a MIC of 4 mg/liter (A) or 8 mg/liter (B).

DISCUSSION

The present work is the first study that reports the pharmacokinetics of ceftazidime in infants. Our results show that a one-compartment model with first-order elimination was optimal for pharmacokinetic data modeling.

In our study, the mean of the estimated CL value was 0.17 liters/h/kg, which seems to be different from previously published values for neonates and similar to those for children. The ceftazidime CL values of various studies are summarized in Table 3. This difference can be explained by renal maturation. Since ceftazidime is eliminated mainly by the renal route (80 to 90% by glomerular filtration) (2), renal anatomical and functional maturation are expected to have a significant influence on ceftazidime clearance and thereby on the dosing regimen in infants. Creatinine clearance (calculated using the serum creatinine concentration), reflecting renal function, was identified as the most important factor influencing the clearance of ceftazidime. The effect of creatinine clearance has also been reported for ciprofloxacin (13), vancomycin (14, 15), and other drugs that are mainly eliminated by the renal route. During the forward and backward selection process, none of the other covariates tested affected the pharmacokinetic parameters.

TABLE 3.

Pharmacokinetics of ceftazidime obtained from different studies

Parameter	Value
	This study	Neonates		Children
		Reference 9	Reference 10	Reference 11	Reference 8
No. of patients	51	41	42	10	8	8
Age (yrs)	0.7 ± 0.6	Neonates	Neonates	2–13	14.9 ± 1.6	8.0 ± 1.4
Wt (kg)	8.0 ± 2.4	1.9 ± 0.8	1.94 ± 0.8	NAa	51.90 ± 7.15	24.95 ± 3.03
CL (liters/h/kg)	0.17 ± 0.03	0.06 ± 0.03	0.01–0.13	0.20 ± 0.05	0.169	0.226

^aNA, not applicable.

Like all other beta-lactam antibiotics, ceftazidime exhibits time-dependent killing of bacteria (16), T_>MIC is the pharmacokinetic/pharmacodynamic (PK/PD) parameter that correlates with therapeutic efficacy. The PK/PD target of 50% fT_>MIC is a well-characterized target for cephalosporins (17–19). More recently, analysis of data from a randomized phase III clinical study found that 45% fT_>MIC predicted favorable outcomes for ceftazidime in hospital-acquired pneumonia patients (20). Although there are few ceftazidime-specific pharmacodynamic data beyond this study, a goal of 70% fT_>MIC would represent a more conservative endpoint. The infants included in this study were those with pneumonia, bronchitis, and other miscellaneous pulmonary infections. A similar experience was reported from Japan, where 248 patients with pulmonary infections received ceftazidime with an overall success rate of 76.2%, infections caused by P. aeruginosa were the least responsive (65.3%), and patients infected with the common Gram-positive respiratory pathogens and H. influenzae had 82 to 94% response rates (21). Haemophilus influenzae was very sensitive to ceftazidime, with a MIC of 2 mg/liter, and Enterobacteriaceae and Pseudomonas spp. showed ceftazidime MICs of 4 and 8 mg/liter, respectively (22). As shown in simulation results, the current dosing regimen (50 mg/kg q12h) is sufficient (40.5% for a MIC of 4 mg/liter and 12.8% for a MIC of 8 mg/liter) to attain >70% fT_>MIC. To improve PK/PD target attainment, an optimized dosing regimen of 25 mg/kg q8h was required for Enterobacteriaceae (MIC, 4 mg/liter) because of a higher PD parameter attainment (67.8%). For more resistant bacterial strains (e.g., Pseudomonas aeruginosa, with a ceftazidime MIC of 8 mg/liter), 50 mg/kg q8h was required.

However, our study had some limitations. The pharmacokinetic model of ceftazidime was only internally validated; external validation was not performed because of the limited number of patients. Ultimately, the optimal dose regimen based on modeling and simulation should be evaluated in clinical practice to confirm its clinical benefits.

Conclusion.

The population pharmacokinetic model developed for ceftazidime after administration of ceftazidime in infants was validated. Body weight and CL_CR were significant covariates influencing ceftazidime clearance. The currently used dose regimen of 50 mg/kg q12h is associated with a high risk of underdosing in infants. To reach the target 70% fT_>MIC, 25 mg/kg q8h is required for a MIC of 4 mg/liter and 50 mg/kg q8h is required for a MIC of 8 mg/liter. The evidence-based dosing regimen of ceftazidime in children was established based on population PK/PD analysis.

MATERIALS AND METHODS

Study design.

A prospective, open-label pharmacokinetic study of ceftazidime was conducted at Children's Hospital of Hebei Province. The inclusion criteria were as follows: infants who were 1 month to 2 years old with a confirmed or suspected bacterial infection and who had received ceftazidime as part of the regular treatment of bacterial infection. The exclusion criteria were as follows: preterm newborn (gestational age of <37 weeks) with expected survival time less than the treatment cycle, enrollment in another clinical trial, and patients with other factors that the researcher considered unsuitable for inclusion. This study was designed by following legal requirements and the Declaration of Helsinki and approved by the institutional ethics board (Children's Hospital of Hebei Province). Informed-consent forms signed by the patients' parents or guardians were obtained.

Dosage regimen and pharmacokinetic sampling.

Ceftazidime for injection (Guangdong Jincheng Jin Su Pharmaceutical Co. Ltd., Zhongshan, China), was administered as an intravenous infusion over 30 to 60 min at a dose of 50 mg/kg twice daily (q12h). The total number of study-specific blood samples was restricted to 2 per participant at steady state. Patients were randomly assigned to one of two predefined sparse pharmacokinetic sampling schedules: group 1, 4 to 8 h after the start of infusion and 3 to 5 min after the end of infusion; group 2, 8 to 12 h after the start of infusion and 1 to 2 h after the end of infusion. Precise infusion and sample times were recorded. The blood volume of samples obtained for pharmacokinetic analyses was 0.2 ml per sample. Blood samples were refrigerated and centrifuged, and plasma was stored at −80°C. Samples were shipped on dry ice to the department of clinical pharmacy at Shandong Provincial Qianfoshan Hospital, where it was stored at −80°C before analysis.

Analytical method for ceftazidime.

Ceftazidime concentrations were determined using high-performance liquid chromatography with UV detection with metronidazole as an internal standard. The range of the method was 0.5 to 200 μg/ml. The inter- and intraday imprecision rates at all tested concentrations were less than 5%. The lower limit of quantification (LLOQ) was 0.5 μg/ml.

Population pharmacokinetic modeling of ceftazidime.

Pharmacokinetic analysis was carried out using the nonlinear mixed effects modeling program NONMEM version 7.2.0 (Icon Development Solutions, USA). A first-order conditional estimation (FOCE) method with interaction was used to estimate pharmacokinetic parameters and their variability.

Interindividual variability of the pharmacokinetic parameters was estimated using an exponential model and was expressed using the equation θ_i = θ_mean · e^ηi, where θ_i represents the parameter value of the i^th subject, θ_mean represents the typical value of the parameter in the population, and ηi represents the variability between subjects, which is assumed to follow a normal distribution with a mean of zero and a variance of ω2. Different residual variability models were evaluated, including exponential, additive, proportional, and combined models.

Covariate analysis followed a forward and backward selection process. The likelihood ratio test was used to test the effect of each variable on model parameters. In the forward step, the results of body weight (allometric coefficients of 0.75 for CL and 1 for V [23]), age, and creatinine clearance (calculated using the serum creatinine concentration collected within 48 h or less of pharmacokinetic sampling [24]) were investigated as potential variables affecting pharmacokinetic parameters. During the first step of covariate model building, a covariate was included if a significant (P < 0.05, χ² distribution with 1 degree of freedom) decrease (reduction of >3.84) in the objective function value (OFV) from the base model was obtained and a reduction in the variability of the pharmacokinetic parameter occurred. Then all the significant covariates were added simultaneously to the model. Subsequently, in the backward step, each covariate was independently removed from the full model. If the increase in the OFV was larger than 6.635 (P < 0.01, χ² distribution), the covariate was considered significantly correlated with the pharmacokinetic parameter and was therefore retained in the final model.

Model validation was based on graphical and statistical criteria. Goodness-of-fit plots, including those of observed (DV) versus population prediction (PRED), observed (DV) versus individual prediction (IPRED), conditional weighted residuals (CWRES) versus time, and CWRES versus PRED, were initially used for diagnostic purposes (25). The stability and performance of the final model were also assessed using a nonparametric bootstrap with resampling and replacement. Resampling was repeated 1,000 times, and the values of estimated parameters from the bootstrap procedure were compared with those determined from the original data set. The final model was also evaluated graphically and statistically by normalized prediction distribution errors (NPDE) (26, 27). One thousand data sets were simulated using the final population model parameters. NPDE results were summarized graphically by default as provided by the NPDE R package (v.3.4.1) (28): (i) a QQ plot of the NPDE and (ii) a histogram of the NPDE. The NPDE is expected to follow the N(0, 1) distribution.

Dosing regimen evaluation and optimization based on a pharmacokinetic model.

The PK/PD relationship of cephalosporin is the time that the free antimicrobial drug concentration exceeds the MIC (fT_>MIC). In order to obtain the maximal bactericidal activity of cephalosporin, a probability of target attainment of 70% was required for a 70% dosing interval (29, 30). The free fraction of ceftazidime was reported to be about 90% (2, 3). According to the EUCAST (31) and CLSI (22), Haemophilus influenzae and Haemophilus parainfluenzae are very sensitive to ceftazidime, with a MIC of 2 mg/liter, while Enterobacteriaceae and Pseudomonas spp. have ceftazidime MICs of 4 and 8 mg/liter, respectively, in patients with normal renal function. Monte Carlo simulations were performed using the parameter estimates obtained from the final model. The pediatric dose of ceftazidime was simulated on a mg/kg basis. The current dosing regimen (50 mg/kg q12 h) in the original data set was simulated. One hundred simulations were performed, and the time that the concentration exceeded the MIC was calculated for each simulated patient. If the current dosing regimen showed underdosing in the majority of patients (>50%), the optimal dosing regimen with an increased dose and/or frequency was given to the virtual patients (32–34). Thus, various dosing regimens (10, 15, 20, 25, 30, 35, 40, 45, and 50 mg/kg q12h and q8h) were simulated in this group. The probability of target attainment is calculated for each dosing regimen to optimize antimicrobial therapy.