Sie können Operatoren mit Ihrer Suchanfrage kombinieren, um diese noch präziser einzugrenzen. Klicken Sie auf den Suchoperator, um eine Erklärung seiner Funktionsweise anzuzeigen.
Findet Dokumente, in denen beide Begriffe in beliebiger Reihenfolge innerhalb von maximal n Worten zueinander stehen. Empfehlung: Wählen Sie zwischen 15 und 30 als maximale Wortanzahl (z.B. NEAR(hybrid, antrieb, 20)).
Findet Dokumente, in denen der Begriff in Wortvarianten vorkommt, wobei diese VOR, HINTER oder VOR und HINTER dem Suchbegriff anschließen können (z.B., leichtbau*, *leichtbau, *leichtbau*).
Population Pharmacokinetics of Cefepime in Critically Ill Children and Young Adults: Model Development and External Validation for Monte Carlo Simulations and Model-Informed Precision Dosing
This study aimed to develop a population pharmacokinetic model for cefepime in critically ill pediatric and young adult patients to inform dosing recommendations and to evaluate the model’s predictive performance for model-informed precision dosing.
Methods
Patients in the pediatric intensive care unit receiving cefepime were prospectively enrolled for clinical data collection and opportunistic plasma sampling for cefepime concentrations. Nonlinear mixed effects modeling was conducted using NONMEM. Allometric body weight scaling was included as a covariate with fixed exponents. Monte Carlo simulations determined optimal initial dosing regimens against susceptible pathogens. The model’s predictions were evaluated with an external dataset.
Results
Data from 510 samples across 100 patients were best fit with a two-compartment model with first-order elimination. Estimated glomerular filtration rate and cumulative percentage of fluid balance were identified as significant covariates on clearance and central volume of distribution, respectively. Internal validation showed no model misspecification. External validation confirmed that bias and precision for both population and individual predictions were within commonly accepted ranges. Monte Carlo simulations suggested that the usual dose of 50 mg/kg may require a 3-h infusion or a 6-h dosing interval to keep concentrations above the Pseudomonas aeruginosa minimum inhibitory concentration (≤ 8 mg/L) throughout the dosing interval for patients with normal or augmented renal clearance.
Conclusion
A cefepime population pharmacokinetic model for critically ill pediatric patients was successfully developed, accounting for patient renal function, fluid status, and body size, using real-world data. The model was internally and externally validated for use in optimal dosing simulations and model-informed precision dosing.
We developed a population pharmacokinetics model of cefepime using real-world data from critically ill pediatric and young adult patients. The model incorporates allometric body weight scaling to account for body size differences, includes estimated glomerular filtration rate to account for kidney function in cefepime clearance, and uses cumulative fluid balance as a covariate on the central volume of distribution.
The model was externally validated and can be used for model-informed precision dosing.
Monte Carlo simulations suggested a dosing regimen of 50 mg/kg every 12–8 h with a 30-min infusion for 50%fT>minimum inhibitory concentration (MIC) of susceptible pathogens for most patients. For 100%fT>MIC, dosing regimens with extended infusion, shorter interval, or continuous infusions are suggested.
1 Introduction
Cefepime is a fourth-generation cephalosporin antibiotic commonly used in the treatment of severe bacterial infections in both children and adults [1]. As a small hydrophilic molecule, cefepime is widely distributed in biological fluids and tissues after administration. It has a protein binding of approximately 20% and is primarily eliminated by the kidneys, with about 85% of the dose excreted unchanged in the urine [2].
Anzeige
As with all β-lactam antibiotics, the bactericidal activity of cefepime is time dependent. Its microbiological and clinical effects depend on the proportion of each dosing interval that free-drug concentrations remain above the target pathogen’s minimum inhibitory concentration (MIC; %fT>MIC) [3]. Although there is no consensus on the optimal therapeutic target for cefepime, a minimum of 50% fT>MIC has been used, and recent guidelines suggest that 100% fT>1–4×MIC is a more optimal goal for critically ill patients [4].
Despite the broad-spectrum activity and widespread use of cefepime, the standard initial cefepime dosing regimen of 50 mg/kg every 8 h with 30-min infusion does not achieve 100% fT>MIC in most critically ill children [5]. Factors such as changes in organ function, inflammation, and therapeutic interventions (e.g., vasoactive infusions, extracorporeal therapies, fluid resuscitation) can affect cefepime pharmacokinetics, leading to an increased variability in cefepime concentrations [6]. Despite this, cefepime concentrations are not measured in routine clinical care, so critically ill patients receive the same standard cefepime dosing regimens as the general patient population based solely on body weight and estimated kidney function without any confirmation of target attainment or dosing adjustments.
Model-informed precision dosing (MIPD) has been emerging as a promising tool to enhance target attainment of drugs, including antibiotics, in clinical practice [7]. By integrating population pharmacokinetic models with drug concentrations and patient-specific factors (e.g., demographic data and laboratory results), MIPD allows for individualized dosing strategies that can improve therapeutic outcomes and minimize toxicity. However, the implementation of MIPD for cefepime in the pediatric intensive care unit (PICU) is currently limited because of the lack of ready access to cefepime assays at most institutions and the absence of a validated population pharmacokinetic model for this patient population [8].
Therefore, the aim of this study was to develop and validate a population pharmacokinetic model of cefepime for critically ill pediatric and young adult patients to derive model-informed starting dose recommendations tailored to consider pathophysiological features of critical illness and to evaluate the predictive performance of the model for its utility for MIPD.
Anzeige
2 Material and Methods
2.1 Study Design
This study was part of a broader prospective and observational research study conducted in the Cincinnati Children’s Hospital Medical Center (CCHMC) PICU between October 2018 and November 2021 to investigate the pharmacokinetics and therapeutic target attainment of ceftriaxone, cefepime, meropenem, and piperacillin/tazobactam. Study day 1 was defined as the first day the patient received a β-lactam antibiotic dose in the PICU as part of the parent β-lactam study. The study was approved by the CCHMC institutional review board with waiver of consent (#2018–3245, Pharmacokinetics of β-lactams in Critically Ill Pediatric Patients during Different Stages of Sepsis).
2.2 Study Population
This analysis included patients in the PICU aged 1 month to 30 years who received at least one dose of cefepime and had at least one total cefepime concentration measurement. Patients who received renal replacement therapy, including intermittent dialysis, continuous renal replacement therapy, peritoneal dialysis, or extracorporeal membrane oxygenation therapy were excluded.
2.3 Drug Dosing and Administration
Cefepime initiation and dosing regimens were determined by the clinical team for each patient. In our institution, cefepime is typically prescribed as a 50 mg/kg/dose (maximum 2000 mg/dose) every 8 h as a 30-min infusion in patients with preserved kidney function.
2.4 Blood Sampling and Cefepime Quantification
Blood samples were collected using a scavenged opportunistic sampling approach [9]. Residual blood from samples drawn as part of standard clinical practice during the first 7 days of β-lactam therapy was requested from the clinical laboratory and stored at 4 °C. Samples collected during cefepime infusion were excluded. Residual blood or plasma samples were centrifuged (2060×g, 4 °C, 10 min) within 5 days of collection from patients, and the supernatant was stored at – 80 °C for up to 120 days until total cefepime concentrations were measured via high-performance liquid chromatography.
Our group has previously described and used high-performance liquid chromatography to quantify cefepime concentrations in a feasibility study of the opportunistic sampling approach [9]. The assay range exhibited linearity from 0.5 to 200 μg/mL. The assay's precision and accuracy were assessed for quality control, with coefficients of variation maintained below 15% for both within-day and between-day measurements.
2.5 Clinical Data Collection
We reviewed electronic medical records to collect demographic and clinical data for up to 7 study days after initiation of cefepime therapy; data were stored in a secure REDCap database (see the electronic supplementary material [ESM] 1) [10].
Since fluid status is dynamic in patients in the PICU and can affect β-lactam pharmacokinetics [11], we carefully characterized fluid status by noting daily net fluid intake and output data to calculate daily net fluid balance (daily difference in all recorded intakes and all outputs). We then defined the cumulative percentage of fluid balance as the sum of all the previous days’ and present day’s percentage of fluid balance (sum of each day’s net fluid balance/PICU admission body weight × 100) [12]. We calculated the estimated glomerular filtration rate (eGFR) using the bedside Schwartz equation for patients aged <18 years and the Chronic Kidney Disease Epidemiology Collaboration equation for patients aged ≥18 years [13, 14].
When a clinical laboratory measurement was absent on a specific study day, we used the ‘last observation carried backward’ method. The most recent observed data point was extended to fill the missing data. If certain laboratory measurements were missing during a patient's hospital stay, it was imputed by substituting the missing data with the population median value [15].
Anzeige
2.6 Pharmacokinetic Analysis
We conducted the population pharmacokinetic analysis using nonlinear mixed-effects modeling in NONMEM, version 7.5 (Icon Development Solutions, Ellicott City, MD, USA), interfaced with Pirana, version 23.1.2 (Certara, Princeton, NJ, USA), with the first-order conditional estimation with the interaction method.
Both one- and two-compartment models with first-order elimination were tested as the structural model. Interindividual variability (IIV) was assessed as a random-effect ETA (ƞ) for each parameter THETA (θ) using an exponential model. IIV values were assumed to follow a normal distribution with a mean of 0 and variance ω2. To characterize the residual variability, we tested additive, proportional, and combined error models. Cefepime concentrations below the limit of quantification were treated with the Beal M6 method, replacing the observation with half of the limit of quantification [15].
We used allometric scaling to include body weight as a covariate in the base model, employing a power function with a fixed exponent of 0.75 for clearance parameters and 1 for volume parameters, and scaled to a typical adult weighing 70 kg [16]. We tested a maturation factor accounting for developmental changes in renal clearance using a Hill function [17] and assessed the effect of categorical and continuous covariates with biological plausibility with stepwise covariate modeling. During the forward inclusion phase, each covariate was added individually to the base model and included if its addition resulted in a reduction of the objective function value (OFV) by more than 3.84 (p < 0.05). In the subsequent backward elimination phase, each covariate was systematically removed from the full model, retaining those whose exclusion increased the OFV by at least 6.63 (p < 0.01). The final model included covariate effects that could be estimated with a relative standard error lower than 40%, ensuring the reliability and robustness of the parameter estimates.
The final model was internally validated through visual inspections of the goodness-of-fit diagnostic plots, including individual fit plots, observed versus population- and individual-predicted concentrations, scatter plots of the residuals, and a prediction-corrected visual predictive check. We performed a non-parametric bootstrap analysis, where a total of 1000 replicate datasets were generated through random sampling with replacement from the original dataset, and the median and 95% confidence intervals of the parameter estimates were compared with the final model estimates.
Anzeige
2.7 External Validation
We obtained an external dataset, distinct from the one used for model development, containing cefepime concentrations and clinical data from patients in the PICU, from a different prospective study conducted in the CCHMC PICU from November 2022 to August 2024. This distinct study investigated urine biomarkers as predictors of cefepime clearance. For this study, patients in the PICU at high risk of (but not necessary subsequently diagnosed with) acute kidney injury were followed for up to 7 days, and residual blood samples were collected for cefepime concentration measurement using the same scavenged opportunistic sampling approach already described. The risk for acute kidney injury was assessed as part of the standard care at our institution through an automated continuous Renal Angina Index assessment [18]. Cefepime initiation and dosing regimens were determined by the clinical team for each patient.
The final model was applied to the external dataset, and the predicted drug concentrations were compared with the corresponding observed concentrations. Goodness-of-fit plots were created to visualize the relationship between observed and predicted concentrations. We calculated the median prediction error (MDPE) to quantify the model’s bias and calculated the median absolute prediction error (MDAPE) to assess the precision of the model's predictions. We also compared the a posteriori predictions, known as Bayesian or individualized predictions, with the corresponding observed concentrations to assess the model's individual-level performance.
We compared the demographics and clinical characteristics of patients in the model-building cohort and the external validation cohort using the Mann–Whitney U test for continuous variables and the chi-squared test for categorical variables. A p value < 0.05 was considered statistically significant for all comparisons.
2.8 Monte Carlo Simulations
The final model was implemented in Simulx software (2023R1 version, Lixoft, Antony, France) for Monte Carlo simulations.
Anzeige
We created a simulation dataset comprising 6000 patients, equally divided among infants (1 month to 2 years), children (2–12 years), adolescents (12–18 years), and young adults (18–30 years), using a random selection from the CDC-NHANES demographic database, with paired age–weight data [19]. Within each age bracket, eGFRs were classified into three categories: kidney impairment, normal eGFR, and augmented renal clearance (ARC). Normal eGFR was defined as values within two standard deviations of the age-specific median for healthy population, kidney impairment was identified as less than two standard deviations below the median, and ARC as values exceeding two standard deviations above the median, up to six standard deviations [13, 20, 21]. For simulations, the cumulative percentage of fluid balance was set to 0% for all patients.
We then assessed the probability of target attainment (PTA) after steady state (24 h after starting antibiotic treatment) for different cefepime dose regimens: 50 mg/kg (maximum 2000 mg/dose) every 6, 8, 12, or 24 h, administered with a 30-min or 3-h infusion duration and continuous infusions of 30 to 180 mg/kg per day (maximum 6000 mg/day). We sought to discern dosing regimens that ensured at least 90% of simulated patients had free concentrations above three distinct pharmacodynamic targets: 50% fT>MIC, 100% fT>MIC, and 100% fT>4×MIC, using the 2023 Clinical & Laboratory Standards Institute cefepime breakpoints of MIC 2 mg/L for Enterobacteriaceae and MIC 8 mg/L for Pseudomonas aeruginosa [22]. We assumed a fixed 20% protein binding to calculate free cefepime concentrations.
3 Results
3.1 Study Population
A total of 520 plasma cefepime concentrations from 100 patients were considered for inclusion in constructing the population pharmacokinetic model. During data cleaning, we excluded 10 (2%) cefepime concentrations that were identified as potential data entry errors or resulting from analytical issues. As a result, 510 samples were used for model building, with a median of four samples (range 1–22) per patient. Of these, three (0.6%) samples were below the limit of quantification so were imputed as half of the limit of quantification.
The median age of the cohort was 7.6 years (interquartile range [IQR] 1.6–16), and the cohort comprised 27 infants (27%), 32 children (32%), 23 adolescents (23%), and 18 adults (18%). At the initiation of cefepime therapy, 39 patients (39%) had normal renal function, 41 patients (41%) exhibited ARC, and 20 patients (20%) presented with kidney impairment. Table 1 presents a summary of the patient demographics and clinical characteristics.
Table 1
Demographic and clinical characteristics of patients receiving cefepime in the model-building cohort and external-validation cohort
Variable
Model-building cohort (n = 100)
External-validation cohort (n = 41)
p value
Demographic data
Age, years
7.6 (1.6–16)
11.9 (6.3–17.5)
0.1085
Infants
27 (27.0)
6 (14.6)
0.1153
Children
32 (32.0)
15 (36.6)
0.5999
Adolescents
23 (23.0)
10 (24.4)
0.8595
Young adults
18 (18.0)
10 (24.4)
0.3877
Male
55 (55.0)
22 (53.7)
0.8845
Female
45 (45.0)
19 (46.3)
0.8845
Body weight, kg
24.8 (11.9–53.6)
28.8 (15.1–50.9)
0.4471
Clinical data
Serum creatininea , mg/dL
0.38 (0.23–0.63)
0.79 (0.42–1.33)
< 0.001*
eGFRa, mL/min/1.73 m2
128.3 (91–171.6)
62.6 (44.7–121.8)
< 0.001*
Serum albumina, g/dL
2.9 (2.4–3.3)
2.9 (2–3.5)
0.8186
Mechanical ventilationb
45 (45.0)
29 (70.7)
0.0055*
Vasopressor treatmentb
41 (41.0)
29 (70.7)
0.0013*
Cumulative % fluid balance
Day 1
3.3 (0.5–6)
2.4 (0.6–7.1)
0.4684
Day 2
4.5 (1.3–11.1)
5.9 (2.9–11.3)
0.2938
Day 3
5.5 (0.3–11.9)
6.4 (2.8–11.2)
0.5139
Day 4
6.2 (0.6–15.1)
6.7 (2.4–11.7)
0.8570
Day 5
4.9 (0.5–11.9)
6 (0.8–9.9)
0.9534
Day 6
5.7 (−0.1–10.4)
4.3 (1.2–7.4)
0.5217
Day 7
4 (0.2–13.2)
4.2 (0.5–6.4)
0.4508
Data are presented as n (%) or median (interquartile range) unless otherwise indicated. Age groups are classified as follows: infants are 1 month to <2 years, children are 2 to <12 years, adolescents are 12 to <18 years, and young adults are 18 to <30 years
eGFR estimated glomerular filtration rate calculated using the Schwartz equation for patients aged <18 years and the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation for those aged ≥18 years
*Statistically significant values, based on the Mann–Whitney U test for continuous variables and the chi-squared test for categorical variables; p value < 0.05
aClinical data at study day 1
bAny day during the patient’s follow up
3.2 Pharmacokinetic Model Building
The final model was best described as a two-compartment model with IIV included on clearance and central volume of distribution. The data did not support the inclusion of random effects on intercompartmental clearance and peripheral volume of distribution, likely because of the sparse sampling approach. The residual variability was best described by the proportional error model. The model employed first-order elimination, as there was no evidence from diagnostic plots or residual analysis to suggest nonlinearity in cefepime elimination.
After the forward inclusion phase of the covariates, eGFR, systolic blood pressure, heart rate, and PRISMIII score had a significant impact on clearance, whereas only the cumulative percentage of fluid balance affected central volume. In the backward elimination phase, the removal of eGFR from clearance and cumulative percentage of fluid balance from central volume resulted in a significant increase in the OFV, with a ∆OFV of 132.3 and 14.1, respectively, so they were retained in the final model.
The estimated parameters and final model equations are presented in Table 2. Clearance was 6.38 L/h/70 kg0.75 with an IIV of 38.1%, and the central volume of distribution was 15 L/70 kg with an IIV of 14.9%. The intercompartmental clearance and peripheral volume of distribution were estimated at 3.65 L/h/70 kg0.75 and 8.91 L/70 kg, respectively.
Table 2.
Population pharmacokinetic parameters of final cefepime model and bootstrap results
Parameter
Stochastic approximation
Bootstrap estimates (n = 1000)
Estimate
RSE (%)
Median
2.5th percentile
97.5th percentile
Fixed effects
Cl = Clpop × (WT/70)0.75 × (eGFR/147.6)β × eηCl
CL (L/h/70kg0.75)
6.38
5
6.41
5.85
7.04
βeGFR
0.66
11
0.66
0.51
0.79
V1 = V1pop × (WT/70) × eβ ×Cum%FB × eηV1
V1 (L/70kg)
15
28
16.41
10.74
20.39
βCum%FB
0.026
44
0.022
0.002
0.041
Q = Qpop × (WT/70)0.75
Q (L/h/70kg0.75)
3.65
54
3.09
1.12
6.70
V2 = V2pop × (WT/70)
V2 (L/70kg)
8.91
21
8.92
6.55
12.98
Random effects
IIV CL (shrinkage)
38.1% (4.2%)
10
37.9%
29.6%
45.3%
IIV V1 (shrinkage)
14.9% (74.5%)
55
16.4%
3.9%
28.8%
Error model parameter: proportional only
B
31.9%
2
31.4%
27.4%
35.7%
ηCl and ηV1 represent the random-effect parameters for IIVs. The IIV is expressed as coefficient of variation (%) calculated as \(\sqrt {e^{{\omega^{2} }} - 1 } \times 100\), where ω2 corresponds to the variance of the random effects
CL clearance, Cum%FB cumulative percentage of fluid balance, eGFR estimated glomerular filtration rate, IIV inter-individual variability, Q intercompartmental clearance, RSE relative standard error, V1 central volume of distribution, V2 peripheral volume, WT body weight
3.3 Internal Validation
Figure 1 presents the goodness-of-fit plots for the final model, revealing a well-fitting model without apparent trends or signs of misspecification. The residuals are well distributed around zero, with most points within the range of −2 and 2, and display no recognizable patterns. The prediction-corrected visual predictive check showed that the 5th, 50th, and 95th percentiles of the observed concentrations were reassuringly within the 95% confidence interval of predicted intervals (Fig. 2). Bootstrap results confirmed model stability, with all parameter estimates within confidence intervals and close to the original estimates (Table 2).
Fig. 1
Diagnostic goodness-of-fit plots of the final cefepime model: observed concentrations versus a population-predicted concentrations and b individual-predicted concentrations. Conditional weighted residuals (CWRES) versus c population-predicted concentrations and d time in hours. Concentrations are expressed in milligrams per liter (mg/L)
Prediction-corrected visual predictive check for cefepime concentrations. Black dots are observed concentrations, solid lines represent the median and the 5th and 95th percentiles of the observed values, and shaded areas represent the spread of 95% prediction intervals calculated from simulations (n = 1000)
We conducted the external validation of the final model using data from 41 pediatric patients with a total of 234 observed cefepime concentrations. This external cohort (Table 1) had a median age of 11.9 years (IQR 6.3–17.5), a median weight of 28.8 kg (IQR 15.1–50.9), and a median initial eGFR of 62.6 mL/min/1.73 m2 (IQR 44.7–121.8). Figure 3 shows the distribution of the prediction errors. For the population-level predictions, MDPE and MDAPE were 6.9% and 34.5%, respectively. For individual predictions, the MDPE was −1.8% and the MDAPE was 18.3%. Goodness-of-fit plots were generated to visualize the relationship between observed and predicted drug concentrations (Fig. 4).
Fig. 3
Scatter plots of population prediction errors (left) and individual prediction errors (right). Each dot represents the prediction error for an observed cefepime concentration in the external dataset. The shaded area represents the error range between − 25 and 25%
Diagnostic goodness-of-fit plots for the external validation of the model: observed concentrations versus a population-predicted concentrations and b individual-predicted concentrations. Conditional weighted residuals (CWRES) versus c population-predicted concentrations and d time in hours. Concentrations are expressed in milligrams per liter (mg/L)
Full results from Monte Carlo simulations investigating different cefepime dosing regimens for patients with normal renal function, kidney impairment, and ARC are available in ESM 2. Table 3 summarizes the suggested dosing regimens stratified by degree of kidney function for different cefepime targets based on the lowest daily doses required to achieve at least 90% PTA after reaching steady state. When the optimal regimen required an extended infusion, we additionally suggested an alternative regimen with a 30-min intermittent infusion that would also achieve 90% PTA, if possible.
Table 3.
Cefepime dosing recommendations according to kidney function for different therapeutic targets, based on the lowest daily doses required to achieve at least 90% target attainment in 6000 simulated patients in the pediatric intensive care unit
50%fT>MIC
100%fT>MIC
100%fT>4×MIC
MIC = 2 mg/L
MIC = 8 mg/L
MIC = 2 mg/L
MIC = 8 mg/L
MIC = 2 mg/L
MIC = 8 mg/L
KI
50 mg/kg q24h TINF 0.5 h
50 mg/kg q24h TINF 3 h
or
50 mg/kg q12h TINF 0.5 h
50 mg/kg q12h TINF 0.5 h
or
30 mg/kg/day CI
50 mg/kg q12h TINF 0.5 h
or
30 mg/kg/day CI
50 mg/kg q12h TINF 0.5 h
or
30 mg/kg/day CI
50 mg/kg q6h TINF 3 h
or
100 mg/kg/day CI
Normal
50 mg/kg q24h TINF 3 h
or
50 mg/kg q12h TINF 0.5 h
50 mg/kg q12h TINF 0.5 h
50 mg/kg q12h TINF 3 h
or
50 mg/kg q8h TINF 0.5 h
or
30 mg/kg/day CI
50 mg/kg q8h TINF 3 h
or
50 mg/kg q6h TINF 0.5 h
or
50 mg/kg/day CI
50 mg/kg q8h TINF 3 h
or
50 mg/kg q6h TINF 0.5 h
or
50 mg/kg/day CI
Not attainable
ARC
50 mg/kg q12h TINF 0.5 h
50 mg/kg q12h TINF 3 h
or
50 mg/kg q8h TINF 0.5 h
50 mg/kg q8h TINF 0.5 h
or
30 mg/kg/day CI
50 mg/kg q6h TINF 3 h
or
50 mg/kg/day CI
50 mg/kg q6h TINF 3 h
or
50 mg/kg/day CI
Not attainable
"Not attainable" indicates that none of the tested dosing regimens achieved a 90% probability of target attainment. Maximum dose considered: 2000 mg/dose
ARC augmented renal clearance, CI continuous infusion, KI kidney impairment, MIC minimum inhibitory concentration, qxh every × hour, TINF duration of infusion, % fT>MIC percentage of time above MIC
When considering the pharmacodynamic target of 50%fT>MIC, an intermittent infusion regimen (30-min infusion) of 50 mg/kg every 12 h was sufficient to achieve 90% PTA for most patients. However, for patients with ARC, a shorter dosing interval of 8 h was necessary against P. aeruginosa (MIC 8 mg/L).
For the pharmacodynamic target of 100%fT>MIC, extended infusion regimens (3 h infusion) or shorter dosing intervals were required to achieve 90% of PTA against P. aeruginosa (MIC 8 mg/L) in patients with normal renal function or ARC. For instance, in this scenario, the optimal regimen for patients with normal renal function is 50 mg/kg every 8 h with 3-h infusion or 50 mg/kg every 6 h with 30-min infusion. If continuous infusion is preferred, the lowest daily doses for achieving 90% PTA are 30 mg/kg per day for patients with kidney impairment and 50 mg/kg per day for patients with normal renal function or ARC.
None of the tested regimens achieved 90% PTA for the pharmacodynamic target of 100%fT>4×MIC in patients with normal renal function or ARC PTA against P. aeruginosa (MIC 8 mg/L).
4 Discussion
In this study, we successfully developed and externally validated a population pharmacokinetic model for cefepime in critically ill children and young adults using real-world data and an opportunistic sampling strategy. We found that commonly used cefepime dosing strategies may be inadequate for meeting stringent pharmacodynamic targets in patients with normal renal clearance or ARC.
In our final model, the mean cefepime clearance and central volume of distribution were 6.4 L/h/700.75 kg and 15 L/70 kg, respectively. The estimated central volume of distribution aligns with values reported in adult studies but is substantially lower than those in younger cohorts. For example, Shoji et al. [23] reported a typical steady-state volume of distribution of 28.4 L/70 kg in a cohort with a median age of 1 month. Zhao et al. [24] (median age 7 days) and de Cacqueray et al. [25] (median age 1.1 years), using a one-compartment model, reported volumes of distribution of 43.2 L/70 kg and 37.3 L/70 kg, respectively. Although direct comparisons across studies are challenging because of differences in model parameterization, the larger volume observed in neonates and infants is consistent with their higher total body water, which decreases with age, explaining the smaller volume of distribution in our older cohort (median age 7.6 years). Our median clearance is similar to those previously reported in adult studies and studies with younger children, such as Zhao et al. [24] (5.8 L/h/700.75) and de Cacqueray et al. [25] (5.6 L/h/700.75), and slightly lower than the values reported by Shoji et al. [23] (9.6 L/h/700.75), which is probably due to different normalization to covariates [2].
As anticipated, based on the renal elimination of cefepime and as commonly reported in both adult and pediatric cefepime pharmacokinetic models, we found that eGFR significantly improved the model as a covariate on clearance [26]. We also observed a significant impact of the cumulative percentage of fluid balance on the central volume of distribution. As cefepime is a small hydrophilic molecule, its distribution is expected to be affected by fluid accumulation. Critically ill patients frequently develop substantial fluid accumulation, leading to overload, which is associated with increased morbidity and mortality [27]. Although this factor is often overlooked in population pharmacokinetic models, we recently developed a model for meropenem in the PICU, where we similarly found that the cumulative percentage of fluid balance significantly affects the central volume of distribution [28]. We previously demonstrated that a positive cumulative percentage of fluid balance may increase the PTA of β-lactam antibiotics in patients receiving intermittent infusion and that patients receiving continuous infusion regimens may require higher loading doses as fluid balance increases to ensure timely attainment of effective drug concentrations [11]. If fluid balance data are unavailable, the model can still be used by assuming a fluid balance of 0, although this could reduce the accuracy of the volume of distribution estimation.
Although there is no consensus on acceptable ranges for external evaluation of population pharmacokinetic models, previous studies have used 20–30% for bias metrics (MDPE) and 30–35% for precision metrics (MDAPE) [29]. In the present study, the bias and precision metrics for both population and individual predictions of our cefepime model fell within these ranges. The inclusion of patients with likely more severe conditions (given the high percentage of mechanical ventilation and vasopressor use) and significantly lower eGFR in the external validation cohort demonstrates the generalizability of the model [30]. These findings underscore the model's robustness in accurately predicting drug concentrations across a diverse pediatric ICU population. To the best of our knowledge, this is the first externally validated population pharmacokinetic model of cefepime for patients in the PICU. The minimal bias and low precision in the individual-level predictions highlight the model's potential reliability and accuracy for supporting MIPD in clinical practice, combining patient-specific factors and serum concentration data to enable individual dose adjustments.
Although some adult studies suggest toxicity thresholds for cefepime based on trough levels (ranging widely from 22 mg/L to 38 mg/L), there is no consensus on the most relevant pharmacokinetic metric—trough levels, peak concentrations, or area under the curve—nor evidence that these thresholds are applicable to pediatric patients [31]. Therefore, our Monte Carlo simulations focused on efficacy targets, with dosing recommendations based on the lowest doses required to achieve these targets and avoid unnecessary high exposures. Our simulations showed that the usual intermittent infusion of 50 mg/kg every 12–8 h would be sufficient to achieve a 90% PTA when considering the pharmacodynamic target of 50%fT>MIC. For a more stringent pharmacodynamic target of 100%fT>MIC, dose regimens with shorter intervals or extended infusion were required for most patients, especially against P. aeruginosa (MIC 8 mg/L). Shoji et al. [23] also recommended a prolonged infusion of 50 mg/kg every 8 h over 3 h to achieve 90% PTA against pseudomonas infections, but they considered an intermediate target of 60%fT>MIC. de Cacqueray et al. [25] highlighted the challenges of achieving 100% fT>MIC with standard intermittent regimens in pediatric patients and recommended continuous infusion of 100 mg/kg/day against bacteria with an MIC ≥2 mg/L. In our simulations, continuous infusion of 30–50 mg/kg/day was suggested as a potential alternative to high-frequency and extended infusion regimens, but clinical studies supporting the routine use of extended infusion of cefepime in children are lacking.
Monte Carlo simulations are a valuable tool for guiding initial dosing regimens based on predictions of drug exposure and clinical covariates [32]. However, despite the strong predictive performance of our model, critically ill patients often present unique and unpredictable pharmacokinetic profiles due to various factors, including organ dysfunction, inflammation, and ICU-specific interventions—that are not fully accounted for by the model—necessitating further individualization of dosing [6]. This underscores the importance of MIPD, particularly for severe infections or pathogens with high MICs, as it integrates population pharmacokinetic models, drug concentrations, and patient-specific factors to describe individual drug concentration–time profiles, enabling tailored dosing to ensure effective concentrations above the MIC for the specific bacteria [33]. Successful implementation of MIPD requires comprehensive training and a multidisciplinary approach, and recent guidelines offer valuable support and address the challenges of its integration into clinical practice [4, 30, 34].
This study has some limitations. Although the model-building and external-validation cohorts were derived from two independent studies, this was a single-center analysis. As an inherent limitation of the scavenged sampling approach, it is possible that some samples used for cefepime measurement were drawn from the same line where cefepime was infused. Although institutional protocol involves wasting a small amount of blood before obtaining samples for clinical laboratory tests, there is a potential risk of inaccurately elevated concentration measurements. Cefepime's low stability may have led to some sample degradation before processing. Although 95% of the samples were processed within 72 h and all were processed within 5 days, our previous data show approximately 15% degradation at 4 °C by day 3 and 24% by day 5 [9]. The sparse opportunistic sampling limited our ability to fully characterize the IIV of intercompartmental clearance and peripheral volume of distribution in the model and contributed to high ETA shrinkage in central volume of distribution (74.5%), so the diagnostic plots should be interpreted with caution [35]. The absence of data from neonates restricted the applicability of the model to this age group, who have distinct maturation-related pharmacokinetic profiles. Moreover, excluding patients receiving renal replacement therapy or extracorporeal membrane oxygenation therapy prevents us from providing specific dosing recommendations for these patients. Finally, because the samples were obtained from observational studies of patients under routine clinical care, we could not dictate the cefepime dosing strategies used to directly validate the results of our simulations.
5 Conclusion
We successfully developed a two-compartment population pharmacokinetic model of cefepime using real-world data from critically ill pediatric and young adult patients with an opportunistic sampling strategy and externally validated the model for MIPD. The model incorporates allometric body weight scaling to account for body size differences and includes eGFR as a covariate on clearance and the cumulative percentage of fluid balance on the central volume of distribution. Monte Carlo simulations suggested that the usual 30-min intermittent infusion of 50 mg/kg every 12–8 h was sufficient when considering the pharmacodynamic target of 50%fT>MIC, but for a more stringent pharmacodynamic target of 100%fT>MIC, extended infusion or higher-frequency regimens were required for most patients to ensure coverage against P. aeruginosa (MIC ≤8 mg/L).
Declarations
Consent to participate
The study was approved by the Institutional Review Board with a waiver of consent.
Consent for publication
Not applicable.
Code availability
The model codes that support the findings of this study are accessible upon reasonable request by contacting the corresponding author.
Data availability
The data and material are accessible upon reasonable request by contacting the corresponding author.
Ethics approval
This research included two studies approved by the CCHMC institutional review board with waiver of consent: “Pharmacokinetics of β-lactams in Critically Ill Pediatric Patients during Different Stages of Sepsis” (IRB #2018-3245), and “Taking Focus 2” (IRB #2018-0724).
Authors’ contributions
R Morales Junior, HR Hambrick, and S Tang Girdwood conceptualized the study. HR Hambrick, KE Pavia, KM Paice, KA Krallman, L Johnson, M Collins, A Gibson, and C Curry collected the data. P Tang and E Schuler conducted the laboratory analyses. R Morales Junior, T Mizuno, and S Tang Girdwood analyzed the pharmacokinetic data. R Morales Junior and HR Hambrick wrote the original draft of the manuscript, and all authors contributed to reviewing and editing the manuscript. J Kaplan, S Goldstein, and S Tang Girdwood supervised the project.
Funding
This work was generously supported by the National Institutes of Health, including funding from the National Institute of General Medical Sciences under an R35 award (R35GM14670), the Eunice Kennedy Shriver National Institute of Child Health and Human Development T32 Training Program in Pediatric Clinical and Developmental Pharmacology (T32HD069054), and the National Institute of Diabetes and Digestive and Kidney Diseases (T32DK007695).
Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License, which permits any non-commercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc/4.0/.
Population Pharmacokinetics of Cefepime in Critically Ill Children and Young Adults: Model Development and External Validation for Monte Carlo Simulations and Model-Informed Precision Dosing
Verfasst von
Ronaldo Morales Junior
H. Rhodes Hambrick
Tomoyuki Mizuno
Kathryn E. Pavia
Kelli M. Paice
Peter Tang
Erin Schuler
Kelli A. Krallman
Luana Johnson
Michaela Collins
Abigayle Gibson
Calise Curry
Jennifer Kaplan
Stuart Goldstein
Sonya Tang Girdwood
Rybak M. The pharmacokinetic profile of a new generation of parenteral cephalosporin. Am J Med. 1996;100:39S-44S.CrossRefPubMed
2.
Pais GM, Chang J, Barreto EF, Stitt G, Downes KJ, Alshaer MH, et al. Clinical pharmacokinetics and pharmacodynamics of cefepime. Clin Pharmacokinet. 2022;61:929–53.CrossRefPubMedPubMedCentral
3.
Downes KJ, Hahn A, Wiles J, Courter JD, Vinks AA. Dose optimisation of antibiotics in children: application of pharmacokinetics/pharmacodynamics in paediatrics. Int J Antimicrob Agents. 2014;43:223–30.CrossRefPubMed
4.
Fratoni AJ, Nicolau DP, Kuti JL. A guide to therapeutic drug monitoring of β-lactam antibiotics. Pharmacotherapy. 2021;41:220–33.CrossRefPubMed
5.
Cies JJ, Moore WS, Enache A, Chopra A. β-Lactam therapeutic drug management in the PICU. Crit Care Med. 2018;46:272–9.CrossRefPubMed
6.
Scaglione F, Paraboni L. Pharmacokinetics/pharmacodynamics of antibacterials in the intensive care unit: setting appropriate dosing regimens. Int J Antimicrob Agents. 2008;32:294-301.e7.CrossRefPubMed
7.
Wicha SG, Märtson A-G, Nielsen EI, Koch BCP, Friberg LE, Alffenaar J-W, et al. From therapeutic drug monitoring to model-informed precision dosing for antibiotics. Clin Pharmacol Ther. 2021;109:928–41.CrossRefPubMed
8.
Tang Girdwood S, Pavia K, Paice K, Hambrick HR, Kaplan J, Vinks AA. β-lactam precision dosing in critically ill children: current state and knowledge gaps. Front Pharmacol. 2022;13:1044683.CrossRefPubMedPubMedCentral
9.
Tang Girdwood SC, Tang PH, Murphy ME, Chamberlain AR, Benken LA, Jones RL, et al. Demonstrating feasibility of an opportunistic sampling approach for pharmacokinetic studies of β-lactam antibiotics in critically ill children. J Clin Pharmacol. 2021;61:565–73.CrossRefPubMed
10.
Harris PA, Taylor R, Thielke R, Payne J, Gonzalez N, Conde JG. Research electronic data capture (REDCap)—a metadata-driven methodology and workflow process for providing translational research informatics support. J Biomed Inform. 2009;42:377–81.CrossRefPubMed
11.
Morales R, Mizuno T, Paice KM, Hambrick HR, Punt N, Girdwood ST. Impact of fluid balance on beta-lactam antibiotics target attainment: insights from a simulation-based meropenem study. Int J Antimicrob Agents. 2024;64:107267.CrossRefPubMed
12.
Selewski DT, Barhight MF, Bjornstad EC, Ricci Z, de Sousa TM, Akcan-Arikan A, et al. Fluid assessment, fluid balance, and fluid overload in sick children: a report from the Pediatric Acute Disease Quality Initiative (ADQI) conference. Pediatr Nephrol. 2024;39:955–79.CrossRefPubMed
13.
Schwartz GJ, Muñoz A, Schneider MF, Mak RH, Kaskel F, Warady BA, et al. New equations to estimate GFR in children with CKD. J Am Soc Nephrol. 2009;20:629–37.CrossRefPubMedPubMedCentral
14.
Delgado C, Baweja M, Crews DC, Eneanya ND, Gadegbeku CA, Inker LA, et al. A unifying approach for GFR estimation: recommendations of the NKF-ASN task force on reassessing the inclusion of race in diagnosing kidney disease. Am J Kidney Dis. 2022;79:268-288.e1.CrossRefPubMed
15.
Irby DJ, Ibrahim ME, Dauki AM, Badawi MA, Illamola SM, Chen M, et al. Approaches to handling missing or “problematic” pharmacology data: pharmacokinetics. CPT Pharmacomet Syst Pharmacol. 2021;10:291–308.CrossRef
16.
Anderson BJ, Holford NHG. Mechanism-based concepts of size and maturity in pharmacokinetics. Annu Rev Pharmacol Toxicol. 2008;48:303–32.CrossRefPubMed
17.
Tod M, Jullien V, Pons G. Facilitation of drug evaluation in children by population methods and modelling. Clin Pharmacokinet. 2008;47:231–43.CrossRefPubMed
18.
Basu RK, Zappitelli M, Brunner L, Wang Y, Wong HR, Chawla LS, et al. Derivation and validation of the renal angina index to improve the prediction of acute kidney injury in critically ill children. Kidney Int. 2014;85:659–67.CrossRefPubMed
Piepsz A, Tondeur M, Ham H. Revisiting normal (51)Cr-ethylenediaminetetraacetic acid clearance values in children. Eur J Nucl Med Mol Imaging. 2006;33:1477–82.CrossRefPubMed
21.
Van Der Heggen T, Dhont E, Peperstraete H, Delanghe JR, Vande Walle J, De Paepe P, et al. Augmented renal clearance: a common condition in critically ill children. Pediatr Nephrol. 2019;34:1099–106.CrossRefPubMed
Shoji K, Bradley JS, Reed MD, van den Anker JN, Domonoske C, Capparelli EV. Population pharmacokinetic assessment and pharmacodynamic implications of pediatric cefepime dosing for susceptible-dose-dependent organisms. Antimicrob Agents Chemother. 2016;60:2150–6.CrossRefPubMedPubMedCentral
24.
Zhao Y, Yao B-F, Kou C, Xu H-Y, Tang B-H, Wu Y-E, et al. Developmental population pharmacokinetics and dosing optimization of cefepime in neonates and young infants. Front Pharmacol. 2020;11:14.CrossRefPubMedPubMedCentral
25.
de Cacqueray N, Hirt D, Zheng Y, Bille E, Leger PL, Rambaud J, et al. Cefepime population pharmacokinetics and dosing regimen optimization in critically ill children with different renal function. Clin Microbiol Infect. 2022;28:1389.e1-1389.e7.CrossRefPubMed
26.
Suttels V, André P, Thoma Y, Veuve F, Decosterd L, Guery B, et al. Therapeutic drug monitoring of cefepime in a non-critically ill population: retrospective assessment and potential role for model-based dosing. JAC Antimicrob Resist. 2022;4:dlac043.CrossRefPubMedPubMedCentral
27.
Selewski DT, Gist KM, Basu RK, Goldstein SL, Zappitelli M, Soranno DE, et al. Impact of the magnitude and timing of fluid overload on outcomes in critically ill children: a report from the multicenter international assessment of worldwide acute kidney injury, renal angina, and epidemiology (AWARE) Study. Crit Care Med. 2023;51:606–18.CrossRefPubMed
28.
Morales Junior R, Mizuno T, Paice KM, Pavia KE, Hambrick HR, Tang P, et al. Identifying optimal dosing strategies for meropenem in the paediatric intensive care unit through modelling and simulation. J Antimicrob Chemother. 2024;79:dkae274.CrossRef
29.
El Hassani M, Marsot A. External evaluation of population pharmacokinetic models for precision dosing: current state and knowledge gaps. Clin Pharmacokinet. 2023;62:533–40.CrossRefPubMed
30.
Taylor ZL, Poweleit EA, Paice K, Somers KM, Pavia K, Vinks AA, et al. Tutorial on model selection and validation of model input into precision dosing software for model-informed precision dosing. CPT Pharmacomet Syst Pharmacol. 2023;12:1827–45.CrossRef
31.
Moscote-Salazar LR, Ghosh A, Pal R, Raj S, Rahman MM, Agrawal A. Neurotoxicity associated with cefepime: an update to neurocritical care: a narrative review. J Transl Crit Care Med. 2020;2:28.CrossRef
32.
Gijsen M, Vlasselaers D, Spriet I, Allegaert K. Pharmacokinetics of antibiotics in pediatric intensive care: fostering variability to attain precision medicine. Antibiotics (Basel). 2021;10:1182.CrossRefPubMedPubMedCentral
33.
Ngougni Pokem P, Vanneste D, Schouwenburg S, Abdulla A, Gijsen M, Dhont E, et al. Dose optimization of β-lactam antibiotics in children: from population pharmacokinetics to individualized therapy. Expert Opin Drug Metab Toxicol. 2024;20:787–804.CrossRefPubMed
34.
Guilhaumou R, Benaboud S, Bennis Y, Dahyot-Fizelier C, Dailly E, Gandia P, et al. Optimization of the treatment with beta-lactam antibiotics in critically ill patients—guidelines from the French Society of Pharmacology and Therapeutics (Société Française de Pharmacologie et Thérapeutique—SFPT) and the French Society of Anaesthesia and Intensive Care Medicine (Société Française d’Anesthésie et Réanimation—SFAR). Crit Care. 2019;23:104.CrossRefPubMedPubMedCentral
35.
Savic RM, Karlsson MO. Importance of shrinkage in empirical bayes estimates for diagnostics: problems and solutions. AAPS J. 2009;11:558–69.CrossRefPubMedPubMedCentral
