Identifying optimal dosing strategies for meropenem in the paediatric intensive care unit through modelling and simulation

Abstract

Background

Meropenem, a β-lactam antibiotic commonly prescribed for severe infections, poses dosing challenges in critically ill patients due to highly variable pharmacokinetics.

Objectives

We sought to develop a population pharmacokinetic model of meropenem for critically ill paediatric and young adult patients.

Patients and methods

Paediatric intensive care unit patients receiving meropenem 20–40 mg/kg every 8 h as a 30 min infusion were prospectively followed for clinical data collection and scavenged opportunistic plasma sampling. Nonlinear mixed effects modelling was conducted using Monolix^®. Monte Carlo simulations were performed to provide dosing recommendations against susceptible pathogens (MIC ≤ 2 mg/L).

Results

Data from 48 patients, aged 1 month to 30 years, with 296 samples, were described using a two-compartment model with first-order elimination. Allometric body weight scaling accounted for body size differences. Creatinine clearance and percentage of fluid balance were identified as covariates on clearance and central volume of distribution, respectively. A maturation function for renal clearance was included. Monte Carlo simulations suggested that for a target of 40% fT > MIC, the most effective dosing regimen is 20 mg/kg every 8 h with a 3 h infusion. If higher PD targets are considered, only continuous infusion regimens ensure target attainment against susceptible pathogens, ranging from 60 mg/kg/day to 120 mg/kg/day.

Conclusions

We successfully developed a population pharmacokinetic model of meropenem using real-world data from critically ill paediatric and young adult patients with an opportunistic sampling strategy and provided dosing recommendations based on the patients’ renal function and fluid status.

Introduction

Meropenem is a β-lactam antibiotic commonly administered to critically ill patients for treatment of severe infections because of its broad antimicrobial spectrum, excellent tissue penetration and favourable safety profile.¹

Meropenem is a small hydrophilic molecule with low degree of protein binding (<2%). It is excreted primarily unchanged by the kidneys through glomerular filtration.² Like all β-lactam antibiotics, the bactericidal activity of meropenem is time-dependent, with microbiological and clinical effects dependent on the proportion of each dosing interval that free drug concentrations remain above the target pathogen’s minimum inhibitory concentration (%fT > MIC).³ Although there is no consensus on the optimal therapeutic target for meropenem, a minimum of 40% fT > MIC has been utilized, while recent guidelines suggest 100% fT > 1–4 × MIC as a more optimal goal for critically ill patients.⁴

Meropenem dosing in children is challenging due to high interindividual variability (IIV) in body composition, metabolism, growth and development.⁵ Current dosing regimens based solely on patient weight have resulted in suboptimal exposure in paediatric patients.⁶ In addition, the physicochemical characteristics of meropenem make its pharmacokinetics sensitive to real-time changes in the dynamic intensive care setting, including changes in renal function, host inflammatory response and therapeutic interventions (e.g. vasoactive infusions, extracorporeal therapies and fluid resuscitation).⁷

Therefore, the aim of this study was to develop a population pharmacokinetic model of meropenem for critically ill paediatric and young adult patients and to derive empiric dosing recommendations tailored to consider pathophysiological features of critical illness.

Materials and methods

Study design

This analysis is part of a larger prospective observational study conducted in the paediatric intensive care unit (PICU) of Cincinnati Children’s Hospital Medical Center between October 2018 and November 2021 designed to investigate pharmacokinetics and therapeutic target attainment of ceftriaxone, cefepime, meropenem, or piperacillin/tazobactam. The study was approved by the Institutional Review Board, which granted a waiver of consent (#2018-3245, Pharmacokinetics of β-lactams in Critically Ill Pediatric Patients during Different Stages of Sepsis).

Study population

This analysis included PICU patients aged 1 month to 30 years receiving meropenem. Patients who received continuous renal replacement therapy (CRRT) or extracorporeal membrane oxygenation therapy (ECMO) were excluded.

Drug dosing and administration

Meropenem initiation and dosing regimen were determined by the clinical team for each patient. In our institution, meropenem is commonly prescribed as 20–40 mg/kg/dose (maximum 2000 mg/dose) every 8 h with 30 min infusion.

Blood sampling and meropenem quantification

Blood samples were obtained using a scavenged opportunistic sampling approach.⁸ 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 centrifuged (2060 × g, 4°C, 10 min) within 7 days of collection. The supernatant was removed and stored at −80°C for up to 120 days until meropenem concentrations were measured via high-performance liquid chromatography method previously described.⁸ We have previously found that meropenem does not degrade more than 15% under these conditions.⁸ Samples collected during meropenem infusion were excluded.

Due to the very low protein binding of meropenem (approximately 2%),⁹ the quantified total concentration was considered as equivalent to the free, unbound concentration.

Clinical data collection

Demographic and clinical data were extracted from electronic medical records for up to 7 study days after starting meropenem and stored in a secure REDCap database.¹⁰ Data included: postmenstrual age (PMA), post-natal age, sex, body weight, height, paediatric mortality risk scores (PRISMIII, PIM2, PIM3),^11–13 vasoactive drug infusion, body temperature, infusion of blood products (red blood cells, platelets, fresh frozen plasma, cryoprecipitate), C-reactive protein, serum albumin, procalcitonin, urea, serum creatinine, serum pH and lactate. Estimated glomerular filtration rate (eGFR) was calculated using the bedside Schwartz equation for patients under 19 years old, and the Chronic Kidney Disease Epidemiology Collaboration equation for patients 19 years and above.^14,15

To characterize fluid status, we extracted daily net fluid intake and output data to calculate daily net fluid balance (daily difference in all intakes and all outputs). The daily percentage of fluid balance was calculated as the daily fluid balance adjusted for body weight [(total fluid in − total fluid out for the day)/PICU admission body weight × 100]. The cumulative percentage of fluid balance was then determined 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).¹⁶

When a clinical laboratory measurement was missing on a given study day, the last observation carried backward method was applied, whereby the most recent observed data point from the study period was extended to fill the gap. Additionally, in cases where a patient had no measurements for specific laboratory data during their hospitalization, we employed an imputation method by considering the median values from the studied population.¹⁷

Pharmacokinetic analysis

The population pharmacokinetic analysis was performed by nonlinear mixed effects modelling using MonolixSuit software (2023R1 Version, Lixoft, Antony, France). Meropenem concentrations below the limit of quantification were treated as interval censored data with a lower limit of 0.^17,18

One-compartment and two-compartment models with first-order elimination were tested to fit the data. We designated each day of meropenem therapy as an occasion in order to align with the daily clinical laboratory assessments commonly performed in critical care. This approach enabled us to assess daily fluctuations in pharmacokinetic parameters while exploring inter-occasion variability (IOV). Both IIV and IOV were tested as random effects eta (η) for each parameter theta (θ) via exponential model. To describe the residual variability, we assessed constant, proportional and combined error models.

Body weight was a priori included as a covariate in the base model using a power function with a fixed exponent of 0.75 for clearance parameters and 1 for volume parameters, scaled to a typical 70 kg adult (i.e. allometric scaling).¹⁹ The effect of categorical and continuous covariates with biological plausibility was assessed with a stepwise regression using the −2 × log-likelihood as objective function value (OFV). In the forward selection phase, each covariate was added to the base model individually and the inclusion criterion was a reduction greater than 3.84 (P < 0.05) of OFV. During backward elimination, each covariate was systematically removed from the model, preserving those for which exclusion increased the OFV in at least 6.63 (P < 0.01). The final model retained covariate effects that could be estimated with a reasonable relative standard error (RSE), ensuring that the model’s parameter estimates are reliable and robust.

We incorporated a maturation factor (MF) to account for developmental changes in renal clearance, using a Hill function based on PMA:

$$\mathrm{MF}={\displaystyle \frac{\mathrm{PM}{\mathrm{A}}^{\mathrm{Hill}}}{\mathrm{PM}{\mathrm{A}}^{\mathrm{Hill}}+\mathrm{PMA}{50}^{\mathrm{Hill}}}}$$

where PMA50 is the PMA at which 50% of the adult clearance is reached, while the Hill coefficient characterizes the slope of the developmental profile. For those whose gestational age was not documented or for patients over 3 years old, gestational age was assumed to be 40 weeks.

The final model was assessed through visual inspections of goodness-of-fit plots and internally validated using the bootstrap method with the R package Rsmlx (R Speaks Monolix) within RStudio. A total of 1000 replicate datasets were generated through random sampling with replacement from the original dataset, and the 95% confidence intervals for the parameters estimates were computed.

Simulation

The final model was implemented in Simulx software (2023R1 Version, Lixoft, Antony, France) for Monte Carlo simulations.

A simulation dataset with 4800 patients, evenly distributed among infants (1 month to 2 years), children (2–12 years), adolescents (12–18 years) and young adults (18–30 years), was generated by random selection from the CDC-NHANES demographic database.²⁰ Within each age group, eGFRs were categorized as kidney impairment, normal eGFR, or 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.^14,21,22 Additionally, for each age group and eGFR category, we alternated the cumulative percentage of fluid balance dichotomously between 0% and 10%.

We assessed the probability of target attainment (PTA) for different meropenem dose regimens: 20 and 40 mg/kg (maximum 2000 mg/dose) every 6 or 8 h, administered with a 30 min or 3 h infusion duration. Additionally, we investigated continuous infusion doses ranging from 30 to 150 mg/kg per day (maximum 6000 mg/day). Our objective was to identify dosing regimens that ensure at least 90% of simulated patients had free concentrations above three distinct pharmacodynamic (PD) targets: 40% fT > MIC, 100% fT > MIC and 100% fT > 4xMIC, using the 2023 CLSI meropenem breakpoints of MIC 1 mg/L for Enterobacteriaceae and MIC 2 mg/L for Pseudomonas aeruginosa.²³

Results

Study population

A total of 48 patients with 304 meropenem plasma concentrations were included in this study. During the data cleaning process, 8 (2%) of the meropenem concentrations were identified as potential data entry errors or resulting from analytical issues. As a result, 296 samples were used for model building, with a median of 5 samples (range 1–24) per patient. Among these, 27 (9.1%) samples were below the limit of quantification.

A summary of demographics and clinical characteristics of the patients is presented in Table 1. Overall median age of the patients was 13.4 years (interquartile range [IQR], 4.1–18.3), but the cohort comprised a mix of 8 infants (17%), 14 children (29%), 13 adolescents (27%) and 13 adults (27%). At the beginning of meropenem therapy, 22 patients (46%) exhibited normal renal function, while 16 patients (35%) had ARC, and 9 patients (19%) had kidney impairment.

Table 1.

Demographics and clinical characteristics of 48 patients receiving meropenem

Variable	Value (n = 48)
Demographic data
Age (years), median (IQR)	13.4 (4.1–18.3)
Infants (1 month to 2 years), n (%)	8 (17%)
Gestational age (weeks), median (IQR)	34.5 (26.5–38.75)
Children (2–12 years), n (%)	14 (29%)
Adolescents (12–18 years), n (%)	13 (27%)
Young adults (18–30 years), n (%)	13 (27%)
Male, n (%)	30 (62%)
Female, n (%)	18 (38%)
Admission body weight (kg), median (IQR)	29 (16.2–55.7)
Height (cm), median (IQR)	128.8 (97.9–162.4)
Clinical data
Serum creatinine (mg/dL), median (IQR)a	0.4 (0.2–0.6)
eGFR (mL/min/1.73 m2), median (IQR)a	133 (95.6–169.2)
Blood urea nitrogen (mg/dL), median (IQR)a	11.5 (7–18)
Serum albumin (g/dL), median (IQR)a	2.8 (2.4–3.2)
Mechanical ventilation, n (%)b	21 (44%)
Vasopressor treatment, n (%)b	16 (33%)
PRISMIII, median (IQR)a	5.5 (3–10)
PIM2, median (IQR)a	−4.8 (−5.1 to −4.3)
PIM3, median (IQR)a	−5.3 (−5.6 to −4.6)
Cumulative % fluid balance
Day 1 (n = 48), median (IQR)	3 (1.1–5.4)
Day 2 (n = 48), median (IQR)	6.7 (2.5–11)
Day 3 (n = 42), median (IQR)	7.1 (3.4–14)
Day 4 (n = 31), median (IQR)	7.7 (4.5–18.5)
Day 5 (n = 30), median (IQR)	10.3 (5.4–20.2)
Day 6 (n = 26), median (IQR)	11.7 (4.9–22.2)
Day 7 (n = 23), median (IQR)	10.9 (7.3–21.7)

IQR, interquartile range; eGFR, estimated glomerular filtration rate.

^aClinical data at admission.

^bAny day during the patient follow-up.

Pharmacokinetic model building

A two-compartment model with first-order elimination best described the data.

Allometric scaling weight was successfully applied for all clearance and volume parameters. Both IIV and IOV were included on clearance and central volume of distribution. The inclusion of random effects on intercompartmental clearance and peripheral volume was not supported by the data, with estimates close to 0 and high RSE. The residual variability was appropriately described by the proportional error model.

After forward inclusion, eGFR, blood urea nitrogen, daily urine output, blood lactate and age had a significant impact on clearance, while lactate, PRISMIII score, PIM2 score, procalcitonin, age, concomitant administration of vasoactive drugs and cumulative percentage of fluid balance impacted central volume. After backward elimination, the removal of the following covariates significantly increased the OFV: eGFR on clearance (ΔOFV of 43.7), cumulative percentage of fluid balance on central volume (ΔOFV of 20.4) and age on central volume (ΔOFV of 20.4). The effect of age on central volume had a high RSE, exceeding 300%, so age was removed from the final model.

Due to the relatively small number of infants in our cohort, we were unable to obtain precise estimates for the PMA50 and Hill coefficient of the maturation factor. Since meropenem is primarily excreted unchanged in the urine, and there is a significant correlation between meropenem clearance and creatinine clearance,²⁴ we assumed the PMA50 and Hill coefficient values to be equal to those representing glomerular filtration rate (GFR) maturation with age, previously described as 47.7 and 3.40, respectively.²⁵ The GFR maturation curve effectively described the meropenem clearance of the studied infants, after normalizing for covariates (Figure 1). The inclusion of the maturation factor improved the model (ΔOFV of −8.0).

Figure 1.

Meropenem clearance of the studied population normalized by 70 kg and estimated glomerular filtration rate (eGFR) of 140 mL/min/1.73 m² as a percentage of a typical adult meropenem clearance versus postmenstrual age. The solid line represents the GFR maturation curve described by Rhodin et al.²⁵ PMA50 is the PMA at which 50% of the adult clearance is reached, while the Hill coefficient characterizes the slope of the developmental profile. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

The final population estimates for meropenem parameters were estimated with good precision (RSE ≤ 20%): clearance of 13.22 L/h/70 kg^0.75, intercompartment clearance of 1.78 L/h/70 kg^0.75, central volume of 16.45 L/70 kg and peripheral volume of 7.66 L/70 kg. The IIV for clearance was 24.3%, with an IOV of 12%. For the central volume, the IIV was 36.1%, and the IOV was 13.1%. Final model equations and parameter estimates are presented in Table 2.

Table 2.

Population pharmacokinetic parameters of final meropenem model and bootstrap results

	Stochastic approximation		Bootstrap estimates (n = 1000)
Parameter	Estimate	RSE (%)	Median	2.5%ile	97.5%ile
Fixed effects
$\mathrm{Cl}={\mathrm{Cl}}_{\mathrm{pop}}\times (\mathrm{WT}/70{)}^{0.75}\times (\mathrm{eGFR}/140{)}^{\mathrm{\beta }}\times {\mathrm{e}}^{\mathrm{\eta }\mathrm{Cl}+\mathrm{\eta }IOV}\times \frac{\mathrm{PM}{\mathrm{A}}^{\mathrm{Hill}}}{\mathrm{PM}{\mathrm{A}}^{\mathrm{Hill}}+\mathrm{PMA}{50}^{\mathrm{Hill}}}$
CL (L/h/70 kg0.75)	13.22	5.28	13.01	11.59	14.55
βeGFR	0.45	14.4	0.43	0.27	0.56
PMA50 (weeks)	47.7	—	—	—	—
Hill	3.4	—	—	—	—
$\mathrm{V}={\mathrm{V}}_{\mathrm{pop}}\times (\mathrm{WT}/70)\times {\mathrm{e}}^{\mathrm{\beta }\times \mathrm{Cum}\text{\%}\mathrm{FB}}\times {\mathrm{e}}^{\mathrm{\eta }\mathrm{V}1+\mathrm{\eta }IOV}$
V1 (L/70 kg)	16.45	12.2	15.56	11.41	20.82
βCum%FB	0.033	21	0.036	0.018	0.053
$\mathrm{Q}={\mathrm{Q}}_{\mathrm{pop}}\times (\mathrm{WT}/70{)}^{0.75}$
Q (L/h/70 kg0.75)	1.78	19.7	1.69	0.97	3.61
$\mathrm{V}2=\mathrm{V}{2}_{\mathrm{pop}}\times (\mathrm{WT}/70)$
V2 (L/70 kg)	7.66	13.6	7.40	4.61	12.01
Random effects
IIV CL	24.3	15.5	23.3	15.1	30.7
IIV V1	36.1	28.8	37.2	16.1	56.9
IOV CL	12.0	15.3	11.0	5.0	17.1
IOV V1	13.1	38	13.1	7.0	24.3
Error model parameter: proportional only
B	0.33	6.65	0.32	0.26	0.37

ηCl and ηV1 represent the random-effect parameters for IIVs, and ηIOV represents the random-effect parameter for IOV. Both IIV and IOV are 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; V1, central volume of distribution; Q, intercompartmental clearance; V2, peripheral volume; eGFR, estimated glomerular filtration rate; PMA50, maturation half-life; Hill, Hill coefficient; Cum%FB, cumulative percentage of fluid balance; WT, body weight; RSE, relative standard error; IIV, interindividual variability; IOV, inter-occasion variability.

Model diagnosis

Figure 2 displays the goodness-of-fit plots for the final model, indicating no evident trends or model misspecification. In the observed versus predicted plots, the scatter of data points is symmetrically clustered around the line of identity. The residuals are well distributed around zero, without any discernible patterns, with most points within the range of −2 and 2. The prediction-corrected visual predictive check showed that most observed concentrations were within the predicted intervals (Figure 3). Bootstrap results confirmed model’s stability and consistency in parameter estimates (Table 2).

Figure 2.

Diagnostic goodness-of-fit plots of the final meropenem model. (a) Population weighted residuals (PWRES) versus time in hours. (b) Population weighted residuals (PWRES) versus population-predicted concentrations. (c) Observed concentrations versus population-predicted concentrations. (d) Observed concentrations versus individual-predicted concentrations. (e) Histogram of the distribution of the normalized prediction distribution errors (NPDE), with the density of the standard Gaussian distribution overlaid. Concentrations are expressed in milligrams per litre (mg/L).

Figure 3.

Prediction-corrected visual predictive check (pcVPC) for meropenem concentrations. The dots are observed concentrations, solid lines represent the median, 2.5th and 97.5th percentile of the observed values, and shaded areas represent the spread of 95% prediction intervals calculated from simulations (n = 1000). This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

Monte Carlo simulations

Using Monte Carlo simulations, we successfully investigated different meropenem dosing regimens for patients with normal renal function, kidney impairment, ARC and varying levels of fluid balance (Tables S1–S3, available as Supplementary data at JAC Online). Table 3 summarizes the suggested dosing regimens according to age group, renal function and target, based on the lowest daily doses required to achieve at least 90% PTA. We suggested continuous infusion only in cases where none of the tested intermittent infusion regimens achieved the 90% PTA threshold.

Table 3.

Meropenem maintenance dosing recommendations according to age group, renal function and target, based on the lowest daily doses required to achieve at least 90% target attainment in 4800 simulated paediatric intensive care unit patients

		40% fT > MIC		100% fT > MIC		100% fT > 4 × MIC
		MIC = 1 mg/L	MIC = 2 mg/L	MIC = 1 mg/L	MIC = 2 mg/L	MIC = 1 mg/L	MIC = 2 mg/L
Infants (1 month to 2 years)	KI	20 mg/kg q8h TINF 0.5h	20 mg/kg q8h TINF 0.5h	20 mg/kg q8h TINF 3h	20 mg/kg q6h TINF 3h	30 mg/kg/day CI	60 mg/kg/day CI
	Normal	20 mg/kg q8h TINF 0.5h	20 mg/kg q8h TINF 3h	40 mg/kg q6h TINF 0.5h	30 mg/kg/day CI	60 mg/kg/day CI	90 mg/kg/day CI
	ARC	20 mg/kg q8h TINF 3h	20 mg/kg q8h TINF 3h	30 mg/kg/day CI	60 mg/kg/day CI	90 mg/kg/day CI	120 mg/kg/day CI
Children (2–12 years)	KI	20 mg/kg q8h TINF 0.5h	20 mg/kg q8h TINF 0.5h	20 mg/kg q8h TINF 3h	20 mg/kg q6h TINF 3h	60 mg/kg/day CI	60 mg/kg/day CI
	Normal	20 mg/kg q8h TINF 0.5h	20 mg/kg q8h TINF 3h	40 mg/kg q6h TINF 0.5h	30 mg/kg/day CI	60 mg/kg/day CI	90 mg/kg/day CI
	ARC	20 mg/kg q8h TINF 3h	20 mg/kg q8h TINF 3h	30 mg/kg/day CI	30 mg/kg/day CI	60 mg/kg/day CI	120 mg/kg/day CI
Adolescents (12–18 years)	KI	20 mg/kg q8h TINF 0.5h	20 mg/kg q8h TINF 0.5h	20 mg/kg q8h TINF 0.5h	20 mg/kg q6h TINF 3h	30 mg/kg/day CI	60 mg/kg/day CI
	Normal	20 mg/kg q8h TINF 0.5h	20 mg/kg q8h TINF 0.5h	20 mg/kg q6h TINF 3h	40 mg/kg q6h TINF 3h	60 mg/kg/day CI	120 mg/kg/day CI
	ARC	20 mg/kg q8h TINF 0.5h	20 mg/kg q8h TINF 3h	40 mg/kg q6h TINF 3h	60 mg/kg/day CI	60 mg/kg/day CI	120 mg/kg/day CI
Young adults (18–30 years)	KI	20 mg/kg q8h TINF 0.5h	20 mg/kg q8h TINF 0.5h	20 mg/kg q8h TINF 0.5h	20 mg/kg q6h TINF 3h	30 mg/kg/day CI	60 mg/kg/day CI
	Normal	20 mg/kg q8h TINF 0.5h	20 mg/kg q8h TINF 0.5h	20 mg/kg q6h TINF 3h	30 mg/kg/day CI	60 mg/kg/day CI	90 mg/kg/day CI
	ARC	20 mg/kg q8h TINF 0.5h	20 mg/kg q8h TINF 3h	30 mg/kg/day CI	30 mg/kg/day CI	60 mg/kg/day CI	120 mg/kg/day CI

Continuous infusion regimens are listed only if none of the tested intermittent regimens achieved 90% target attainment.

KI, kidney impairment; ARC, augmented renal clearance; MIC, minimum inhibitory concentration; % fT > MIC, percentage of time above minimum inhibitory concentration; T_inf, duration of infusion; CI, continuous infusion.

When considering the PD target of 40% fT > MIC, the commonly used dose of 20 mg/kg every 8 h over 30 min was insufficient for 90% PTA against susceptible pathogens (MIC ≤ 2 mg/L) for infants and children with normal renal function, with fluid balance of 0%, and for those with ARC in all age groups. Extending the infusion to 3 h, while maintaining the dosing regimen at 20 mg/kg every 8 h, achieved 90% PTA for all age groups with varying renal function.

If 100% fT > MIC is desired, only continuous infusion regimens had PTA greater than 90% against susceptible pathogens (MIC ≤ 2 mg/L) in patients with normal renal function or ARC, except for adolescents with normal renal function, who had acceptable PTA with 40 mg/kg q6h with a 3 h extended infusion. Continuous infusion regimens ranging from 30 to 60 mg/kg/day resulted in concentrations above MIC of 2 mg/L throughout the entire dosing interval for all patients. For patients with kidney impairment, an alternative approach involves increasing the frequency of administration and extending the infusion duration to 20 mg/kg every 6 h with a 3 h infusion instead of continuous infusion. For a target of 100% fT > 4 × MIC 2 mg/L, all patients needed continuous infusion, ranging from 60 mg/kg/day to 120 mg/kg/day.

For intermittent regimens, patients with a cumulative percentage of fluid balance of 10% had an increased likelihood of achieving the therapeutic target compared to those with a fluid balance of 0, by up to 25%.

Discussion

We describe a population pharmacokinetic model of meropenem developed using real-world data from critically ill patients in the PICU. In our final model, the mean values of meropenem clearance and central volume of distribution were 13.22 L/h/70^0.75 kg and 16.45 L/70 kg, respectively. Cies et al.²⁶ also studied meropenem in critically ill children and reported higher median clearance of 17.3 L/h/70 kg^0.75 and median central volume of 31.5 L/70 kg, likely due to their focus on patients with systemic inflammatory response syndrome, as well as their exclusion of patients with acute or chronic renal failure. Rapp et al., who investigated critically ill children with different renal functions, reported meropenem clearance of 6.82 L/h/70 kg^0.75, and central volume of 40.6 L/70 kg, while Saito et al. found even higher values for clearance and central volume: 19.2 L/h/70 kg^0.75 and 34.3 L/70 kg, respectively.^27,28 Although their inclusion of patients receiving ECMO and CRRT makes direct parameter comparison challenging but likely contributes to the higher volume of distribution, the wide range of pharmacokinetic parameters highlights the complexity of meropenem pharmacokinetics and the need to individualize dosing regimens.

Allometric body weight scaling is a standard practice in paediatrics to scale pharmacokinetic parameters according to body size.²⁹ Besides that, the inclusion of a maturation function accounts for the age-related increases in clearance.³⁰ Tod et al.³⁰ suggest that when the factor cannot be estimated, the value may be fixed according to prior knowledge. We assumed that the maturation factor for meropenem clearance is similar to that reflecting GFR maturation, and this approach has been used for other renally eliminated antibiotics.^31,32 Similarly to the study by Padari et al.,³³ who also incorporated GFR maturation parameters in their meropenem model, the inclusion of the maturation factor improved the model and was retained in the final model.

As expected, eGFR significantly improved the model as a covariate on clearance, since meropenem is mainly eliminated by the kidneys. eGFR or other markers of renal function are the most common covariates in paediatric and adult meropenem models.^27,34 We also observed a significant impact of the cumulative percentage of fluid balance on the central volume of distribution. Critically ill patients frequently develop substantial fluid overload, which is associated with increased morbidity and mortality.³⁵ Although it is well known that volume status can influence drug distribution, especially for hydrophilic antibiotics such as β-lactams,³⁶ this factor is infrequently integrated into population pharmacokinetic models. In simulations, the presence of positive fluid balance did not compromise target attainment of meropenem, even slightly increasing the likelihood of achieving the therapeutic target for intermittent dosing regimens. Nehus et al.³⁷ had similar findings in children receiving CRRT, observing that fluid overload tended to increase meropenem PTA.

The results of the Monte Carlo simulations suggested that the commonly used dose regimen of 20–40 mg/kg every 8 h administered with a 30 min infusion is insufficient to ensure target attainment of 40% fT > MIC against susceptible Gram-negative bacteria (MIC ≤ 2 mg/L), especially in patients with preserved renal function or ARC. Previously published studies have highlighted the risk of meropenem target attainment failure in patients with ARC, which is highly prevalent in PICU mostly due to inflammatory response and increased cardiac output.^22,38,39 Therefore, the optimal dose suggested by our simulations is 20 mg/kg every 8 h via 3 h extended infusion for this least stringent target. Previously published simulations performed based on population PK analyses of various β-lactam agents have consistently found that use of extended infusion increases the PTA in children compared with rapid infusions.⁴⁰ Although clinical studies supporting routine use of extended infusion of meropenem in children are lacking, a recent meta-analysis concluded that prolonged β-lactam antibiotic infusions are associated with increased clinical cure rates and reduced risk of ICU mortality among severely ill adult patients.⁴¹ Rapid infusions demonstrated favourable PTA only in patients with kidney impairment, as their reduced meropenem clearance leads to sustained concentrations above MIC for longer.

If more stringent PD targets are considered (100% fT > 1–4 × MIC), continuous infusion regimens are necessary to ensure target attainment against susceptible pathogens (MIC ≤ 2 mg/L). For most patients, a continuous infusion of 30–60 mg/kg/day was sufficient to result in concentrations above MIC for susceptible pathogens throughout the entire dosing interval. However, a recent international consensus recommended at least 100% fT > 4 × MIC when administering β-lactam as continuous infusions.⁴⁰ In this case, higher doses of 90–120 mg/kg/day are necessary for optimal PTA. Simulations performed by Rapp et al.²⁷ suggested that continuous infusion is the best scheme for patients with normal and augmented renal clearance, ranging from 60 to 120 mg/kg/day. Cies et al.²⁶ suggested even higher daily doses of 120–160 mg/kg/day as continuous infusion, which is expected due to higher meropenem clearance in their population.

Further studies regarding the clinical benefits, feasibility and safety of continuous infusion regimens in paediatrics are warranted. Most studies suggesting the use of continuous infusion of β-lactams in children are simulation-based studies.⁴² The BLING III trial, the largest international randomized clinical trial conducted in adult ICU patients receiving either continuous (n = 3498) or intermittent (n = 3533) infusion of β-lactam antibiotics, showed a higher rate of clinical cure in the continuous infusion group (55.7% versus 50%, P < 0.05). Although the 90-day mortality difference was not statistically significant, the confidence interval indicated a potential clinically important benefit with continuous infusions.⁴³

This study has some limitations. First, as an inherent limitation of the scavenged sampling approach, it is possible that some samples used for meropenem measurement were drawn from the same line where meropenem was infused. Although a small amount of blood is usually wasted prior to obtaining blood for clinical labs by institutional protocol, there may be potential inaccurately elevated concentration measurements. Also due to the sparse opportunistic sampling, we were not able to describe the IIV of intercompartmental clearance and peripheral volume of distribution. Although the inclusion of a maturation factor aligns with existing literature and improved the model, assuming that the maturation factor for meropenem clearance is the same as GFR may oversimplify the complex process of drug elimination and not fully capture the intricacies of meropenem clearance changes during development. Finally, since we did not include patients receiving continuous renal replacement or extracorporeal membrane oxygenation therapy, which are known to significantly alter meropenem pharmacokinetics, we are unable to provide dosing recommendations for these vulnerable subgroups.

Conclusions

We successfully developed a two-compartment population pharmacokinetic model of meropenem using real-world data from critically ill paediatric and young adult patients with an opportunistic sampling strategy. Allometric body weight scaling was included to account for body size differences, and a maturation function was incorporated to accommodate developmental changes in renal clearance. Incorporating eGFR as a covariate on clearance and cumulative percentage of fluid balance on the central volume of distribution significantly improved our model. Monte Carlo simulations suggested that the most effective dosing regimen is 20 mg/kg every 8 h with a 3 h infusion for 40% fT > MIC. If more stringent PD targets are considered, continuous infusion regimens ensure optimal target attainment for most patients, ranging from 30 mg/kg/day to 120 mg/kg/day.

Funding

This work was supported by the National Institutes of Health (NIH), including funding from the National Institute of General Medical Sciences (NIGMS) under an R35 award (R35GM14670), the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) T32 Training Program in Pediatric Clinical and Developmental Pharmacology (T32HD069054) and the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) (T32DK007695).

Transparency declarations

None to declare.

Supplementary data

Tables S1–S3 are available as Supplementary data at JAC Online.