Augmented Renal Clearance Implies a Need for Increased Amoxicillin-Clavulanic Acid Dosing in Critically Ill Children

Abstract

There is little data available to guide amoxicillin-clavulanic acid dosing in critically ill children. The primary objective of this study was to investigate the pharmacokinetics of both compounds in this pediatric subpopulation. Patients admitted to the pediatric intensive care unit (ICU) in whom intravenous amoxicillin-clavulanic acid was indicated (25 to 35 mg/kg of body weight every 6 h) were enrolled. Population pharmacokinetic analysis was conducted, and the clinical outcome was documented. A total of 325 and 151 blood samples were collected from 50 patients (median age, 2.58 years; age range, 1 month to 15 years) treated with amoxicillin and clavulanic acid, respectively. A three-compartment model for amoxicillin and a two-compartment model for clavulanic acid best described the data, in which allometric weight scaling and maturation functions were added a priori to scale for size and age. In addition, plasma cystatin C and concomitant treatment with vasopressors were identified to have a significant influence on amoxicillin clearance. The typical population values of clearance for amoxicillin and clavulanic acid were 17.97 liters/h/70 kg and 12.20 liters/h/70 kg, respectively. In 32% of the treated patients, amoxicillin-clavulanic acid therapy was stopped prematurely due to clinical failure, and the patient was switched to broader-spectrum antibiotic treatment. Monte Carlo simulations demonstrated that four-hourly dosing of 25 mg/kg was required to achieve the therapeutic target for both amoxicillin and clavulanic acid. For patients with augmented renal function, a 1-h infusion was preferable to bolus dosing. Current published dosing regimens result in subtherapeutic concentrations in the early period of sepsis due to augmented renal clearance, which risks clinical failure in critically ill children, and therefore need to be updated. (This study has been registered at Clinicaltrials.gov as an observational study [NCT02456974].)

INTRODUCTION

Appropriate antibiotic treatment is a cornerstone in the pharmacological treatment of critically ill children. Pediatric sepsis and septic shock reportedly affect 30% of children admitted to pediatric intensive care units (ICUs), with a 25% mortality rate (1). Due to their broad antimicrobial spectrum and relatively low toxicity, β-lactam antibiotics such as amoxicillin-clavulanic acid are commonly used in pediatric critical care for treating community-acquired infections (2). Typical indications include community-acquired pneumonia, skin, soft tissue, and abdominal infections.

During childhood, many developmental changes occur, which influence both drug exposure and drug response (3). Moreover, pathophysiological changes during critical illness frequently affect pharmacokinetics (PK) and pharmacodynamics (PD) (4–6). To date, only one report on the pharmacokinetics of amoxicillin-clavulanic acid in a limited number of critically ill children older than 2 years (n = 15 patients) is available (7).

Broader-spectrum and newer antibiotics are now being studied more extensively in this patient population, which may predispose clinicians toward the use of such agents (8). Therefore, research on more-targeted and well-established therapies like amoxicillin-clavulanic acid is highly relevant.

The primary aims of this study were (i) to investigate the pharmacokinetics of intravenous amoxicillin-clavulanic acid in critically ill infants and children and (ii) to evaluate the efficiency of current and alternative dosing regimens in this population.

MATERIALS AND METHODS

Study design.

A prospective, open-label, pharmacokinetic study was conducted at the pediatric ICU of the Ghent University Hospital, Ghent, Belgium, between May 2012 and December 2013. Patients between 1 month and 15 years of age admitted to the pediatric ICU were included in whom treatment with intravenous amoxicillin-clavulanic acid was the standard of care. Patients were excluded if they required an extracorporeal circuit or did not have arterial or intravenous access other than the drug infusion line available for blood sampling. This research was conducted in accordance with the guidelines of the Declaration of Helsinki and was approved by the institutional Ethics Committee (EC/2012/172). Written informed consent was obtained from the parents or legal representatives and also from the patients if they were older than 12 years.

Drug dosing and administration.

Amoxicillin-clavulanic acid (Augmentin P 500/50 mg and Augmentin 1,000/200 mg; GlaxoSmithKline, Genval, Belgium; Amoxiclav Sandoz 1,000/200 mg; Sandoz NV, Vilvoorde, Belgium) was prescribed in a dose range of 25 to 35 mg amoxicillin per kilogram of body weight (maximum 1,000 mg) every 6 h and administered intravenously over 5 to 30 min using a calibrated syringe driver, according to current dosing guidelines (9). According to a standardized procedure, infusion lines were flushed with normal saline immediately after drug administration with a minimum of twice the dead space volume.

Blood sampling.

Serial blood samples were obtained from the first and/or assumed steady-state doses from an indwelling catheter other than the drug infusion line (median of four blood samples per dose). The total number of samples collected (per patient) was limited by the predefined total maximum blood volume permitted for PK sampling per individual patient, defined as 2.4 ml/kg of body weight (10). A full sampling scheme per dose typically included a sample just before dosing (t = 0), a sample immediately after dosing and flush, a distribution sample between 5 and 70 min after the start of the drug infusion, a mid-dose-interval sample 3 h after the drug infusion start time, and a trough sample just prior to the next dose. All samples were immediately transferred on ice to the chemistry laboratory and centrifuged (8 min, 1,885 × g) after which the resulting plasma was frozen at −80°C for a maximum of 3 months before assay.

Drug and biochemical assays.

Total plasma amoxicillin and clavulanic acid concentrations were quantified simultaneously using a validated ultraperformance ultrahigh-pressure liquid chromatography (UPLC)-tandem mass spectrometry method (11). The lower limit of quantification (LLOQ) was 0.5 mg/liter for both compounds, and the imprecision was <15% at all levels. For the first 24 patients, only the amoxicillin compound was quantified. Creatinine was measured in serum (Scr) and urine (Ucr) using the rate-blanked compensated Jaffe technique (Modular P and Cobas 6000, Roche Diagnostics GmbH, Mannheim, Germany). Twenty-four-hour creatinine clearance (CrCL) was calculated using the following formula: CrCL = urine volume × Ucr/(1,440 × Scr) (whenever a urinary catheter and 24-h urine collection were available). Plasma cystatin C (CysC) was measured using the N Latex cystatin C assay on the Behring nephelometer II (Siemens Healthcare Diagnostics Products GmbH, Marburg, Germany) (intraassay coefficient of variation [CV], 1.4%; interassay CV, 5.4%) and was standardized according to the ERM-DA471/IFCC reference material (12).

Pharmacokinetic analysis.

Amoxicillin-clavulanic acid pharmacokinetics were evaluated using the nonlinear mixed-effects modeling software NONMEM version 7.3 (ICON PLC, Ellicott City, MD). The first-order conditional estimation method with the interaction option (FOCE-I) was used to estimate PK parameters and variability. R (version 3.0.2) and PsN (version 5.18.2) tools were used for pre- and postprocessing. One-, two-, and three-compartment linear models were tested to fit plasma concentrations of both compounds independently using the NONMEM library ADVAN subroutines (13). Following this, a simultaneous fit of concentrations of both compounds was evaluated; when both analyte measurements were obtained from the same sample, correlations in residual error were handled using the L2 method in NONMEM (13). The L2 data item is used to group observations within an individual to indicate that there may be a degree of correlation in the residual variability (usually residual variability assumes all observations are independent). All clearance parameters were scaled a priori using an allometric weight (WT) approach with a fixed exponent of 0.75 in combination with a Hill model using postmenstrual age (PMA) to describe the maturation process on clearance (14):

$${\text{CL}}_i={\text{CL}}_{\text{pop}}\times {\left(\frac{\text{WT}}{{\text{WT}}_{\text{med}}}\right)}^{0.75}\times \left(\frac{{\text{PMA}}^{\text{Hill}}}{{{\text{TM}}_{50}}^{\text{Hill}}+{\text{PMA}}^{\text{Hill}}}\right)$$

where CL_i is the individual clearance, CL_pop is the population clearance, WT_med is the median weight, PMA is postmenstrual age, Hill is the Hill coefficient, and TM₅₀ is the maturation half-life. All volume of distribution parameters were scaled a priori with linear weight. Between-subject variability (BSV) was described using an exponential error model, and a proportional error model was used to describe residual variability. Between-occasion variability (BOV) was tested on clearance (CL). For model evaluation, decrease in objective function value (OFV), plots of observed versus population predicted concentrations, observed versus individual predicted concentrations, conditional weighted residuals (CWRES) versus time after dose, and CWRES versus population predicted concentration were utilized. Parameter estimates were compared using three different methods for handling data below quantification limit (BQL): omitting BQL samples, setting values to half the limit of quantification (LOQ/2), and the M3 method, enabling estimation of the likelihood of BQL measurements being real BQL data (15). While body weight and age were included a priori as described above, CysC was then further tested, since it is known that amoxicillin and clavulanic acid are renally cleared (16). After this, a stepwise covariate model (SCM) building exercise was performed with a forward inclusion criterion of P < 0.01 and backwards elimination criterion of P < 0.005. The following covariates were tested in the SCM: primary reason for admission, measures of organ function, and patient severity of illness as described by the PELOD (pediatric logistic organ dysfunction) score, PRISM II (pediatric risk of mortality) score (17, 18), presence of surgery, presence of mechanical ventilation, cotreatment with vasopressors and nephrotoxic medications (aminoglycosides, glycopeptides, diuretics, angiotensin-converting enzyme [ACE] inhibitors, nonsteroidal anti-inflammatory drugs, tacrolimus, cyclosporin, methotrexate), fluid resuscitation (>60 ml/kg per 24 h), type of catheter used for drug administration and blood sampling, and C-reactive protein (CRP). Given that 35% of Scr samples were BQL, Scr and CrCL could not be tested as covariates on drug clearances. The final population model was evaluated in two ways: a nonparametric bootstrap sampling procedure (n = 1,000) and a visual predictive check (VPC) (n = 1,000).

Clinical outcome assessment.

Clinical failure in our study population was defined as premature termination of amoxicillin-clavulanic acid with change of antibiotic therapy or additional antibiotics commenced within 48 h of completion of therapy. Duration of antibiotic therapy was dependent on the type of infection and the patient's clinical evolution and was determined by the attending physician.

Assessment of dose-exposure relationship.

Monte Carlo simulations (n = 1,000 patients) were performed for three dosing regimens (Table 1) (9, 19). Based on these simulations, the fraction of time during which the unbound drug concentration is above the MIC (fT_>MIC) was calculated for the first dose and over the first 48 h of treatment. The target efficacy exposure was defined as 40% fT_>MIC (20), and a target MIC of 8 mg/liter for amoxicillin was chosen as the worst-case scenario, according to the European Committee on Antimicrobial Susceptibility Testing (EUCAST) and Clinical and Laboratory Standards Institute (CLSI) clinical breakpoints for Escherichia coli (21). As the EUCAST clinical breakpoint for amoxicillin was determined using a fixed concentration of 2 mg/liter clavulanic acid, this was chosen as the target concentration for clavulanic acid (note that the CLSI susceptibility testing concentration was 4 mg/liter). The mean protein binding values of amoxicillin and clavulanic acid are 18 and 25%, respectively (16), and these values were used to simulate unbound concentration.

TABLE 1.

Simulated dosing scenarios

Dosing regimena	Drug formulary
25 mg/kg q12h for children between 1 and 3 mo old	British National Formulary for Children (19)
25 mg/kg q8h for children older than 3 mo	British National Formulary for Children (19)
25 mg/kg q6hb	Sanford Guide for Antimicrobial Therapy (9)
25 mg/kg q4hc

^aThe dosing regimen was based on the amoxicillin component and a fixed ratio of amoxicillin-clavulanic acid of 5:1. All simulations included a bolus and 1-h infusion regimen. Abbreviations: q12h, every 12 h; q8h, every 8 h; q6h, every 6 h; q4h, every 4 h.

^bStudy dosing regimen.

^cNewly tested dosing regimen.

RESULTS

A total of 50 patients were included in this study; demographic, clinical, and treatment characteristics are summarized in Table 2. Patients younger than 2 years old accounted for 44% of the study population (n = 22). A total of 325 amoxicillin and 151 clavulanic acid concentrations in plasma were available for population PK analysis. A three-compartment model for amoxicillin and a two-compartment model for clavulanic acid best described the data (Fig. 1). BSV was supported for amoxicillin and clavulanic acid central clearance and volume only. Central volume values for amoxicillin and clavulanic acid were highly correlated (R² > 0.99), so a single random effect with a scaling factor for clavulanic acid was used (Table 3). The correlation between amoxicillin and clavulanic acid clearance was also estimated. BOV was supported for amoxicillin clearance and clavulanic acid clearance. Since these values were highly correlated, a single random effect with scaling factor was also used (Table 3). The Hill coefficient on the maturation model was estimated to be close to 1, and no significant change in OFV was noted when the coefficient was fixed at 1. When CysC was included as a covariate on amoxicillin and clavulanic acid clearance, it significantly improved the fit of the model with drops in OFV of 6.63 points for amoxicillin and 6.61 points for clavulanic acid. Model building to this stage was initially undertaken with the exclusion of BQL data, which comprised 9% of the observations. BQL data were then included to compare parameter estimates as described above. Given that no significant parameter estimate differences were observed between BQL handling methods and the run time for the M3 method was more than 3-fold longer than the LOQ/2 method, the latter method was chosen for handling BQL. This model, with mechanistic covariates, was then taken forward to the SCM, and vasopressor treatment on amoxicillin clearance further significantly improved the model fit for amoxicillin and therefore was retained in the final model. The impact of covariates is illustrated in Fig. 2. The population PK parameter estimates and their precision are summarized in Table 3. The VPC plots are presented in Fig. 3; the 5th, 50th, and 95th percentiles of the predicted concentrations closely follow the percentiles of the observed data, suggesting a good model fit in both cases.

TABLE 2.

Demographic, clinical, and treatment characteristics

Characteristica	Value for characteristicb
Sex
Male	30 (60)
Female	20 (40)
Age (yr)	2.58 (0.08–15)
Wt (kg)	14.4 (4.07–65)
Total length of ICU stay (days)	9.5 (3–72)
PRISM II score	6.5 (0–32)
Primary reason for ICU admission
Postoperative	16 (32)
Respiratory	10 (20)
Gastrointestinal	10 (20)
Neurologic	7 (14)
Cardiovascular	6 (12)
Other	1 (2)
Reason for antibiotic treatment
Treatment of infection	33 (66)
Postoperative prophylaxis	17 (44)
Mechanical ventilationc	29 (58)
Vasopressor treatmentc	16 (32)
PELOD scored	1 (0–31)
Serum creatinine (mg/dl)d	0.21(<0.17–1.89)
Plasma cystatin Cd,e (mg/liter)	0.63 (0.33–1.23)
Serum CRPd (mg/liter)	5.4 (0.40–28.79)

^aAbbreviations: ICU, intensive care unit; PRISM, pediatric risk of mortality; PELOD, pediatric logistic organ dysfunction; CRP, C-reactive protein.

^bValues are median (range) or number of patients (percentage of total number of patients).

^cDuring ICU stay.

^dAt day(s) of sampling.

^eBased on values from 49 patients.

FIG 1.

Goodness-of-fit diagnostic plots for amoxicillin and clavulanic acid. The plots show observations versus population predictions and individual predictions and conditional weighted residuals (CWRES) versus time after dose and population predictions. In the observation versus prediction plots, a line of identity (black solid line) and a Loess smooth line (red solid line) were included as a reference. In the CWRES plots, dashed lines at +2 and −2 standard deviations from the mean (solid line) were included to indicate the expected region of approximately 95% of the data.

TABLE 3.

Population pharmacokinetic estimates of amoxicillin-clavulanic acid

Parametera	Estimate	Median bootstrap estimate	2.5th percentile from bootstrap estimateb (n = 1,000)	95th percentile from bootstrap estimateb (n = 1,000)
Amoxicillin
CL (liters/h/70 kg)	17.97	17.83	15.33	21.30
V1 (liters/70 kg)	9.07	9.00	6.41	11.66
V2 (liters/70 kg)	5.43	5.75	3.57	13.45
V3 (liters/70 kg)	11.24	11.00	7.01	13.78
Q1 (liters/h/70 kg)	35.88	34.73	12.03	60.09
Q2 (liters/h/70 kg)	5.52	5.36	1.49	8.03
θ in (CYSC/MCYSC)Θ	−0.54	−0.54	−1.01	−0.14
θ in COVVASO = (1 + θ)	−0.18	−0.18	−0.28	−0.07
BSV on CL (% CV)	18.9	18.1	8.10	28.6
BSV on V1 (% CV)	48.6	46.8	30.9	66.7
BOV on CL (% CV)	14.6	13.8	7.70	20.2
Residual error (% CV)	28.7	28.4	23.4	33.1
Clavulanic acid
CLclav (liters/h/70 kg)	12.20	12.09	10.54	14.55
V1clav (liters/70 kg)	11.60	11.39	8.42	13.76
V2clav (liters/70 kg)	9.85	10.22	8.05	13.87
Q1clav (liters/h/70 kg)	6.22	6.81	3.94	23.9
θ in CYSCOV = (CYSC/MCYSC)Θ	−0.37	−0.36	−0.85	−0.08
BSV on CLclav (% CV)	15.6	15.6	7.8	22.6
θ in BSV on V1clav = θ*BSV on V1	0.77	0.77	0.40	0.97
θ in BOV on CLclav = θ*BOV on CL	0.52	0.48	−0.02	1.34
Residual error (% CV)	35	34	27	40
Amoxicillin-clavulanic acid
Hill coefficientc	1	1	1	1
TM50 (wk)	39.90	39.10	20.66	64.29

^aAbbreviations: CL, clearance; clav, clavulanic acid; V1, central volume of distribution; V2 and V3, peripheral volumes of distribution; Q1 and Q2, intercompartmental clearances; θ, model parameter in NONMEM code; CYSCOV, cystatin C covariate on clearance; CYSC, plasma cystatin C value; MCYSC, median plasma cystatin C value; COVVASO, vasopressor covariate on clearance (θ = 0 if no coadministration with vasopressors); BSV, between-subject variability; BOV, between-occasion variability; CV, coefficient of variation; TM₅₀, maturation half-life.

^bNonparametric bootstrap of 953 successful runs.

^cFixed value.

FIG 2.

Impact of covariates in the final model. The left-hand plot shows weight- and cystatin C-standardized amoxicillin clearance with age split by whether patients were on vasopressors. The right-hand plot shows weight- and age-standardized amoxicillin clearance plotted against measured cystatin C split by whether patients were on vasopressors. Note there is more than one clearance value per subject, since between-occasion variability was included. Open circles, patients not on vasopressors; open squares, patients on vasopressors. Solid line, population predicted values if the patients were not on vasopressors; broken line, population predicted values if the patients were on vasopressors.

FIG 3.

Stratified visual predictive check for amoxicillin (CLAV = 0) and clavulanic acid (CLAV = 1). The gray shaded areas show the 95% confidence intervals of the simulated 5th, 50th, and 95th percentiles. The lines show the 5th, 50th and 95th percentiles of raw data.

Pathogens were grown in only 50% of patients and in 63.6% of patients treated for infection. The clinical failure rate in patients receiving amoxicillin-clavulanic acid treatment or prophylaxis was 32%; in those patients treated for infection alone, it was 34.4%. The main pathogens identified in patients with clinical failure were Pseudomonas aeruginosa (37.5%) (after which amoxicillin-clavulanic acid was switched to piperacillin-tazobactam) and Enterobacteriaceae (25%). No pathogen could be identified in 31.3% of patients.

Probability of target attainment for amoxicillin against MIC after the first dose are presented in Fig. 4. Since no significant drug accumulation was seen, the results from the first 48 h of treatment were similar (data not shown). When doses were prescribed as a bolus according to the British National Formulary for Children (BNFc) dosing regimen, the Sanford Guide dosing regimen, or the four-hourly dosing regimen (Table 1), the median target attainment values for clavulanic acid after the first dose were 48%, 66%, and 96%, respectively; when the doses were given as a 1-h infusion, the median target attainment values were 53%, 73%, and 99%, respectively.

FIG 4.

Probability of target attainment (PTA) (n = 1,000 patients) for amoxicillin according to the dosing regimen, presence/absence of vasopressor therapy, and plasma cystatin C value. The three simulated dosing regimens were as follows: (i) 25 mg/kg every 12 h if under 3 months of age, otherwise every 8 h (British National Formulary for Children [BNF-C] [19]), (ii) 25 mg/kg every 6 h (Sanford Guide for Antimicrobial Therapy [9]), (iii) 25 mg/kg every 4 h (alternative dosing regimen) for bolus administration and 1-h infusion. Amoxicillin target was defined as 40% of time above a MIC of 8 mg/liter.

DISCUSSION

This is, to our knowledge, the first report characterizing amoxicillin-clavulanic acid disposition in (critically ill) children using a population PK-PD modeling approach. Besides growth and maturation, renal function was found to be a significant covariate on amoxicillin and clavulanic acid clearance. This finding is in concordance with data in critically ill adults showing both clearances to be proportional to CrCL (22). As previously discussed, we were not able to test Scr or CrCL as a covariate on drug clearance. However, the use of Scr and CrCL as markers for renal function have some major disadvantages, especially in children, as they are sensitive to changes in age, muscle mass, feeding, and disease status (23, 24). Moreover, in younger children, Scr is generally underestimated when using the standardized Jaffe analysis method (25). Finally, creatinine undergoes tubular secretion leading to an overestimation of the glomerular filtration rate (GFR). CysC, a newer endogenous renal biomarker, was previously shown to be superior to plasma creatinine in estimating renal function in critically ill children (26, 27). Second, it has been shown that CysC is a good renal biomarker in children with sepsis (28) and that CysC-based GFR estimations remain accurate in children presenting with hyperfiltration (in contrast to Scr-based formulae) (24). Finally, CysC was found to better predict elimination of renally excreted drugs in adults compared to Scr or CrCL (29–32). This is only the second report to identify this new biomarker as a predictor of renal drug clearance in children at the expense of Scr or CrCL (33). Treatment with vasopressors (mainly norepinephrine; 14/15 patients) was also associated with an 18% decrease in amoxicillin clearance. Although high-quality scientific evidence is lacking, available data for norepinephrine suggest a positive effect on renal blood flow in patients presenting with sepsis (34). One may hypothesize that treatment with vasopressors should be considered an overall parameter reflecting critical illness severity, irrespective of renal function.

The observed population estimate for amoxicillin clearance is much higher than previously reported in critically ill adults (10 liters/h/75 kg) (22) and somewhat variable between doses (BOV 14.6% CV). A high mean clearance was previously reported in 15 seriously ill children aged 2 to 14 years using a noncompartmental (NCA) PK analysis (16.99 liters/h/1.70 m²) (7). The elevated clearance could be explained by a state of “augmented renal clearance” (ARC) in our study's patient population. Although the underlying pathophysiological mechanisms are yet to be revealed (35), this phenomenon has been increasingly investigated in the critically ill adult population, including its impact on the PK and PD of renally cleared antimicrobials (36), but ARC has—to the best of our knowledge—never before been reported in critically ill children. The hypothesis of a “hyperdynamic” status of our study is supported by the fact that a large proportion of measured renal biomarkers was undetectable (Scr) or low (CysC) compared to age-corrected reference values (37). A plausible explanation, besides the analytical challenges for creatinine described above, could be a faster renal clearance of these endogenous compounds. Moreover, as we observed trough concentrations from maintenance doses that remained very low in most patients, we could conclude that no accumulation in steady-state conditions was attained, probably due to the enhanced renal capacity. It is also worth noting that although one would expect a correlation between age and CysC below the age of 2 years (37), we were not able to identify such a relationship in our patient population. It is possible that maturational changes in CysC were masked in our study due to ARC.

Regarding volumes of distribution, the observed population estimate for amoxicillin is similar to what has been reported in critically ill adults (27.4 liters/75 kg) (22) and slightly lower than that found by Jones et al. (7) Similar trends in clearance and volume of distribution were observed for the clavulanic acid compound.

Of particular importance, our data challenge current pediatric dosing recommendations, as they could lead to subtherapeutic treatment in severe infections (Fig. 4). Although our study has a small sample size, clinical outcome data from our study suggest that underdosing of amoxicillin-clavulanic acid could result in clinical failure in severe infections with Enterobacteriaceae (EUCAST and CLSI MIC breakpoints are 8 mg/liter). We have shown that, at minimum, a dosing regimen of 25 mg/kg (based on the amoxicillin component) every 4 h is warranted in those infections (given as a bolus in children with CysC above 1 mg/liter and as a 1-h infusion to children with CysC under 1 mg/liter). It was decided not to simulate longer infusion times in order to maximize tissue penetration and to circumvent potential drug incompatibilities, drug stability issues, and drug administration errors with more-complex dosing regimens (38). This dosing recommendation is fully in concordance with Jones et al., suggesting that higher/more-frequent amoxicillin-clavulanic acid dosing might be justified and needed in pediatric intensive care units (7). Moreover, it should be highlighted that 40% fT_>MIC is a rather conservative target, as 100% fT_>MIC has been associated with better outcomes in critically ill adults (39).

With regard to clavulanic acid, no clear-cut PK/PD targets are reported. As EUCAST clinical susceptibility breakpoints were determined in the presence of 2 mg/liter clavulanic acid, this was chosen as the target concentration (21). We hypothesized that lower targets of beta-lactamase inhibitors could potentially result in higher MIC breakpoints for the combined penicillin antibiotic. This hypothesis is supported by the study of Liu et al. demonstrating that for equal piperacillin exposure, different tazobactam half-lives have a significant effect on antimicrobial outcome (40). Although a larger interpatient variability was observed than for amoxicillin, our dosing recommendation (as above) resulted in an acceptable target attainment with no accumulation after 48 h of treatment. With regard to potential toxicity of higher cumulative amoxicillin-clavulanic acid daily doses, it should be noted that, for our optimized dosing regimen, we have specifically chosen not to select higher individual doses (25 mg/kg based on the amoxicillin compound) to eliminate potential safety risks related to higher peak concentrations. Finally, we feel confident that our optimized dosing regimen will not increase idiosyncratic clavulanic acid-induced liver toxicity, as it is known that the mechanism of its toxicity is immunoallergic, to some extent genetically controlled, and dose independent (41).

This research has some notable limitations. First, the studied population included a heterogeneous group of children with regard to possible differences in the (suspected) infecting organism and tissue involvement/penetration. Second, total drug plasma concentrations were mathematically corrected for protein binding instead of free drug concentration measurement in plasma or drug measurement at the site of infection. However, this simplification was previously found to be acceptable for β-lactam antibiotics with low protein binding like amoxicillin and clavulanic acid (42). Third, MIC values were not prospectively determined. Instead, a worst-case scenario using the clinical breakpoints for E. coli was chosen to challenge dosing regimens regardless of the infecting organism. This approach is justifiable, as β-lactam antibiotics have a wide therapeutic index, and the consequences of potentially supratherapeutic dosing are therefore of less concern. Fourth, notwithstanding that a substantial number of younger patients were recruited, PK data from additional neonates and infants are needed to estimate maturation parameters more precisely on both clearances and refine dosing regimens in these age categories.

In conclusion, this is the first population PK study demonstrating that the current dosing recommendations for amoxicillin-clavulanic acid can result in subtherapeutic treatment in critically ill children, thereby risking treatment failure. Besides developmental changes, CysC as a (new) biomarker for renal function and cotreatment with vasopressors were found to be significant covariates influencing drug disposition. The findings from this study make a significant contribution to knowledge regarding how to optimize the clinical use of amoxicillin-clavulanic acid in critically ill children. Whether these results of augmented renal clearance can be extrapolated to other renally cleared (β-lactam) antibiotics—or indeed other classes of medication—remains speculative and needs to be investigated in future research.