Population Pharmacokinetics and Dosage Optimization of Vancomycin in Pediatric Patients with Skin and Soft Tissue Infections, Bone, and Joint Infections

ABSTRACT

Vancomycin is recommended for the treatment of skin and soft tissue infections (SSTI) and bone and joint infections (BJI). However, a detailed investigation of the pharmacokinetic profile and optimal dosing regimens of vancomycin in pediatric patients with SSTI and BJI is lacking. We successfully developed a new PopPK model of vancomycin in this population by using scavenged blood samples with the typical values for clearance (CL) of 0.14 L/h/kg and volume of distribution (V) of 0.5 L/kg. Body weight was confirmed as the significant covariate on CL and V. The optimal dosing regimens of 75 mg/kg/day and 80 mg/kg/day were recommended for this specific population.

INTRODUCTION

Methicillin-resistant Staphylococcus aureus (MRSA) is still a leading cause of skin and soft tissue infections (SSTI) and bone and joint infections (BJI), which poses a formidable clinical threat (1, 2). Vancomycin is a glycopeptide antimicrobial agent widely used for the treatment of bacteremia and endocarditis, pneumonia, and central nervous system infections, as well as SSTI and BJI (3, 4). The ratio of area under the curve over 24 h to MIC (AUC/MIC) has been recommended as the primary pharmacokinetic/pharmacodynamic (PK/PD) predictor of vancomycin activity (5–7). To accurately estimate the AUC of vancomycin with limited PK sampling, one such widely applied approach involves the use of the well-developed population PK (PopPK) model (8). Nevertheless, it is worth mentioning that the estimation of vancomycin clearance (CL) and volume of distribution (V) varies a lot in different pediatric populations (9). The reported coefficient of variation of interindividual variability on CL and the central V compartment can reach as high as 50.4% and 232%, respectively (9). SSTI and BJI encompass various pathophysiological conditions, which range from simple mild infections to severe life-threatening infections (10). The pathophysiological changes associated with severe life-threatening infections and the penetration of antimicrobials into infection sites may significantly change its PK patterns (11). Until now, there is a paucity of clinical trial data that describes vancomycin PK profile at steady state in pediatric patients with SSTI and BJI (9). This study aimed to develop a new PopPK model from scavenged blood samples and characterize the vancomycin PK profile via covariate analysis.

We conducted a prospective PK study of vancomycin in an orthopedic ward at Children’s Hospital Affiliated to Zhengzhou University. This study was approved by the ethics committee of the hospital (2020-KS-01). Pediatric patients aged from 1 month to 18 years with suspected or confirmed SSTI and BJI receiving intravenous vancomycin as part of their routine medical care were enrolled. Vancomycin (Vancocin) was administered by intravenous infusion at the dose of 40 mg/kg/day divided into three or four times within 1 to 2 h. All samples for PopPK analysis were collected at steady state (at least 2 days after administration of vancomycin) and obtained from the scavenged blood immediately after routine biochemical tests. The infusion and sampling time were precisely recorded. The samples were centrifuged at 5,000 rpm for 5 min at 4°C, and plasma samples were stored for up to 1 month at −70°C for further analysis. The total concentration of vancomycin was determined by the enzyme-multiplied immunoassay technique (EMIT) using the SYVA Viva-E clinical biochemical analyzer (SYVA Viva-E) with a calibration range of 2.0 to 50 mg/L.

A total of 25 participants and corresponding 75 scavenged blood samples meeting inclusion criteria were included and evaluated for analysis from May 2020 to December 2021. The median (interquartile range [IQR]) age and median body weight were 2.68 (1.56, 7.54) years and 12.25 (10.8, 19.1) kg, respectively. Forty-four percent (11 of 25) of the patients were male. All of the patients had either or both SSTI and BJI. MRSA was isolated from bone marrow fluid in 16 patients with an MIC of 0.5 or 1 mg/L. The baseline information of the participants is summarized in Table 1. Of the 75 values of vancomycin concentrations, two were below the lower limit of quantification and excluded from further analysis. The 73 detectable vancomycin concentrations used for the development of the PopPK model ranged from 2.1 mg/L to 26.2 mg/L (Fig. 1).

TABLE 1.

Baseline characteristics of the 25 pediatric patients included in the study

Characteristics	Indexa	Median (IQR)	Range
No. of patients	25	/	/
Male	11	/	/
No. of observations per patient	2.92 ± 1.76	2 (1, 4)	1–7
Age (yr)	4.1 ± 3.4	2.68 (1.56, 7.54)	0.1–10.8
Body wt (kg)	17.5 ± 10.7	12.25 (10.8, 19.1)	4.9–51.0
Height (meter)	1.0 ± 0.28	0.90 (0.81, 1.10)	0.52–1.65
Albumin (g/L)	37.7 ± 5.0	40.1 (34.2, 41.2)	28.8–45.8
Aspartate aminotransferase (U/L)	25.9 ± 9.1	23.7 (19.8, 28.2)	11.8–50.2
Alanine aminotransferase (U/L)	20.5 ± 12.6	20.5 (11.0, 25.3)	6.6–81.4
Direct bilirubin (μmol/L)	1.7 ± 0.7	1.6 (1.1, 2.5)	0.4–2.8
Total bilirubin (μmol/L)	5.7 ± 2.8	5.1 (3.7, 7.2)	2.2–15.8
Alkaline phosphatase (U/L)	170.5 ± 75.0	184.5 (130.6, 216.8)	11.0–401.0
Serum creatinine (μmol/L)	25.7 ± 10.8	22.1 (16, 32.2)	12–57.5
Blood urea nitrogen (mmol/L)	3.5 ± 1.6	2.9 (2.3, 4.0)	1.7–7.1
Vancomycin daily dose (mg)	723.1 ± 412.8	580 (400, 910)	196–2,000.0
Vancomycin daily dose (mg/kg)	41.8 ± 6.2	40 (40, 40)	36.4–60
Concentrations (mg/L)	11.7 ± 5.7	11.5 (6.6, 15.4)	2.1–26.2
Sampling time after last dose (h)	3.6 ± 1.5	3.2 (2.4, 5.3)	1.1–6.5

^aValues are means ± SD.

FIG 1.

Observed concentrations of vancomycin versus time after last dose. (A) Linear plot. (B) Semilog plot.

The PopPk analysis was conducted using a nonlinear mixed effects modeling program named NONMEM with the first-order conditional estimation method with interaction (FOCE-I). The vancomycin PopPK profile was best characterized by a one-compartment model with first-order (linear) elimination. Interindividual variability and residual variability were best fitted by an exponential model and a combined model (Table 2), respectively. The estimated CL and V of vancomycin in the final model led to a lower objective function value (OFV) (ΔOFV = 44.80) compared with the base model.

TABLE 2.

Estimated parameters of the final vancomycin PopPK model and the bootstrap results based on the final model^a

Parametersb	Full data set of the final model		Bootstrap (n = 1,000)		Bias (%)
	Estimate	RSE (%)	Median	95% CI
θCL (L/h)	2.53	4.3	2.54	2.31–2.75	0.4
θV (L)	8.84	7.4	8.89	7.48–10.20	0.6
θwt-CL	0.771	9.6	0.775	0.61–0.93	0.5
θwt-V	0.957	9.8	0.963	0.74–1.18	0.6
Between-subject variability ηCL (%)	18.2	21.1	17.2	5.29–24.4	–5.5
Residual variability ε1 (%)	1.07	26.1	1.00	0.08–2.07	–6.5
Residual variability ε2	2.7	17.8	2.62	0.70–4.69	−3.0

^aThe final PopPK model of vancomycin was as follows:$\text{CL} \left(\text{L}/\text{h}\right) = 2.53\times {\left(\text{Wt}/17.5\right)}^{0.771}\times e^{\eta \text{CL}}$$\text{V}�\left(\text{L}\right)�=�\text{8}.\text{84}\times {\left(\text{Wt}/17.5\right)}^{0.957}$

^bCL, clearance; V, volume of distribution; θ_CL, θ_V, typical population value of CL and V; θ_wt-CL, weight effect on CL; θ_wt-V, weight effect on V; RSE%, relative standard error of the estimate; Bias%, prediction error [Bias% = (bootstrap median − NONMEM estimate)/NONMEM estimate ×100%]; CI, confidence interval.

Various covariates including sex, age, height, body weight, body mass index (BMI), serum creatinine, aspartate aminotransferase (AST), alanine aminotransferase (ALT), albumin, and glomerular filtration rate were tested by stepwise modeling. If the objective function value (OFV) decreased by more than 3.84 compared with the value of the basic model in the forward inclusion process, covariates were deemed statistically significant (P < 0.05, χ² distribution; 1 degree of freedom) and added to the model. Subsequently, each covariate was independently examined in the backward elimination process and retained in the model if the OFV increased more than 6.63 (P < 0.01, χ² distribution; 1 degree of freedom). Serum creatinine concentrations were measured by a sarcosine oxidase enzymatic method using Beckman Coulter AU5800 with the lower limit of quantification of 2.4 μmol/L. The estimated glomerular filtration rate (eGFR) was calculated with the modified Schwartz equation (equation 1) (12, 13), where k is 0.45 for children under 1 year and 0.413 for children between 1 and 18 years, height is expressed in centimeters, Scr is serum creatinine concentration (mg/dL), age is expressed in years, and weight is expressed in kilograms.

$$\mathit{eGFR}=\frac{k\times \text{height}}{\text{Scr}}$$ (1)

Among all the tested covariates, age, height, and body weight had a significant effect on CL and V in the forward inclusion process (Fig. 2). Age, height, and body weight appeared to have a similar effect on CL and V. Considering the high correlation of body weight with age and height (r = 0.86 and 0.95) (Fig. 2), body weight was the single significant covariate incorporating into the final model to avoid a collinearity effect (Table 2). In our final model, the weight-normalized CL and V of vancomycin at steady state were estimated to be 0.14 L/h/kg and 0.5 L/kg, respectively.

FIG 2.

The Correlation plot between covariates and CL and V. Age, height, and body weight had a significant effect on CL and V in the forward inclusion process. HT, height; BW, body weight.

Internal and external evaluations were used to evaluate the accuracy and robustness of the final model. The results of the goodness-of-fit plots are shown in Fig. 3A to D. The final model had an adequate fit with minimal bias as illustrated by the even distribution of the population predicted concentrations (PRED) and individual prediction concentrations (IPRED), as well as the trend between conditional weighted residuals (CWRES) and PRED or CWRES and time after dose (TAD). In addition, the visual predictive check (VPC) as shown in Fig. 3E further demonstrated an adequate predictive performance of the final model. Nonparametric bootstrap was used to further assess the stability and reliability of the final model, where resampling was performed 1,000 times. As summarized in Table 2, the estimated PopPK parameters generated from bootstrap analysis were consistent with those in the final model, which indicated the stability and reliability of the final model. We then validated the performance of the final model by external data. Here, we had 34 pediatric patients with SSTI and BJI at our center between 2017 and 2021, and 49 vancomycin trough concentrations at steady state used for routine therapeutic drug monitoring were collected (data for building the PopPK model were not included). The median (IQR) age and median body weight of the patients were 3.8 (1.4, 6.9) years and 14.0 (10.1, 23.2) kg, respectively. IPRED versus observed concentrations were shown in Fig. 4. The mean relative prediction error (MPE%) and mean relative absolute prediction error (MAPE%) values calculated by equations 2 and 3 were −2.0% and 18.9%, respectively, proving the good prediction with a low bias of the final model.

$$\mathit{MPE}\%=\frac{1}{N}{\displaystyle \sum }_1^N\left(\frac{\mathit{pre}{d}_i-\mathit{ob}{s}_i}{\mathit{ob}{s}_i}\right)\times 100$$ (2)

$$\mathit{MAPE}\%=\frac{1}{N}{\displaystyle \sum }_1^N\left(\frac{\left|\mathit{pre}{d}_i-\mathit{ob}{s}_i\right|}{\mathit{ob}{s}_i}\right)\times 100$$ (3)

FIG 3.

Internal evaluation of the final vancomycin PopPK model. (A) Individual observed concentrations versus predicted concentrations. (B) Population observed concentrations versus predicted concentrations. (C) Conditional weighted residuals versus population predicted concentrations. (D) Conditional weighted residuals versus time after dose. (E) Prediction-corrected visual predictive check by time after dose. The red solid and dashed lines are the 50th and 5th and 95th percentiles of observed concentrations, respectively. The three shaded areas highlight the simulation-based 95% confidence intervals for corresponding percentiles.

FIG 4.

External evaluation of the final vancomycin PopPK model. IPRED versus observed concentrations.

An AUC value ranging from 400 to 800 mg·h/L (assuming MIC with a common value of 1 mg/L) was considered the PK/PD target for dose optimization of vancomycin. Different dosing regimens (40, 60, 70, 75, 80, 90, and 100 mg/kg/day divided into two, three, or four times) were simulated for patients aged from 1 month to 18 years by using parameters derived from the final model. AUC was calculated as in equation 4, where “DOSE” represents the daily dose of vancomycin, CL is the individual clearance calculated based on the developed PopPK model of vancomycin, and AUC is the area under the concentration-time curve for 0 to 24 h at steady state. The probability of target attainment (PTA) results for different dosing regimens is shown in Table 3. The doses of 75 mg/kg/day and 80 mg/kg/day resulted in an adequate PTA above 85%, which may be the optimal dosing regimens and should be recommended.

$$\mathit{AUC}=\frac{\mathit{DOSE}}{\mathit{CL}}$$ (4)

TABLE 3.

The simulated results of different regimes^a

Dosing regimen	AUC (mg · h/L; mean ± SD)	800 ≥ AUC ≥ 400 (%)	AUC > 800 (%)	Trough (mg/L; mean ± SD)
10 mg/kg q6h	301.7 ± 71.0	8.4	0	6.9 ± 4.0
13 mg/kg q6h	380.3 ± 85.1	36.6	0	8.4 ± 4.5
15 mg/kg q6h	451.5 ± 106.5	65.5	0.3	10.1 ± 5.8
20 mg/kg q6h	601.9 ± 142.0	84.6	8.2	13.5 ± 7.6
25 mg/kg q6h	730.0 ± 164.0	69.8	30.1	16.9 ± 9.1
13 mg/kg q8h	287.2 ± 63.4	5.4	0	4.7 ± 2.9
20 mg/kg q8h	440.3 ± 97.7	62.6	0.1	7.2 ± 4.4
25 mg/kg q8ha	549.6 ± 122.1	86.4	3.6	9.1 ± 5.6
30 mg/kg q8h	658.6 ± 146.6	82.7	15.3	11.2 ± 6.9
20 mg/kg q12h	299.5 ± 64.9	7.7	0	2.9 ± 2.0
25 mg/kg q12h	373.3 ± 80.8	33.7	0	3.5 ± 2.5
30 mg/kg q12h	446.4 ± 96.6	65.3	0.1	4.1 ± 3.0
40 mg/kg q12ha	592.6 ± 128.6	87.9	7.2	5. 6 ± 4.2
50 mg/kg q12h	737.5 ± 161.2	68.5	31.2	7.3 ± 5.7

^aOptimal dosing regimens.

In previous reports, the values of CL and V in different pediatric populations ranged from 0.0155 to 0.255 (median 0.103) L/h/kg and 0.32 to 5.89 (median 0.64) L/kg (11). In the study, we observed median values of CL with 0.14 L/h/kg and V with 0.5 L/kg, which were consistent with the reporting median values of CL and V and suggested a moderate tissue diffusion of vancomycin in this population. The large variability of CL and V could be explained by diverse clinical manifestations and the differing age groups of the children. Additionally, none of the reported models described the characteristics of vancomycin in pediatric patients with SSTI and BJI.

It is widely accepted that growth and development are two features of children that are not observed in adults (14). In the present work, we incorporated body weight into the final model. However, renal function descriptors including serum creatinine and glomerular filtration rate were tested but did not confirm a significant relationship with vancomycin clearance in our study, despite vancomycin is eliminated primarily via the renal route (15). This finding was consistent with the previously published study in pediatric patients (16–18). The limited changes in creatinine levels in our population probably led to the clinical outcome.

Due to the narrow therapeutic window, therapeutic drug monitoring is required for vancomycin to optimize the dosage regimens and guarantee effective and safe therapy. Maintaining AUC/MIC above 400 has been recommended for pediatric patients to achieve clinical efficiency according to the current guidelines (5, 7). Of note, AUC above 800 mg · h/L increased the risk of acute kidney injury (19). Therefore, AUC/MIC ranging from 400 to 800 was considered the therapeutic PK/PD target and used to determine the optimal dosing regimens of vancomycin in our study. The PTA results showed that 75 mg/kg/day and 80 mg/kg/day were the optimal dosing regimens for vancomycin, which was consistent with the guidelines recommended vancomycin dosage with 60 to 80 mg/kg/day divided into three or four times for children with normal renal function (5, 7). We successfully established detailed vancomycin dosing regimens in pediatric patients with SSTI and BJI, which will especially benefit the patients unavailable to perform therapeutic drug monitoring.

Moreover, trough concentration at steady state was no longer recommended from an efficiency point of view (5, 6). In the study, we found that although vancomycin trough concentrations were positively correlated with AUC as illustrated by Fig. 5, the troughs corresponding to the target AUC varied substantially. For example, 60 mg/kg/day divided into two, three, and four times resulted in a similar PTA (Table 3). However, the vancomycin trough concentrations do not necessarily need to be above 10 mg/L to achieve the target AUC. Therefore, a merely trough-guiding dosing optimization may be not accurate and optimal for vancomycin (20).

FIG 5.

AUC versus trough concentration for simulated dosing regimens of 15 mg/kg/dose every 6 h, 20 mg/kg/dose every 8 h, and 30 mg/kg/dose every 12 h.

However, there are some limitations in this study. First, this is a single-center study that was limited to pediatric patients with SSTI and BJI. The internal and external validity of the final model was still constrained by the similar type populations. Second, the enrollment size was relatively small. Nevertheless, the blood samples collected from the majority of patients were across different time points, which provided sufficient PK information to estimate the PK parameters with good precision.

In conclusion, we successfully developed a PopPK model of vancomycin in pediatric patients with SSTI and BJI by using scavenged blood samples. To the best of our knowledge, it is the first vancomycin PopPK model established in pediatric patients with SSTI and BJI. A one-compartment model with first-order elimination was proven to have the best fit for the present concentration-time data. Body weight showed a significant impact on vancomycin pharmacokinetics. Model-based dosing regimens were proposed based on body weight in this specific population and need to be further validated in clinical practice.