Evaluation of Bayesian predictability of vancomycin concentration using population pharmacokinetic parameters in pediatric patients

An Ensemble Model for Prediction of Vancomycin Trough Concentrations in Pediatric Patients

Purpose

This study aimed to establish an optimal model to predict vancomycin trough concentrations by using machine learning.

Patients and Methods

We enrolled 407 pediatric patients (age < 18 years) who received vancomycin intravenously and underwent therapeutic drug monitoring from June 2013 to April 2020 at Xinhua Hospital affiliated to Shanghai Jiaotong University School of Medicine. The median (interquartile range) age and weight of the patients were 2 (0.63–5) years and 12 (7.8–19) kg. Vancomycin trough concentrations were considered as the target variable, and eight different algorithms were used for predictive performance comparison. The whole dataset (407 cases) was divided into training group and testing group at the ratio of 80%: 20%, which were 325 and 82 cases, respectively.

Results

Ultimately, five algorithms (XGBoost, GBRT, Bagging, ExtraTree and decision tree) with high R² (0.657, 0.514, 0.468, 0.425 and 0.450, respectively) were selected and further ensembled to establish the final model and achieve an optimal result. For missing data, through filling the missing values and model ensemble, we obtained R²=0.614, MAE=3.32, MSE=24.39, RMSE=4.94 and a prediction accuracy of 51.22% (predicted trough concentration within ±30% of the actual trough concentration). In comparison with the pharmacokinetic models (R²=0.3), the machine learning model works better in model fitting and has better prediction accuracy.

Conclusion

Therefore, the ensemble model is useful for the vancomycin concentration prediction, especially in the population of children with great individual variation. As machine learning methods evolve, the clinical value of the ensemble model will be demonstrated in the clinical practice.

Pharmacokinetic-pharmacodynamic correlation of imipenem in pediatric burn patients using a bioanalytical liquid chromatographic method

<p>A bioanalytical method was developed and applied to quantify the free imipenem concentrations for pharmacokinetics and PK/PD correlation studies of the dose adjustments required to maintain antimicrobial effectiveness in pediatric burn patients. A reverse-phase Supelcosil LC18 column (250 x 4.6 mm 5 micra), binary mobile phase consisting of 0.01 M, pH 7.0 phosphate buffer and acetonitrile (99:1, v/v), flow rate of 0.8 mL/min, was applied. The method showed good absolute recovery (above 90%), good linearity (0.25-100.0 µg/mL, r²=0.999), good sensitivity (LLOQ: 0.25 µg/mL; LLOD: 0.12 µg/mL) and acceptable stability. Inter/intraday precision values were 7.3/5.9%, and mean accuracy was 92.9%. A bioanalytical method was applied to quantify free drug concentrations in children with burns. Six pediatric burn patients (median 7.0 years old, 27.5 kg), normal renal function, and 33% total burn surface area were prospectively investigated; inhalation injuries were present in 4/6 (67%) of the patients. Plasma monitoring and PK assessments were performed using a serial blood sample collection for each set, totaling 10 sets. The PK/PD target attained (40%T>MIC) for each minimum inhibitory concentration (MIC: 0.5, 1.0, 2.0, 4.0 mg/L) occurred at a percentage higher than 80% of the sets investigated and 100% after dose adjustment. In conclusion, the purification of plasma samples using an ultrafiltration technique followed by quantification of imipenem plasma measurements using the LC method is quite simple, useful, and requires small volumes for blood sampling. In addition, a small amount of plasma (0.25 mL) is needed to guarantee drug effectiveness in pediatric burn patients. There is also a low risk of neurotoxicity, which is important because pharmacokinetics are unpredictable in these critical patients with severe hospital infection. Finally, the PK/PD target was attained for imipenem in the control of sepsis in pediatric patients with burns.</p>

Vancomycin: a review of population pharmacokinetic analyses

Despite nearly five decades of clinical use, vancomycin has retained a significant and uncontested niche in the antibacterial arsenal because of its consistent activity against almost all Gram-positive bacteria. Nevertheless, major vancomycin toxicities have been reported in the literature - in particular, nephrotoxicity and ototoxicity. Vancomycin pharmacokinetics have been described in numerous studies for 25 years. This review presents a synthesis of the reported population pharmacokinetic models of vancomycin. The objective was to determine if there was a consensus on a structural model and which covariates were identified. A literature search was conducted from the PubMed database, from its inception through December 2010, using the following terms: 'vancomycin', 'pharmacokinetic(s)', 'population', 'model(ling)' and 'nonlinear mixed effect'. Articles were excluded if they were not pertinent. The reference lists of all selected articles were also evaluated. Twenty-five articles were included in this review: 15 models concerned paediatric patients and ten models were conducted in adults. In neonates and infants, the pharmacokinetics of vancomycin were mainly described by a one-compartment model, whereas in adults, a two-compartment model was preferentially used. Various covariates were tested but only three (age, creatinine clearance [CL(CR)] and body weight) were included in almost all of the described models. After inclusion of these covariates, the mean (range) values of the interindividual variability in the clearance and volume of distribution were 30% (15.6-45%) and 23% (12.6-48%), respectively. The mean (range) value of the residual variability was 20% (7-39.6%). This review highlights the numerous population pharmacokinetic models of vancomycin developed in recent decades and concludes with relevant information for clinicians and researchers. To optimize vancomycin dosage, this review points out the relevant covariates according to the target population. In adults, dosage optimization depends on CL(CR) and body weight, while in children, it depends on age, body weight and CL(CR). For future population pharmacokinetic studies, a sensitive liquid chromatography-tandem mass spectrometry method could be used and new covariates such as cardiac output or possible renal transporters could be tested. Finally, we suggest that external evaluation should be the first step in a pharmacokinetic analysis of vancomycin rather than describing a new model.

Population pharmacokinetic analysis of vancomycin in neonates. A new proposal of initial dosage guideline

AIM

To determine the population pharmacokinetic parameters of vancomycin in neonatal patients with a wide range of gestational age and birth weight, and subsequently to design an initial dosing schedule for vancomycin in neonates.

METHODS

Using nonlinear mixed-effects modelling (NONMEM VI), the pharmacokinetics of vancomycin were investigated in 70 neonates with postmenstrual age and body weight ranging 25.1-48.1 weeks and 0.7-3.7kg, respectively. A one-compartment linear disposition model with zero order input and first-order elimination was used to describe the data. Nine demographic characteristics and 21 co-administered drugs were evaluated as covariates of clearance (CL) and distribution volume (V(d) ) of vancomycin.

RESULTS

Weight-normalized clearance of vancomycin was influenced by postmenstrual age (PMA) and co-administration of amoxicillin-clavulanic acid. Weight-normalized volume of distribution was influenced by co-administration of spironolactone. CL and V(d) of the typical individual in this study population (PMA = 34.6 weeks, weight = 1.7kg) were estimated to be 0.066lh(-1) kg(-1) (95% CI 0.059, 0.073lh(-1) kg(-1) ) and 0.572lkg(-1) (95% CI 0.505, 0.639lkg(-1) ), respectively. This model was used to predict a priori serum vancomycin concentrations in a validation group (n= 41), which were compared with observed concentrations to determine the predictive performance of the model. The 95% confidence interval of mean prediction error included zero for both peak and trough vancomycin concentrations.

CONCLUSIONS

Postmenstrual age, co-administration of amoxicillin-clavulanic acid and spironolactone have a significant effect on the weight-normalized CL and V(d) . An initial dosage guideline for vancomycin is proposed for preterm and full-term neonates, whereas the population pharmacokinetic model can be used for dosage individualization of vancomycin.

A retrospective analysis of vancomycin pharmacokinetics in Japanese cancer and non-cancer patients based on routine trough monitoring data

The pharmacokinetics of vancomycin was retrospectively examined based on trough concentrations obtained during routine therapeutic drug monitoring to examine possible pharmacokinetic differences between adult Japanese cancer and non-cancer patients with various degrees of renal function. A total of 231 data points from 65 cancer patients and 41 non-cancer patients were collected, and patients' background, vancomycin dose, and vancomycin clearance estimated by an empirical Bayesian method were summarized. Regarding the patients' characteristics and clinical laboratory test data, no clear differences were found between the two groups. The relationship between vancomycin clearance and creatinine clearance were similar between the groups, suggesting little effect of malignancy on vancomycin clearance. After the sub-group comparisons regarding fluid retention and cancer type, no clear differences were found in the vancomycin clearance versus creatinine clearance relationship. We conclude that the initial dose of vancomycin should not necessarily be adjusted for cancer patients. For individualized vancomycin-based therapy, dose adjustment at the appropriate time is important according to information from routine therapeutic drug monitoring and clinical laboratory tests, and to observations of the efficacy, nephrotoxicity, and other conditions in each patient.