Systematic external evaluation of reported population pharmacokinetic models of vancomycin in Chinese children and adolescents

Optimization of Vancomycin Initial Dosing Regimen in Neonates Using an Externally Evaluated Population Pharmacokinetic Model

BACKGROUND

Vancomycin therapeutic monitoring guidelines were revised in March 2020, and a population pharmacokinetics-guided Bayesian approach to estimate the 24-hour area under the concentration-time curve to the minimum inhibitory concentration ratio has since been recommended instead of trough concentrations. To comply with these latest guidelines, we evaluated published population pharmacokinetic models of vancomycin using an external dataset of neonatal patients and selected the most predictive model to develop a new initial dosing regimen.

METHODS

The models were identified from the literature and tested using a retrospective dataset of Canadian neonates. Their predictive performance was assessed using prediction- and simulation-based diagnostics. Monte Carlo simulations were performed to develop the initial dosing regimen with the highest probability of therapeutic target attainment.

RESULTS

A total of 144 vancomycin concentrations were derived from 63 neonates in the external population. Five of the 28 models retained for evaluation were found predictive with a bias of 15% and an imprecision of 30%. Overall, the Grimsley and Thomson model performed best, with a bias of -0.8% and an imprecision of 20.9%; therefore, it was applied in the simulations. A novel initial dosing regimen of 15 mg/kg, followed by 11 mg/kg every 8 hours should favor therapeutic target attainment.

CONCLUSIONS

A predictive population pharmacokinetic model of vancomycin was identified after an external evaluation and used to recommend a novel initial dosing regimen. The implementation of these model-based tools may guide physicians in selecting the most appropriate initial vancomycin dose, leading to improved clinical outcomes.

Prospective validation of a model-informed precision dosing tool for vancomycin treatment in neonates

We recruited 48 neonates (50 vancomycin treatment episodes) in a prospective study to validate a model-informed precision dosing (MIPD) software. The initial vancomycin dose was based on a population pharmacokinetic model and adjusted every 36-48 h. Compared with a historical control group of 53 neonates (65 episodes), the achievement of a target trough concentration of 10-15 mg/L improved from 37% in the study to 62% in the MIPD group (P = 0.01), with no difference in side effects.

External validation of population pharmacokinetic models of vancomycin in postoperative neurosurgical patients

OBJECTIVE

Vancomycin is commonly used in the prevention and treatment of intracranial infections in postoperative neurosurgical patients with narrow therapeutic window and large pharmacokinetic variations. Several population pharmacokinetic (PPK) models of vancomycin have been established for neurosurgical patients. But comprehensive external evaluation has not been performed for almost all models. The objective of this study was to evaluate the predictive ability of published vancomycin PPK models in adult postoperative neurosurgical patients using an independent dataset.

METHOD

PubMed, Embase and China National Knowledge Internet databases were searched to identify published vancomycin PPK models in adult postoperative neurosurgical patients. Prediction-based and simulation-based diagnostics were used to evaluate model predictability. Bayesian forecasting was used to assess the influence of prior concentration on model prediction performance.

RESULT

A total of 763 vancomycin plasma concentrations from 493 postoperative neurosurgical patients were included in the external dataset. Eight population pharmacokinetic models of vancomycin in postoperative neurosurgical patients were included and evaluated. The model by Zhang et al. exhibited the best predictive performance in prediction-based diagnostics and prediction-corrected visual predictive checks, followed by the model by Shen et al. The predictive performance of other models was not satisfactory. The normalized predictive distribution error test shows that none of the models is suitable to describe our data. The predictive performance of vancomycin models was obviously improved by maximum a posteriori Bayesian forecasting.

CONCLUSION

The published PPK models for adult postoperative neurosurgical patients show extensive variation in predictive performance in our patients. Although it is challenging to recommend initial doses of vancomycin from these predictive models, the combination of model-based prediction and therapeutic drug monitoring can be used for dose optimization.

External Evaluation of Population Pharmacokinetic Models for Precision Dosing: Current State and Knowledge Gaps

Predicting drug exposures using population pharmacokinetic models through Bayesian forecasting software can improve individual pharmacokinetic/pharmacodynamic target attainment. However, selecting the most adapted model to be used is challenging due to the lack of guidance on how to design and interpret external evaluation studies. The confusion around the choice of statistical metrics and acceptability criteria emphasises the need for further research to fill this methodological gap as there is an urgent need for the development of standards and guidelines for external evaluation studies. Herein we discuss the scientific challenges faced by pharmacometric researchers and opportunities for future research with a focus on antibiotics.

Machine learning advances the integration of covariates in population pharmacokinetic models: Valproic acid as an example

Background and Aim: Many studies associated with the combination of machine learning (ML) and pharmacometrics have appeared in recent years. ML can be used as an initial step for fast screening of covariates in population pharmacokinetic (popPK) models. The present study aimed to integrate covariates derived from different popPK models using ML.

Methods: Two published popPK models of valproic acid (VPA) in Chinese epileptic patients were used, where the population parameters were influenced by some covariates. Based on the covariates and a one-compartment model that describes the pharmacokinetics of VPA, a dataset was constructed using Monte Carlo simulation, to develop an XGBoost model to estimate the steady-state concentrations (Css) of VPA. We utilized SHapley Additive exPlanation (SHAP) values to interpret the prediction model, and calculated estimates of VPA exposure in four assumed scenarios involving different combinations of CYP2C19 genotypes and co-administered antiepileptic drugs. To develop an easy-to-use model in the clinic, we built a simplified model by using CYP2C19 genotypes and some noninvasive clinical parameters, and omitting several features that were infrequently measured or whose clinically available values were inaccurate, and verified it on our independent external dataset.

Results: After data preprocessing, the finally generated combined dataset was divided into a derivation cohort and a validation cohort (8:2). The XGBoost model was developed in the derivation cohort and yielded excellent performance in the validation cohort with a mean absolute error of 2.4 mg/L, root-mean-squared error of 3.3 mg/L, mean relative error of 0%, and percentages within ±20% of actual values of 98.85%. The SHAP analysis revealed that daily dose, time, CYP2C19*2 and/or *3 variants, albumin, body weight, single dose, and CYP2C19*1*1 genotype were the top seven confounding factors influencing the Css of VPA. Under the simulated dosage regimen of 500 mg/bid, the VPA exposure in patients who had CYP2C19*2 and/or *3 variants and no carbamazepine, phenytoin, or phenobarbital treatment, was approximately 1.74-fold compared to those with CYP2C19*1/*1 genotype and co-administered carbamazepine + phenytoin + phenobarbital. The feasibility of the simplified model was fully illustrated by its performance in our external dataset.

Conclusion: This study highlighted the bridging role of ML in big data and pharmacometrics, by integrating covariates derived from different popPK models.

Clinical application of vancomycin population pharmacokinetics model in patients with hematological diseases and neutropenia

To explore the clinical application of a population pharmacokinetics (PPK) model of vancomycin in patients with hematological diseases and neutropenia. Patients with hematological diseases and neutropenia were included in the PPK model study. Nonlinear mixed effect modeling approach (NONMEM) was used for model establishment. Monte Carlo simulation was carried out. A total of 74 patients were divided into model group and non‐model group for clinical application research. The model group was given the initial dose of 1g q8h, and the non‐model group was given 1g q12h as an empiric initial dosage. The follow‐up dose adjustments were made according to the concentration results. This two‐compartment model showed good stability and accuracy. The first trough concentration (C₀) and the compliance rate of the first C₀ were much higher in the model group than that in the non‐model group (14.30 ± 4.73 μg/ml and 59.38% vs. 8.02 ± 2.61 μg/ml, 35.71%). Less patients needed dose adjustments and fewer adjustment times in the model group than those in the non‐model group (12.50% and 0.13 ± 0.34 times vs. 50.00% and 0.61 ± 0.66 times). This suggested that for those patients who had a Creatinine clearance rate (CLCR) ≥ 90 ml/min/1.73 m², the initial dose of 1g q8h may help to reach the target C₀ (10∼20 μg/ml) quickly. It also helped to reduce the times and number of patients who need dose adjustments. Our PPK model of vancomycin in patients with hematologic diseases and neutropenia can be used to shorten the time to reach the target concentration and reduce the number of dose adjustments.Clinical trial registration: Not applicable (Retrospective study).