Hospitalized Patients With and Without Hemodialysis Have Markedly Different Vancomycin Pharmacokinetics: A Population Pharmacokinetic Model-Based Analysis

Impact of a 20 mg/kg vancomycin loading dose on early AUC target attainment

This retrospective chart review evaluated whether 20 mg/kg vancomycin loading doses increase early area under the curve (AUC) target attainment within 48 hours in comparison to non-loading dose regimens. There were no differences between groups for the primary outcome (46 % vs. 50 %; P = 0.58).

Vancomycin: An analysis and evaluation of eight population pharmacokinetic models for clinical application in general adult population

INTRODUCTION

Based on the recent guidelines for vancomycin therapeutic drug monitoring (TDM), the area under the curve to minimum inhibitory concentration ratio was to be employed combined with the usage of population pharmacokinetic (popPK) model for dosing adaptation. Yet, deploying these models in a clinical setting requires an external evaluation of their performance.

OBJECTIVES

This study aimed to evaluate existing vancomycin popPK models from the literature for the use in TDM within the general patient population in a clinical setting.

METHODS

The models under external evaluation were chosen based on a review of literature covering vancomycin popPK models developed in general adult populations. Patients' data were collected from Charles-Le Moyne Hospital (CLMH). The external evaluation was performed with NONMEM® (v7.5). Additional analyses such as evaluating the impact of number of samples on external evaluation, Bayesian forecasting, and a priori dosing regimen simulations were performed on the best performing model.

RESULTS

Eight popPK models were evaluated with an independent dataset that included 40 patients and 252 samples. The model developed by Goti and colleagues demonstrated superior performance in diagnostic plots and population predictive performance, with bias and inaccuracy values of 0.251% and 22.7%, respectively, and for individual predictive performance, bias and inaccuracy were -4.90% and 12.1%, respectively. When limiting the independent dataset to one or two samples per patient, the Goti model exhibited inadequate predictive performance for inaccuracy, with values exceeding 30%. Moreover, the Goti model is suitable for Bayesian forecasting with at least two samples as prior for the prediction of the next trough concentration. Furthermore, the vancomycin dosing regimen that would maximize therapeutic targets of area under the curve to minimum inhibitory concentration ratio (AUC24/MIC) and trough concentrations (Ctrough) for the Goti model was 20 mg/kg/dose twice daily.

CONCLUSION

Considering the superior predictive performance and potential for Bayesian forecasting in the Goti model, future research aims to test its applicability in clinical settings at CLMH, both in a priori and a posteriori scenario.

Validation and development of population pharmacokinetic model of vancomycin using a real-world database from a nationwide free web application

INTRODUCTION

Vancomycin requires a population pharmacokinetic (popPK) model to estimate the area under the concentration-time curve (AUC), and an AUC-guided dosing strategy is necessary. This study aimed to develop a popPK model for vancomycin using a real-world database pooled from a nationwide web application (PAT).

METHODS

In this retrospective study, the PAT database between December 14, 2022 and April 6, 2023 was used to develop a popPK model. The model was validated and compared with six existing models based on the predictive performance of datasets from another PAT database and the Kumamoto University Hospital. The developed model determined the dosing strategy for achieving the target AUC.

RESULTS

The modeling populations consisted of 7146 (13,372 concentrations from the PAT database), 3805 (7540 concentrations from the PAT database), and 783 (1775 concentrations from Kumamoto University Hospital) individuals. A two-compartment popPK model was developed that incorporated creatinine clearance as a covariate for clearance and body weight for central and peripheral volumes of distribution. The validation demonstrated that the popPK model exhibited the smallest mean absolute prediction error of 5.07, outperforming others (ranging from 5.10 to 5.83). The dosing strategies suggested a first dose of 30 mg/kg and maintenance doses adjusted for kidney function and age.

CONCLUSIONS

This study demonstrated the updating of PAT through the validation and development of a popPK model using a vast amount of data collected from anonymous PAT users.

Differences in drug removal between standard high-flux and medium cut-off dialyzers in a case of severe vancomycin toxicity

Vancomycin is a widely used glycopeptide antibiotic with the need for therapeutic drug monitoring to avoid renal toxicity. We report a case of severe vancomycin-associated anuric acute kidney injury managed with successful drug-removal by hemodialysis (HD) using different types of dialyzers. Medium cut-off (MCO) and high-flux dialyzers were effective in drug removal. Higher vancomycin elimination rate and lower plasma half-life were achieved with MCO dialyzer despite low-flow vascular access and intolerance to ultrafiltration. MCO dialyzers may be reasonable for drug removal in patients with intolerance of ultrafiltration, low-flow vascular access or impracticality of hemodiafiltration. Future studies should explore the use of MCO dialyzers in comparison with high-flux HD and hemodiafiltration in both the acute and chronic setting.

External Validation of Obese/Critically Ill Vancomycin Population Pharmacokinetic Models in Critically Ill Patients Who Are Obese

Obesity combined with critical illness might increase the risk of acquiring infections and hence mortality. In this patient population the pharmacokinetics of antimicrobials vary significantly, making antimicrobial dosing challenging. The objective of this study was to assess the predictive performance of published population pharmacokinetic models of vancomycin in patients who are critically ill or obese for a cohort of critically ill patients who are obese. This was a multi-center retrospective study conducted at 2 hospitals. Adult patients with a body mass index of ≥30 kg/m2 were included. PubMed was searched for published population pharmacokinetic studies in patients who were critically ill or obese. External validation was performed using Monolix software. A total of 4 models were identified in patients who were obese and 5 models were identified in patients who were critically ill. In total, 138 patients who were critically ill and obese were included, and the most accurate models for these patients were the Goti and Roberts models. In our analysis, models in patients who were critically ill outperformed models in patients who were obese. When looking at the most accurate models, both the Goti and the Roberts models had patient characteristics similar to ours in terms of age and creatinine clearance. This indicates that when selecting the proper model to apply in practice, it is important to account for all relevant variables, besides obesity.

Population pharmacokinetics of antibacterial agents in the older population: a literature review

INTRODUCTION

Older individuals face an elevated risk of developing bacterial infections. The optimal use of antibacterial agents in this population is challenging because of age-related physiological alterations, changes in pharmacokinetics (PK) and pharmacodynamics (PD), and the presence of multiple underlying diseases. Therefore, population pharmacokinetics (PPK) studies are of great importance for optimizing individual treatments and prompt identification of potential risk factors.

AREA COVERED

Our search involved keywords such as 'elderly,' 'old people,' and 'geriatric,' combined with 'population pharmacokinetics' and 'antibacterial agents.' This comprehensive search yielded 11 categories encompassing 28 antibacterial drugs, including vancomycin, ceftriaxone, meropenem, and linezolid. Out of 127 studies identified, 26 (20.5%) were associated with vancomycin, 14 (11%) with meropenem, and 14 (11%) with piperacillin. Other antibacterial agents were administered less frequently.

EXPERT OPINION

PPK studies are invaluable for elucidating the characteristics and relevant factors affecting the PK of antibacterial agents in the older population. Further research is warranted to develop and validate PPK models for antibacterial agents in this vulnerable population.

Model-Informed Precision Dosing Improves Outcomes in Patients Receiving Vancomycin for Gram-Positive Infections

Background

Consensus guidelines for dosing and monitoring of vancomycin recommend collection of 2 serum concentrations to estimate an area under the curve/minimum inhibitory concentration ratio (AUC/MIC). Use of Bayesian software for AUC estimation and model-informed precision dosing (MIPD) enables pre-steady state therapeutic drug monitoring using a single serum concentration; however, data supporting this approach are limited.

Methods

Adult patients with culture-proven gram-positive infections treated with vancomycin ≥72 hours receiving either trough-guided or AUC-guided therapy were included in this retrospective study. AUC-guided therapy was provided using MIPD and single-concentration monitoring. Treatment success, vancomycin-associated acute kidney injury (VA-AKI), and inpatient mortality were compared using a desirability of outcome ranking analysis. The most desirable outcome was survival with treatment success and no VA-AKI, and the least desirable outcome was death.

Results

The study population (N = 300) was comprised of an equal number of patients receiving AUC-guided or trough-guided therapy. More patients experienced the most desirable outcome in the AUC-guided group compared to the trough-guided group (58.7% vs 46.7%, P = .037). Rates of VA-AKI were lower (21.3% vs 32.0%, P = .037) and median hospital length of stay was shorter (10 days [interquartile range {IQR}, 8-20] vs 12 days [IQR, 8-25]; P = .025) among patients receiving AUC-guided therapy.

Conclusions

AUC-guided vancomycin therapy using MIPD and single-concentration monitoring improved outcomes in patients with culture-proven gram-positive infections. Safety was improved with reduced incidence of VA-AKI, and no concerns for reduced efficacy were observed. Moreover, MIPD allowed for earlier assessment of AUC target attainment and greater flexibility in the collection of serum vancomycin concentrations.

Combining pharmacokinetic and electrophysiological models for early prediction of drug-induced arrhythmogenicity

BACKGROUND AND OBJECTIVE

In silico methods are gaining attention for predicting drug-induced Torsade de Pointes (TdP) in different stages of drug development. However, many computational models tended not to account for inter-individual response variability due to demographic covariates, such as sex, or physiologic covariates, such as renal function, which may be crucial when predicting TdP. This study aims to compare the effects of drugs in male and female populations with normal and impaired renal function using in silico methods.

METHODS

Pharmacokinetic models considering sex and renal function as covariates were implemented from data published in pharmacokinetic studies. Drug effects were simulated using an electrophysiologically calibrated population of cellular models of 300 males and 300 females. The population of models was built by modifying the endocardial action potential model published by O'Hara et al. (2011) according to the experimentally measured gene expression levels of 12 ion channels.

RESULTS

Fifteen pharmacokinetic models for CiPA drugs were implemented and validated in this study. Eight pharmacokinetic models included the effect of renal function and four the effect of sex. The mean difference in action potential duration (APD) between male and female populations was 24.9 ms (p<0.05). Our simulations indicated that women with impaired renal function were particularly susceptible to drug-induced arrhythmias, whereas healthy men were less prone to TdP. Differences between patient groups were more pronounced for high TdP-risk drugs. The proposed in silico tool also revealed that individuals with impaired renal function, electrophysiologically simulated with hyperkalemia (extracellular potassium concentration [K+]o = 7 mM) exhibited less pronounced APD prolongation than individuals with normal potassium levels. The pharmacokinetic/electrophysiological framework was used to determine the maximum safe dose of dofetilide in different patient groups. As a proof of concept, 3D simulations were also run for dofetilide obtaining QT prolongation in accordance with previously reported clinical values.

CONCLUSIONS

This study presents a novel methodology that combines pharmacokinetic and electrophysiological models to incorporate the effects of sex and renal function into in silico drug simulations and highlights their impact on TdP-risk assessment. Furthermore, it may also help inform maximum dose regimens that ensure TdP-related safety in a specific sub-population of patients.

Feasibility of individualised patient modelling for continuous vancomycin infusions in outpatient antimicrobial therapy, a retrospective study

BACKGROUND

The area under the curve (AUC) to minimum inhibitory concentration (MIC) ratio is proposed as a therapeutic drug-monitoring parameter for dosing vancomycin continuous infusion in methicillin-resistant Staphylococcus aureus (MRSA) infection. Individualised pharmacokinetic-pharmacodynamic (PK/PD) calculation of AUC24 may better represent therapeutic dosing than current Therapeutic Drug Monitoring (TDM) practices, targeting a Steady State Concentration of 15-25 mg/L.

AIM

To compare real world TDM practice to theoretical, individualised, PK/PD target parameters utilising Bayesian predictions to steady state concentrations (Css) for outpatients on continuous vancomycin infusions.

METHOD

A retrospective single centre study was conducted at a tertiary hospital on adult patients, enrolled in an outpatient parenteral antimicrobial therapy (OPAT) program, receiving vancomycin infusions for MRSA infection. Retrospective Bayesian dosing was modelled to target PK/PD parameters and compared to real world data.

RESULTS

Fifteen patients were evaluated with 53% (8/15) achieved target CSS during hospitalisation, and 83% (13/15) as outpatient. Median Bayesian AUC/MIC was 613 mg.h/L with CSS 25 mg/L. Patients suffering an Acute Kidney Injury (33%) had higher AUC0-24/MIC values. Retrospective Bayesian modelling demonstrated on median 250 mg/24 h lower doses than that administered was required (R2 = 0.81) which achieved AUC24/MIC median 444.8 (range 405-460) mg.h/L and CSS 18.8 (range 16.8-20.4) mg/L.

CONCLUSION

Bayesian modelling could assist in obtaining more timely target parameters at lower doses for patients receiving continuous vancomycin infusion as part of an OPAT program, which may beget fewer adverse effects. Utilisation of personalised predictive modelling may optimise vancomycin prescribing, achieving earlier target concentrations as compared to empiric dosing regimens.

Model-informed precision dosing of vancomycin for rapid achievement of target area under the concentration-time curve: A simulation study

In this study, we aimed to evaluate limited sampling strategies for achieving the therapeutic ranges of the area under the concentration-time curve (AUC) of vancomycin on the first and second day (AUC0-24 , AUC24-48 , respectively) of therapy. A virtual population of 1000 individuals was created using a population pharmacokinetic (PopPK) model, which was validated and incorporated into our model-informed precision dosing tool. The results were evaluated using six additional PopPK models selected based on a study design of prospective or retrospective data collection with sufficient concentrations. Bayesian forecasting was performed to evaluate the probability of achieving the therapeutic range of AUC, defined as a ratio of estimated/reference AUC within 0.8-1.2. The Bayesian posterior probability of achieving the AUC24-48 range increased from 51.3% (a priori probability) to 77.5% after using two-point sampling at the trough and peak on the first day. Sampling on the first day also yielded a higher Bayesian posterior probability (86.1%) of achieving the AUC0-24 range compared to the a priori probability of 60.1%. The Bayesian posterior probability of achieving the AUC at steady-state (AUCSS ) range by sampling on the first or second day decreased with decreased kidney function. We demonstrated that second-day trough and peak sampling provided accurate AUC24-48 , and first-day sampling may assist in rapidly achieving therapeutic AUC24-48 , although the AUCSS should be re-estimated in patients with reduced kidney function owing to its unreliable predictive performance.

Model-informed precision dosing in vancomycin treatment

Introduction: While vancomycin remains a widely prescribed antibiotic, it can cause ototoxicity and nephrotoxicity, both of which are concentration-associated. Overtreatment can occur when the treatment lasts for an unnecessarily long time. Using a model-informed precision dosing scheme, this study aims to develop a population pharmacokinetic (PK) and pharmacodynamic (PD) model for vancomycin to determine the optimal dosage regimen and treatment duration in order to avoid drug-induced toxicity. Methods: The data were obtained from electronic medical records of 542 patients, including 40 children, and were analyzed using NONMEM software. For PK, vancomycin concentrations were described with a two-compartment model incorporating allometry scaling. Results and discussion: This revealed that systemic clearance decreased with creatinine and blood urea nitrogen levels, history of diabetes and renal diseases, and further decreased in women. On the other hand, the central volume of distribution increased with age. For PD, C-reactive protein (CRP) plasma concentrations were described by transit compartments and were found to decrease with the presence of pneumonia. Simulations demonstrated that, given the model informed optimal doses, peak and trough concentrations as well as the area under the concentration-time curve remained within the therapeutic range, even at doses smaller than routine doses, for most patients. Additionally, CRP levels decreased more rapidly with the higher dose starting from 10 days after treatment initiation. The developed R Shiny application efficiently visualized the time courses of vancomycin and CRP concentrations, indicating its applicability in designing optimal treatment schemes simply based on visual inspection.

Population Pharmacokinetics of Vancomycin in Patients Receiving Hemodialysis in a Malian and a French Center and Simulation of the Optimal Loading Dose

PURPOSE

Vancomycin dosing remains challenging in patients receiving intermittent hemodialysis, especially in developing countries, where access to therapeutic drug monitoring and model-based dose adjustment services is limited. The objectives of this study were to describe vancomycin population PK in patients receiving hemodialysis in a Malian and French center and examine the optimal loading dose of vancomycin in this setting.

METHODS

Population pharmacokinetic analysis was conducted using Pmetrics in 31 Malian and 27 French hemodialysis patients, having a total of 309 vancomycin plasma concentrations. Structural and covariate analyses were based on goodness-of-fit criteria. The final model was used to perform simulations of the vancomycin loading dose, targeting a daily area under the concentration-time curve (AUC) of 400-600 mg.h/L or trough concentration of 15-20 mg/L at 48 hours.

RESULTS

After 48 hours of therapy, 68% of Malian and 63% of French patients exhibited a daily AUC of <400. The final model was a 2-compartment model, with hemodialysis influencing vancomycin elimination and age influencing the vancomycin volume distribution. Younger Malian patients exhibited a lower distribution volume than French patients. Dosing simulation suggested that loading doses of 1500, 2000, and 2500 mg would be required to minimize underexposure in patients aged 30, 50, and 70 years, respectively.

CONCLUSIONS

In this study, a low AUC was frequently observed in hemodialysis patients in Mali and France after a standard vancomycin loading dose. A larger dose is necessary to achieve the currently recommended AUC target. However, the proposed dosing algorithm requires further clinical evaluation.

Optimizing the dosing of vancomycin in patients receiving intermittent haemodialysis with low-flux filters, and the potential impact of dosing software

AIM

Iterative approaches to vancomycin dosing (e.g., dosing when trough concentrations <15-20 mg/L) can be inadequate. Computer-guided dosing may be superior but has not been evaluated in patients with kidney failure receiving replacement therapy. We evaluated vancomycin concentrations using a hospital protocol and a pharmacokinetic software. We measured vancomycin clearance by the FX8 low-flux filter because data are absent.

METHODS

We retrospectively reviewed records of adults with kidney failure requiring replacement therapy receiving vancomycin and dialysed with the FX8 low-flux filter, and calculated the proportion of pre-dialysis concentrations that were within, above or below a specified range. One and two-compartment models in the pharmacokinetic software were assessed by computing mean prediction error (MPE) and root mean square error (RMSE) of observed versus predicted concentrations. Vancomycin extracorporeal clearance was prospectively determined using the extraction method.

RESULTS

In 24 patients (34 courses; 139 paired observed and predicted concentrations), 62/139 (45%) pre-dialysis concentrations were 15-25 mg/L, 29/139 (21%) were above, and 48/139 (35%) were below. MPE for the one-compartment model was -0.2 mg/L, RMSE 5.3 mg/L. MPE for the two-compartment model was 2.0 mg/L, RMSE 5.6 mg/L. Excluding the first paired concentrations, the subsequent MPE (n = 105) using one-compartment model was -0.5 mg/L, RMSE 5.6 mg/L. The MPE for the two-compartment model was 2.1 mg/L, RMSE 5.8 mg/L. The median extracorporeal clearance was 70.7 mL/min (range: 10.3-130.3; n = 22).

CONCLUSIONS

Vancomycin dosing was suboptimal and the pharmacokinetic software was not sufficiently predictive. These may improve with a loading dose. The substantial removal of vancomycin by low-flux filters is not accounted for by the models tested.

In silico Evaluation of a Vancomycin Dosing Guideline Among Adults with Serious Infections

BACKGROUND

This study aimed to compare the achievement of pharmacokinetic-pharmacodynamic (PK-PD) exposure targets for vancomycin using a newly developed dosing guideline with product-information-based dosing in the treatment of adult patients with serious infections.

METHODS

In silico product-information- and guideline-based dosing simulations for vancomycin were performed across a range of doses and patient characteristics, including body weight, age, and renal function at 36-48 and 96 hours, using a pharmacokinetic model derived from a seriously ill patient population. The median simulated concentration and area under the 24-hour concentration-time curve (AUC 0-24 ) were used to measure predefined therapeutic, subtherapeutic, and toxicity PK-PD targets.

RESULTS

Ninety-six dosing simulations were performed. The pooled median trough concentration target with guideline-based dosing at 36 and 96 hours was achieved in 27.1% (13/48) and 8.3% (7/48) of simulations, respectively. The pooled median AUC 0-24 /minimum inhibitory concentration ratio with guideline-based dosing at 48 and 96 hours was attained in 39.6% (19/48) and 27.1% (13/48) of simulations, respectively. Guideline-based dosing simulations yielded improved trough target attainment compared with product-information-based dosing at 36 hours and significantly less subtherapeutic drug exposure. The toxicity threshold was exceeded in 52.1% (25/48) and 0% (0/48) for guideline- and product-information-information-based dosing, respectively ( P < 0.001).

CONCLUSIONS

A Critical care vancomycin dosing guideline appeared slightly more effective than standard dosing, as per product information, in achieving PK-PD exposure associated with an increased likelihood of effectiveness. In addition, this guideline significantly reduced the risk of subtherapeutic exposure. The risk of exceeding toxicity thresholds, however, was greater with the guideline, and further investigation is suggested to improve dosing accuracy and sensitivity.

Acute Kidney Injury Incidence With Bayesian Dosing Software Versus 2-Level First-Order Area Under the Curve-Based Dosing of Vancomycin With Piperacillin-Tazobactam

Background: Two methods of area under the curve (AUC) dosing are recommended in vancomycin consensus guidelines: first-order calculations utilizing 2 vancomycin concentrations or a Bayesian approach. It is unknown if there is a difference in acute kidney injury (AKI) between the 2 dosing strategies for patients receiving concomitant piperacillin-tazobactam and vancomycin (VPT). Objective: The objective of this study was to compare incidence of AKI in patients being administered VPT with first-order calculations versus model-informed precision dosing (MIPD)/Bayesian dosing. Methods: This was a single-center, retrospective, observational study at a community hospital. Patients who received VPT therapy for at least 48 hours were included. The primary outcome was overall incidence of AKI. Secondary outcomes included percentage target attainment with initial regimen, average serum creatinine increase, time to AKI, usable vancomycin levels, and need for temporary dialysis or intensive care unit admission. Results: There were 100 patients included (50 in the first-order group and 50 in the MIPD/Bayesian group). The overall incidence of AKI was lower in the MIPD/Bayesian group (12% vs 28%, P = 0.046). There was no difference in average serum creatinine increase, time to AKI, need for temporary dialysis, or intensive care unit admission. Patients in the MIPD/Bayesian group had a higher percentage of target attainment (46% vs 18%, P = 0.003) and usable vancomycin levels (98% vs 60%, P < 0.001). Conclusion and Relevance: In patients receiving VPT, model-informed precision dosing with Bayesian modeling resulted in a lower rate of AKI, higher target attainment, and more usable vancomycin levels compared with first-order AUC dosing. The small sample and retrospective nature of this study reinforces the need for additional data.

Model-Informed Precision Dosing of Vancomycin in Adult Patients Undergoing Hemodialysis

Model-informed precision dosing (MIPD) maximizes the probability of successful dosing in patients undergoing hemodialysis. In these patients, area under the concentration-time curve (AUC)-guided dosing is recommended for vancomycin. However, this model is yet to be developed. The purpose of this study was to address this issue. The overall mass transfer-area coefficient (KoA) was used for the estimation of vancomycin hemodialysis clearance. A population pharmacokinetic (popPK) model was developed, resulting in a fixed-effect parameter for nonhemodialysis clearance of 0.316 liters/h. This popPK model was externally evaluated, with a resulting mean absolute error of 13.4% and mean prediction error of -0.17%. KoA-predicted hemodialysis clearance was prospectively evaluated for vancomycin (n = 10) and meropenem (n = 10), with a correlation equation being obtained (slope of 1.099, intercept of 1.642; r = 0.927, P < 0.001). An experimental evaluation using an in vitro hemodialysis circuit validated the developed model of KoA-predicted hemodialysis clearance using vancomycin, meropenem, vitamin B6, and inulin in 12 hemodialysis settings. This popPK model indicated a maximum a priori dosing for vancomycin-a loading dose of 30 mg/kg, which achieves the target AUC for 24 h after first dose with a probability of 93.0%, ensured by a predialysis concentration of >15 μg/mL. Maintenance doses of 12 mg/kg after every hemodialysis session could achieve the required exposure, with a probability of 80.6%. In conclusion, this study demonstrated that KoA-predicted hemodialysis clearance may lead to an upgrade from conventional dosing to MIPD for vancomycin in patients undergoing hemodialysis.

Implementing Vancomycin Population Pharmacokinetic Models: An App for Individualized Antibiotic Therapy in Critically Ill Patients

In individualized therapy, the Bayesian approach integrated with population pharmacokinetic models (PopPK) for predictions together with therapeutic drug monitoring (TDM) to maintain adequate objectives is useful to maximize the efficacy and minimize the probability of toxicity of vancomycin in critically ill patients. Although there are limitations to implementation, model-informed precision dosing (MIPD) is an approach to integrate these elements, which has the potential to optimize the TDM process and maximize the success of antibacterial therapy. The objective of this work was to present an app for individualized therapy and perform a validation of the implemented vancomycin PopPK models. A pragmatic approach was used for selecting the models of Llopis, Goti and Revilla for developing a Shiny app with R. Through ordinary differential equation (ODE)-based mixed effects models from the mlxR package, the app simulates the concentrations' behavior, estimates whether the model was simulated without variability and predicts whether the model was simulated with variability. Moreover, we evaluated the predictive performance with retrospective trough concentration data from patients admitted to the adult critical care unit. Although there were no significant differences in the performance of the estimates, the Llopis model showed better accuracy (mean 80.88%; SD 46.5%); however, it had greater bias (mean -34.47%, SD 63.38%) compared to the Revilla et al. (mean 10.61%, SD 66.37%) and Goti et al. (mean of 13.54%, SD 64.93%) models. With respect to the RMSE (root mean square error), the Llopis (mean of 10.69 mg/L, SD 12.23 mg/L) and Revilla models (mean of 10.65 mg/L, SD 12.81 mg/L) were comparable, and the lowest RMSE was found in the Goti model (mean 9.06 mg/L, SD 9 mg/L). Regarding the predictions, this behavior did not change, and the results varied relatively little. Although our results are satisfactory, the predictive performance in recent studies with vancomycin is heterogeneous, and although these three models have proven to be useful for clinical application, further research and adaptation of PopPK models is required, as well as implementation in the clinical practice of MIPD and TDM in real time.

Hemoperfusion with Seraph 100 Microbind Affinity Blood Filter Unlikely to Require Increased Antibiotic Dosing: A Simulations Study Using a Pharmacokinetic/Pharmacodynamic Approach

INTRODUCTION

The Seraph® 100 Microbind® Affinity Blood Filter (Seraph 100) is a hemoperfusion device that can remove pathogens from central circulation. However, the effect of Seraph 100 on achieving pharmacodynamic (PD) targets is not well described. We sought to determine the impact of Seraph 100 on ability to achieve PD targets for commonly used antibiotics.

METHODS

Estimates of Seraph 100 antibiotic clearance were obtained via literature. For vancomycin and gentamicin, published pharmacokinetic models were used to explore the effect of Seraph 100 on ability to achieve probability of target attainment (PTA). For meropenem and imipenem, the reported effect of continuous kidney replacement therapy (CKRT) on achieving PTA was used to extrapolate decisions for Seraph 100.

RESULTS

Seraph 100 antibiotic clearance is likely less than 0.5 L/h for most antibiotics. Theoretical Seraph 100 clearance up to 0.5 L/h and 2 L/h had a negligible effect on vancomycin PTA in virtual patients with creatinine clearance (CrCl) = 14 mL/min and CrCl >14 mL/min, respectively. Theoretical Seraph 100 clearance up to 0.5 L/h and 2 L/h had a negligible effect on gentamicin PTA in virtual patients with CrCl = 120 mL/min and CrCl <60 mL/min, respectively. CKRT intensity resulting in antibiotic clearance up to 2 L/h generally does not require dose increases for meropenem or imipenem. As Seraph 100 is prescribed intermittently and likely contributes far less to antibiotic clearance, dose increases would also not be required.

CONCLUSION

Seraph 100 clearance of vancomycin, gentamicin, meropenem, and imipenem is likely clinically insignificant. There is insufficient evidence to recommend increased doses. For aminoglycosides, we recommend extended interval dosing and initiating Seraph 100 at least 30 min to 1 h after completion of infusion to avoid the possibility of interference with maximum concentrations.

Bayesian method application: Integrating mathematical modeling into clinical pharmacy through vancomycin therapeutic monitoring

The most recent consensus guidelines for dosing and monitoring vancomycin recommended the use of area-under-the-curve with Bayesian estimation for therapeutic monitoring. As this is a modern concept in the practice of clinical pharmacy, the main objective of this review is to introduce the fundamentals of Bayesian estimation and its mathematical application as it relates to vancomycin therapeutic drug monitoring. In addition, we aim to identify pharmacokinetic (PK) software programs that incorporate Bayesian estimation for vancomycin dosing and to describe the PK models utilized in those software programs for the adult population. Twelve software programs that utilize Bayesian estimation were identified, which included: Adult and Pediatric Kinetics, Best Dose, ClinCalc, DoseMeRx, ID-ODS, InsightRx, MwPharm++, NextDose, PrecisePK, TDMx, Tucuxi, and VancoCalc. The software programs varied in the population PK models used as the Bayesian a priori. With the presence of various vancomycin Bayesian software programs, it is important to choose those that utilize PK models reflective of the specific patient population.

Discrepancies Between Bayesian Vancomycin Models Can Affect Clinical Decisions in the Critically Ill

Purpose

To assess the agreement in 24-hour area under the curve (AUC₂₄) value estimates between commonly used vancomycin population pharmacokinetic models in the critically ill.

Materials and Methods

Adults admitted to intensive care who received intravenous vancomycin and had a serum vancomycin concentration available were included. AUC₂₄ values were determined using Tucuxi (revision cd7bd7a8) for dosing intervals with a vancomycin concentration using three models (Goti 2018, Colin 2019, and Thomson 2009) previously evaluated in the critically ill. AUC₂₄ values were categorized as subtherapeutic (<400 mg·h/L), therapeutic (400-600 mg·h/L), or toxic (>600 mg·h/L), assuming a minimum inhibitory concentration of 1 mg/L. AUC₂₄ value categorization was compared across the three models and reported as percent agreement.

Results

Overall, 466 AUC₂₄ values were estimated in 188 patients. Overall, 52%, 42%, and 47% of the AUC₂₄ values were therapeutic for the Goti, Colin, and Thomson models, respectively. The agreement of AUC₂₄ values between all three models was 48% (223/466), Goti-Colin 59% (193/466), Goti-Thomson 68% (318/466), and Colin-Thomson 67% (314/466).

Conclusion

In critically ill patients, vancomycin AUC₂₄ values obtained from different pharmacokinetic models are often discordant, potentially contributing to differences in dosing decisions. This highlights the importance of selecting the optimal model.

Systematic Comparison of Hospital-Wide Standard and Model-Based Therapeutic Drug Monitoring of Vancomycin in Adults

We aimed to evaluate the predictive performance and predicted doses of a single-model approach or several multi-model approaches compared with the standard therapeutic drug monitoring (TDM)-based vancomycin dosing. We performed a hospital-wide monocentric retrospective study in adult patients treated with either intermittent or continuous vancomycin infusions. Each patient provided two randomly selected pairs of two consecutive vancomycin concentrations. A web-based precision dosing software, TDMx, was used to evaluate the model-based approaches. In total, 154 patients contributed 308 pairs. With standard TDM-based dosing, only 48.1% (148/308) of all of the second concentrations were within the therapeutic range. Across the model-based approaches we investigated, the mean relative bias and relative root mean square error varied from −5.36% to 3.18% and from 24.8% to 28.1%, respectively. The model averaging approach according to the squared prediction errors showed an acceptable bias and was the most precise. According to this approach, the median (interquartile range) differences between the model-predicted and prescribed doses, expressed as mg every 12 h, were 113 [−69; 427] mg, −70 [−208; 120], mg and 40 [−84; 197] mg in the case of subtherapeutic, supratherapeutic, and therapeutic exposure at the second concentration, respectively. These dose differences, along with poor target attainment, suggest a large window of opportunity for the model-based TDM compared with the standard TDM-based vancomycin dosing. Implementation studies of model-based TDM in routine care are warranted.

Application of Machine Learning Classification to Improve the Performance of Vancomycin Therapeutic Drug Monitoring

Bayesian therapeutic drug monitoring (TDM) software uses a reported pharmacokinetic (PK) model as prior information. Since its estimation is based on the Bayesian method, the estimation performance of TDM software can be improved using a PK model with characteristics similar to those of a patient. Therefore, we aimed to develop a classifier using machine learning (ML) to select a more suitable vancomycin PK model for TDM in a patient. In our study, nine vancomycin PK studies were selected, and a classifier was created to choose suitable models among them for patients. The classifier was trained using 900,000 virtual patients, and its performance was evaluated using 9000 and 4000 virtual patients for internal and external validation, respectively. The accuracy of the classifier ranged from 20.8% to 71.6% in the simulation scenarios. TDM using the ML classifier showed stable results compared with that using single models without the ML classifier. Based on these results, we have discussed further development of TDM using ML. In conclusion, we developed and evaluated a new method for selecting a PK model for TDM using ML. With more information, such as on additional PK model reporting and ML model improvement, this method can be further enhanced.

Comparison of Bayesian-derived and first-order analytic equations for calculation of vancomycin area under the curve

INTRODUCTION

Consensus guidelines recommend targeting a vancomycin area under the curve to minimum inhibitory concentration (AUC₂₄ :MIC) ratio of 400-600 to improve therapeutic success and reduce nephrotoxicity. Although guidelines specify either Bayesian software or first-order equations may be used to estimate AUC₂₄ , there are currently no large studies directly comparing these methods.

OBJECTIVE

To compare calculated vancomycin AUC₂₄ using first-order equations with two-drug concentrations at steady state to Bayesian two- and one-concentration estimations.

METHODS

This was a single-center, retrospective cohort study of 978 adult hospitalized patients receiving intravenous vancomycin between 2017 and 2019. Patients were included if they received at least 72 h of vancomycin and had two-serum drug concentrations obtained. AUC₂₄ was calculated using first-order analytic (linear), Bayesian two-concentration, and Bayesian one-concentration methods for each patient. The InsightRx™ software platform was used to calculate Bayesian AUC₂₄ . Pearson's correlation and clinical agreement (based on AUC₂₄  classified as subtherapeutic, therapeutic, or supratherapeutic) were used to assess agreement between methods. Bland-Altman plots were used to assess mean difference (MD) and 95% limits of agreement (LOA).

RESULTS

Excellent agreement was observed between linear and Bayesian two-concentration methods (r = 0.963, clinical agreement = 87.4%) and Bayesian two-concentration and one-concentration methods (r = 0.931, clinical agreement = 88.5%); however, a degree of variability was noted with 95% LOA -99 to 76 (MD = -11.5 mg*h/L) and -92 to 113 (MD = -10.4 mg*h/L), for the respective comparisons. The agreement between linear and Bayesian one-concentration approaches was less than prior comparisons (r = 0.823, clinical agreement = 76.8%) and demonstrated the greatest amount of variability with 95% LOA -197 to 153 (MD = -21.9 mg*h/L).

CONCLUSIONS

Linear and Bayesian two-concentration methods demonstrated high-level agreement with acceptable variability and may be considered comparable to estimate vancomycin AUC₂₄ . As linear and Bayesian one-concentration methods demonstrated significant variability and suboptimal agreement, concerns exist surrounding the interchangeability of these methods in clinical practice, particularly at higher extremes of AUC₂₄ .

A Free Open-Source Bayesian Vancomycin Dosing App for Adults: Design and Evaluation Study

Background

It has been suggested that Bayesian dosing apps can assist in the therapeutic drug monitoring of patients receiving vancomycin. Unfortunately, Bayesian dosing tools are often unaffordable to resource-limited hospitals. Our aim was to improve vancomycin dosing in adults. We created a free and open-source dose adjustment app, VancoCalc, which uses Bayesian inference to aid clinicians in dosing and monitoring of vancomycin.

Objective

The aim of this paper is to describe the design, development, usability, and evaluation of a free open-source Bayesian vancomycin dosing app, VancoCalc.

Methods

The app build and model fitting process were described. Previously published pharmacokinetic models were used as priors. The ability of the app to predict vancomycin concentrations was performed using a small data set comprising of 52 patients, aged 18 years and over, who received at least 1 dose of intravenous vancomycin and had at least 2 vancomycin concentrations drawn between July 2018 and January 2021 at Lakeridge Health Corporation Ontario, Canada. With these estimated and actual concentrations, median prediction error (bias), median absolute error (accuracy), and root mean square error (precision) were calculated to evaluate the accuracy of the Bayesian estimated pharmacokinetic parameters.

Results

A total of 52 unique patients’ initial vancomycin concentrations were used to predict subsequent concentration; 104 total vancomycin concentrations were assessed. The median prediction error was –0.600 ug/mL (IQR –3.06, 2.95), the median absolute error was 3.05 ug/mL (IQR 1.44, 4.50), and the root mean square error was 5.34.

Conclusions

We described a free, open-source Bayesian vancomycin dosing calculator based on revisions of currently available calculators. Based on this small retrospective preliminary sample of patients, the app offers reasonable accuracy and bias, which may be used in everyday practice. By offering this free, open-source app, further prospective validation could be implemented in the near future.

Model-informed precision dosing of vancomycin via continuous infusion: a clinical fit-for-purpose evaluation of published PK models

BACKGROUND

Model-informed precision dosing (MIPD) is an innovative approach used to guide bedside vancomycin dosing. The use of Bayesian software requires suitable and externally validated population pharmacokinetic (popPK) models.

OBJECTIVES

Therefore, we aimed to identify suitable popPK models for a priori prediction and a posteriori forecasting of vancomycin in continuous infusion. Additionally, a model averaging (MAA) and a model selection approach (MSA) were compared with the identified popPK models.

METHODS

. Clinical PK data were retrospectively collected from patients receiving continuous vancomycin therapy and admitted to a general ward of three large Belgian hospitals. The predictive performance of the popPK models, identified in a systematic literature search, as well as the MAA/MSA was evaluated for the a priori and a posteriori scenarios using bias, root mean square errors, normalized prediction distribution errors and visual predictive checks.

RESULTS

The predictive performance of 23 popPK models was evaluated based on clinical data from 169 patients and 923 therapeutic drug monitoring samples. Overall, the best predictive performance was found using the Okada model (bias < -0.1 mg/L), followed by the Colin model. The MAA/MSA predicted with a constantly high precision and low inaccuracy and were clinically acceptable in the Bayesian forecasting.

CONCLUSION

This study identified the two-compartmental models of Okada et al. and Colin et al. as most suitable for non-ICU patients to forecast individual exposure profiles after continuous vancomycin infusion. The MAA/MSA performed equally good as the individual popPK models. Both approaches could therefore be used in clinical practice to guide dosing decisions.

Comparison of area under the curve for vancomycin from one- and two-compartment models using sparse data

BACKGROUND AND OBJECTIVE

Vancomycin pharmacokinetics have been described by both one- and two-compartment models. One-compartment models are widely used to predict the area under the curve (AUC), a useful parameter for determining the efficacy and safety of vancomycin, based on sparse data collected during therapeutic drug monitoring. It is uncertain whether AUCs from one-compartment models with sparsely sampled data can sufficiently represent the true AUC. This study aimed to compare AUC estimates from one- and two-compartment models using sparse data. The reliability of AUCs from models constructed with trough-only data was also assessed.

METHODS

A previously published robust model was used to simulate vancomycin concentration points at 15 min intervals in 100 patients. From these simulated data, the reference AUC (AUC_ref) was calculated and two depleted dataset versions (trough-only and peak-trough datasets) were also created. One- and two-compartment models were built from the depleted datasets with the use of NONMEM. Vancomycin 24-hour AUC was calculated from concentration-time profiles of each model by a linear trapezoidal formula at three different time periods: 0-24 hours (AUC_0-24), 24-48 hours (AUC_24-48) and 0-48 hours (AUC_avg). The deviation of each of the AUCs from the AUC_ref was examined to assess the AUC predictability of models from sparse data. The difference in AUCs between one- and two-compartment models was analysed from statistical and clinical perspectives.

RESULTS

When assessing the deviation of each AUC from the AUC_ref, the one-compartment model from both peak-trough and trough-only data could adequately represent the true AUC with no statistically significant differences. Two-compartment model from peak-trough data also provided similar AUC estimates with the AUCref. However, AUCs from the two-compartment model with trough-only data did not adequately represent the true AUC, with significant differences of 25.16% for AUC_0-24, 15.92% for AUC_24-48 and 19.45% for AUC_avg.

CONCLUSION

Regardless of statistically significant differences between AUCs from one- and two-compartment models, the level of difference was acceptable from the clinical perspective, being <17% in models from peak-trough data. Therefore, both one- and two-compartment models with sparse data having at least a pair of peak-trough data per patient could be reliable for predicting AUC. Furthermore, AUCs of the one-compartment model from trough-only data did not show a significant difference from the AUC_ref. Hence, one-compartment models developed from trough-only data could be useful for predicting AUC when models with rich data are not available for the intended population. However, it is suggested that the use of the two-compartment model built from trough-only data should be avoided.

Clinical Practice Guidelines for Therapeutic Drug Monitoring of Vancomycin in the Framework of Model-Informed Precision Dosing: A Consensus Review by the Japanese Society of Chemotherapy and the Japanese Society of Therapeutic Drug Monitoring

Background: To promote model-informed precision dosing (MIPD) for vancomycin (VCM), we developed statements for therapeutic drug monitoring (TDM). Methods: Ten clinical questions were selected. The committee conducted a systematic review and meta-analysis as well as clinical studies to establish recommendations for area under the concentration-time curve (AUC)-guided dosing. Results: AUC-guided dosing tended to more strongly decrease the risk of acute kidney injury (AKI) than trough-guided dosing, and a lower risk of treatment failure was demonstrated for higher AUC/minimum inhibitory concentration (MIC) ratios (cut-off of 400). Higher AUCs (cut-off of 600 μg·h/mL) significantly increased the risk of AKI. Although Bayesian estimation with two-point measurement was recommended, the trough concentration alone may be used in patients with mild infections in whom VCM was administered with q12h. To increase the concentration on days 1–2, the routine use of a loading dose is required. TDM on day 2 before steady state is reached should be considered to optimize the dose in patients with serious infections and a high risk of AKI. Conclusions: These VCM TDM guidelines provide recommendations based on MIPD to increase treatment response while preventing adverse effects.

Comparison of Mw\Pharm 3.30 and Mw\Pharm ++, a Windows version of pharmacokinetic software for PK/PD monitoring of vancomycin. Part 1: A-posteriori modelling

BACKGROUND AND OBJECTIVES

For a long time, the Mw\Pharm software suite (MEDI\WARE, Prague, Czech Republic/ Groningen, Netherlands) has been used for PK/PD modelling in therapeutic drug monitoring (TDM). The aim of this study was to find the best model in the newer Windows Mw\Pharm++ 1.3.5.558 version (WIN).

METHODS

25 patients were repeatedly examined for vancomycin (mean age 63±14 years, body weight 88±21 kg, median dose 1 g/12 h). Trough concentrations predicted a-posteriori by WIN models "vancomycin_adult_k_C2", "#vancomycin_adult_C2", "vancomycin_adult_C2" were compared with the measured value and "vancomycin adult" DOS 3.30 model (DOS).

STATISTICS

Percentage prediction error (%PE) calculated as (predicted-measured)/measured values, or WIN-DOS/DOS - data presented as mean±SD, RMSE, Blandt-Altman plot - data presented as bias±SD (95% limits of agreement), Pearson's coefficient of rank correlation (R), Student's t-test. Statistical analysis was performed using GraphPad Prism version 5.00 for Windows.

RESULTS

The mean%PE in vancomycin predicted values varied from -4.5% ± 33.6 to -8.2% ± 39.3. The%PE between WIN and DOS models varied from -0.2% ± 24.5% to 4.4 ± 21.4%. Model "vancomycin_adult_C2" was closest both to measured vancomycin trough concentration and DOS model:%PE -4.5 ± 33.6% vs +4.2 ± 20.3%, RMSE 33.7 vs 20.6, Blandt-Altman bias +2.19 ± 6.17 (-9.9 - 14.3) vs -0.29 ± 3.25 (-6.7 - 6.1), resp. "#vancomycin_adult_C2" model produced largest%PE (-8.2%), RMSE (40.0) as well as Blandt-Altman bias +2.82 ± 6.76 (-10.4 - 16.1). The Pearson's R of predicted and measured vancomycin concentration, and of values predicted by WIN and DOS models, varied from 0.5135 to 0.5854, P<0.0001 and from 0.7869 to 0.8462, P<0.0001, resp.

CONCLUSIONS

Three Windows vancomycin models and one DOS model in the Mw\Pharm software were compared. The best outcomes, i.e. lowest%PE, RMSE and highest Pearson's R, were reached with "vancomycin_adult_C2" model.

Establishment and application of population pharmacokinetics model of vancomycin in infants with meningitis

BACKGROUND

To establish a population pharmacokinetics (PPK) model of vancomycin (VCM) for dose individualization in Chinese infants with meningitis.

METHODS

We collected the data of 82 children with meningitis in hospital from July 2014 to June 2016. The initial vancomycin dosage regimen for children was 10 or 15 mg/kg for q12 h, q8 h or q6 h. Serum concentrations were determined by Viva-E Analyzer before and after the fifth administration. The PPK model was developed by nonlinear mixed-effect model software, assessed by the bootstrap method and then tested in 20 infant patients.

RESULTS

The VCM clearance (CL) was increased by body weight (WT) and decreased by blood urea nitrogen (BUN). Pharmacokinetic parameters of VCM were not influenced by co-administered drugs. The trough concentrations of VCM were accurately predicted by the PPK model, with the prediction errors less than 32%.

CONCLUSION

A new individual strategy for VCM regimens was proposed and validated by the PPK model.

Dose Tailoring of Vancomycin Through Population Pharmacokinetic Modeling Among Surgical Patients in Pakistan

Background: Vancomycin is a narrow therapeutic agent, and it is necessary to optimize the dose to achieve safe therapeutic outcomes. The purpose of this study was to identify the significant covariates for vancomycin clearance and to optimize the dose among surgical patients in Pakistan. Methods: Plasma concentration data of 176 samples collected from 58 surgical patients treated with vancomycin were used in this study. A population pharmacokinetic model was developed on NONMEM^® using plasma concentration-time data. The effect of all available covariates was evaluated on the pharmacokinetic parameters of vancomycin by stepwise covariate modeling. The final model was evaluated using bootstrap, goodness-of-fit plots, and visual predictive checks. Results: The pharmacokinetics of vancomycin followed a one-compartment model with first-order elimination. The vancomycin clearance (CL) and volume of distribution (Vd) were 2.45 L/h and 22.6 l, respectively. Vancomycin CL was influenced by creatinine clearance (CRCL) and body weight of the patients; however, no covariate was significant for its effect on the volume of distribution. Dose tailoring was performed by simulating dosage regimens at a steady state based on the CRCL of the patients. The tailored doses were 400, 600, 800, and 1,000 mg for patients with a CRCL of 20, 60, 100, and 140 ml/min, respectively. Conclusion: Vancomycin CL is influenced by CRCL and body weight of the patient. This model can be helpful for the dose tailoring of vancomycin based on renal status in Pakistani patients.

Population pharmacokinetics of cefotaxime in intensive care patients

PURPOSE

To characterise the pharmacokinetics and associated variability of cefotaxime in adult intensive care unit (ICU) patients and to assess the impact of patient covariates.

METHODS

This work was based on data from cefotaxime-treated patients included in the ACCIS (Antibiotic Concentrations in Critical Ill ICU Patients in Sweden) study. Clinical data from 51 patients at seven different ICUs in Sweden, given cefotaxime (1000-3000 mg given 2-6 times daily), were collected from the first day of treatment for up to three consecutive days. In total, 263 cefotaxime samples were included in the population pharmacokinetic analysis.

RESULTS

A two-compartment model with linear elimination, proportional residual error and inter-individual variability (IIV) on clearance and central volume of distribution best described the data. The typical individual was 64 years, with body weight at ICU admission of 92 kg and estimated creatinine clearance of 94 mL/min. The resulting typical value of clearance was 11.1 L/h, central volume of distribution 5.1 L, peripheral volume of distribution 18.2 L and inter-compartmental clearance 14.5 L/h. The estimated creatinine clearance proved to be a significant covariate on clearance (p < 0.001), reducing IIV from 68 to 49%.

CONCLUSION

A population pharmacokinetic model was developed to describe cefotaxime pharmacokinetics and associated variability in adult ICU patients. The estimated creatinine clearance partly explained the IIV in cefotaxime clearance. However, the remaining unexplained IIV is high and suggests a need for dose individualisation using therapeutic drug monitoring where the developed model, after evaluation of predictive performance, may provide support.

A hybrid machine learning/pharmacokinetic approach outperforms maximum a posteriori Bayesian estimation by selectively flattening model priors

Model‐informed precision dosing (MIPD) approaches typically apply maximum a posteriori (MAP) Bayesian estimation to determine individual pharmacokinetic (PK) parameters with the goal of optimizing future dosing regimens. This process combines knowledge about the individual, in the form of drug levels or pharmacodynamic biomarkers, with prior knowledge of the drug PK in the general population. Use of “flattened priors” (FPs), in which the weight of the model priors is reduced relative to observations about the patient, has been previously proposed to estimate individual PK parameters in instances where the patient is poorly described by the PK model. However, little is known about the predictive performance of FPs and when to apply FPs in MIPD. Here, FP is evaluated in a data set of 4679 adult patients treated with vancomycin. Depending on the PK model, prediction error could be reduced by applying FPs in 42–55% of PK parameter estimations. Machine learning (ML) models could identify instances where FPs would outperform MAPs with a specificity of 81–86%, reducing overall root mean squared error (RMSE) of PK model predictions by 12–22% (0.5–1.2 mg/L) relative to MAP alone. The factors most indicative of the use of FPs were past prediction residuals and bias in past PK predictions. A more clinically practical minimal model was developed using only these two features, reducing RMSE by 5–18% (0.20–0.93 mg/L) relative to MAP. This hybrid ML/PK approach advances the precision dosing toolkit by leveraging the power of ML while maintaining the mechanistic insight and interpretability of PK models.

Predictive Performance of Bayesian Vancomycin Monitoring in the Critically Ill

OBJECTIVES

It is recommended that therapeutic monitoring of vancomycin should be guided by 24-hour area under the curve concentration. This can be done via Bayesian models in dose-optimization software. However, before these models can be incorporated into clinical practice in the critically ill, their predictive performance needs to be evaluated. This study assesses the predictive performance of Bayesian models for vancomycin in the critically ill.

DESIGN

Retrospective cohort study.

SETTING

Single-center ICU.

PATIENTS

Data were obtained for all patients in the ICU between 1 January, and 31 May 2020, who received IV vancomycin. The predictive performance of three Bayesian models were evaluated based on their availability in commercially available software. Predictive performance was assessed via bias and precision. Bias was measured as the mean difference between observed and predicted vancomycin concentrations. Precision was measured as the SD of bias, root mean square error, and 95% limits of agreement based on Bland-Altman plots.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

A total of 466 concentrations from 188 patients were used to evaluate the three models. All models showed low bias (-1.7 to 1.8 mg/L), which was lower with a posteriori estimate (-0.7 to 1.8 mg/L). However, all three models showed low precision in terms of SD (4.7-8.8 mg/L) and root mean square error (4.8-8.9 mg/L). The models underpredicted at higher observed vancomycin concentrations (bias 0.7-3.2 mg/L for < 20 mg/L; -5.1 to -2.3 for ≥ 20 mg/L) and the Bland-Altman plots showed a great deviation between observed and predicted concentrations.

CONCLUSIONS

Bayesian models of vancomycin show not only low bias, but also low precision in the critically ill. Thus, Bayesian-guided dosing of vancomycin in this population should be used cautiously.

Accuracy of documented administration times for intravenous antimicrobial drugs and impact on dosing decisions

AIMS

Accurate documentation of medication administration time is imperative for many therapeutic decisions, including dosing of intravenous antimicrobials. The objectives were to determine (1) the discrepancy between actual and documented administration times for antimicrobial infusions and (2) whether day of the week, time of day, nurse-to-patient ratio and drug impacted accuracy of documented administration times.

METHODS

Patient and dosing data were collected (June-August 2019) for 55 in-patients receiving antimicrobial infusions. "Documented" and "actual" administration times (n = 660) extracted from electronic medication management systems and smart infusion pumps, respectively, were compared. Influence of the day (weekday/weekend), time of day (day/evening/night), nurse-to-patient ratio (high 1:1/low 1:5) and drug were examined. Monte Carlo simulation was used to predict the impact on dose adjustments for vancomycin using the observed administration time discrepancies compared to the actual administration time.

RESULTS

The median discrepancy between actual and documented administration times was 16 min (range, 2-293 min), with discrepancies greater than 60 minutes in 7.7% of administrations. Overall, discrepancies (median [range]) were similar on weekends (17 [2-293] min) and weekdays (16 [2-188] min), and for high (16 [2-157] min) and low nurse-to-patient ratio wards (16 [2-293] min). Discrepancies were smallest for night administrations (P < .05), and antimicrobials with shorter half-lives (P < .0001). The observed discrepancies in vancomycin administration time resulted in a different dose recommendation in 58% of cases (30% higher, 28% lower).

CONCLUSIONS

Overall, there were discrepancies between actual and documented antimicrobial infusion administration times. For vancomycin, these discrepancies in administration time were predicted to result in inappropriate dose recommendations.

Population pharmacokinetic modeling and clinical application of vancomycin in Chinese patients hospitalized in intensive care units

Management of vancomycin administration for intensive care units (ICU) patients remains a challenge. The aim of this study was to describe a population pharmacokinetic model of vancomycin for optimizing the dose regimen for ICU patients. We prospectively enrolled 466 vancomycin-treated patients hospitalized in the ICU, collected trough or approach peak blood samples of vancomycin and recorded corresponding clinical information from July 2015 to December 2017 at Tai Zhou Hospital of Zhejiang Province. The pharmacokinetics of vancomycin was analyzed by nonlinear mixed effects modeling with Kinetica software. Internal and external validation was evaluated by the maximum likelihood method. Then, the individual dosing regimens of the 92 patients hospitalized in the ICU whose steady state trough concentrations exceeded the target range (10–20 μg/ml) were adjusted by the Bayes feedback method. The final population pharmacokinetic model show that clearance rate (CL) of vancomycin will be raised under the conditions of dopamine combined treatment, severe burn status (Burn-S) and increased total body weight (TBW), but reduced under the conditions of increased serum creatinine (Cr) and continuous renal replacement therapy status; Meanwhile, the apparent distribution volume (V) of vancomycin will be enhanced under the terms of increased TBW, however decreased under the terms of increased age and Cr. The population pharmacokinetic parameters (CL and V) according to the final model were 3.16 (95%CI 2.83, 3.40) L/h and 60.71 (95%CI 53.15, 67.46). The mean absolute prediction error for external validation by the final model was 12.61% (95CI 8.77%, 16.45%). Finally, the prediction accuracy of 90.21% of the patients’ detected trough concentrations that were distributed in the target range of 10–20 μg/ml after dosing adjustment was found to be adequate. There is significant heterogeneity in the CL and V of vancomycin in ICU patients. The constructed model is sufficiently precise for the Bayesian dose prediction of vancomycin concentrations for the population of ICU Chinese patients.

Do Vancomycin Pharmacokinetics Differ Between Obese and Non-obese Patients? Comparison of a General-Purpose and Four Obesity-Specific Pharmacokinetic Models

Over the past decade, numerous obesity-specific pharmacokinetic (PK) models and dosage regimens have been developed. However, it is unclear whether vancomycin PKs differ between obese and other patients after accounting for weight, age, and kidney function. In this study, the authors investigated whether using obesity-specific population PK models for vancomycin offers any advantage in accuracy and precision over using a recently developed general-purpose model.

A Model Averaging/Selection Approach Improves the Predictive Performance of Model-Informed Precision Dosing: Vancomycin as a Case Study

Many important drugs exhibit substantial variability in pharmacokinetics and pharmacodynamics leading to a loss of the desired clinical outcomes or significant adverse effects. Forecasting drug exposures using pharmacometric models can improve individual target attainment when compared with conventional therapeutic drug monitoring (TDM). However, selecting the "correct" model for this model-informed precision dosing (MIPD) is challenging. We derived and evaluated a model selection algorithm (MSA) and a model averaging algorithm (MAA), which automates model selection and finds the best model or combination of models for each patient using vancomycin as a case study, and implemented both algorithms in the MIPD software "TDMx." The predictive performance (based on accuracy and precision) of the two algorithms was assessed in (i) a simulation study of six distinct populations and (ii) a clinical dataset of 180 patients undergoing TDM during vancomycin treatment and compared with the performance obtained using a single model. Throughout the six virtual populations the MSA and MAA (imprecision: 9.9-24.2%, inaccuracy: less than ± 8.2%) displayed more accurate predictions than the single models (imprecision: 8.9-51.1%; inaccuracy: up to 28.9%). In the clinical dataset, the predictive performance of the single models applying at least one plasma concentration varied substantially (imprecision: 28-62%, inaccuracy: -16 to 25%), whereas the MSA or MAA utilizing these models simultaneously resulted in unbiased and precise predictions (imprecision: 29% and 30%, inaccuracy: -5% and 0%, respectively). MSA and MAA approaches implemented in TDMx might thereby lower the burden of fit-for-purpose validation of individual models and streamline MIPD.

Population Pharmacokinetic Modeling of Vancomycin in Thai Patients With Heterogeneous and Unstable Renal Function

BACKGROUND

Vancomycin is widely used to treat gram-positive bacterial infections. However, given significant interpatient variability in its pharmacokinetics, maintaining plasma concentrations is difficult within its characteristically narrow therapeutic window. This is especially challenging in patients with unstable renal function. Thus, the aim of this study was to develop a population pharmacokinetic model for vancomycin that is suitable for Thai patients with variable renal functions, including those with unstable renal function.

METHODS

Data from 213 patients, including 564 blood samples, were retrospectively collected; approximately 70% patients exhibited unstable renal function during vancomycin treatment. The model building group was randomly assigned 108 patients and the remaining 33 patients comprised the validation group. A population pharmacokinetic model was developed that incorporated drug clearance (CL) as a function of time-varying creatine clearance (CrCL). The predictive ability of the resulting population model was evaluated using the validation data set, including its ability to forecast serum concentrations within a Bayesian feedback algorithm.

RESULTS

A 2-compartment model with drug CL values that changed with time-varying CrCL adequately described vancomycin pharmacokinetics in the evaluated heterogeneous patient population with unstable renal function. Vancomycin CL was related to time-varying CrCL as follows: CL (t) = 0.11 + 0.021 × CrCL (t) (CrCL <120 mL/min. Using the population model, Bayesian estimation with at least one measured serum concentration resulted in a forecasting error of small bias (-2.4%) and adequate precision (31.5%).

CONCLUSIONS

In hospitals with a high incidence of unstable renal function, incorporating time-varying CrCL with Bayesian estimation and at least one measured drug concentration, along with frequent CrCL monitoring, improves the predictive performance of therapeutic drug monitoring of vancomycin.

Assessing the accuracy of two Bayesian forecasting programs in estimating vancomycin drug exposure

Background

Current guidelines for intravenous vancomycin identify drug exposure (as indicated by the AUC) as the best pharmacokinetic (PK) indicator of therapeutic outcome.

Objectives

To assess the accuracy of two Bayesian forecasting programs in estimating vancomycin AUC0–∞ in adults with limited blood concentration sampling.

Methods

The application of seven vancomycin population PK models in two Bayesian forecasting programs was examined in non-obese adults (n = 22) with stable renal function. Patients were intensively sampled following a single (1000 mg or 15 mg/kg) dose. For each patient, AUC was calculated by fitting all vancomycin concentrations to a two-compartment model (defined as AUCTRUE). AUCTRUE was then compared with the Bayesian-estimated AUC0–∞ values using a single vancomycin concentration sampled at various times post-infusion.

Results

Optimal sampling times varied across different models. AUCTRUE was generally overestimated at earlier sampling times and underestimated at sampling times after 4 h post-infusion. The models by Goti et al. (Ther Drug Monit 2018;

40

212–21) and Thomson et al. (J Antimicrob Chemother 2009;

63

1050–7) had precise and unbiased sampling times (defined as mean imprecision &lt;25% and &lt;38 mg·h/L, with 95% CI for mean bias containing zero) between 1.5 and 6 h and between 0.75 and 2 h post-infusion, respectively. Precise but biased sampling times for Thomson et al. were between 4 and 6 h post-infusion.

Conclusions

When using a single vancomycin concentration for Bayesian estimation of vancomycin drug exposure (AUC), the predictive performance was generally most accurate with sample collection between 1.5 and 6 h after infusion, though optimal sampling times varied across different population PK models.

Towards precision dosing of vancomycin in critically ill patients: an evaluation of the predictive performance of pharmacometric models in ICU patients

OBJECTIVES

Vancomycin dose recommendations depend on population pharmacokinetic models. These models have not been adequately assessed in critically ill patients, who exhibit large pharmacokinetic variability. This study evaluated model predictive performance in intensive care unit (ICU) patients and identified factors influencing model performance.

METHODS

Retrospective data from ICU adult patients administered vancomycin were used to evaluate model performance to predict serum concentrations a priori (no observed concentrations included) or with Bayesian forecasting (using concentration data). Predictive performance was determined using relative bias (rBias, bias) and relative root mean squared error (rRMSE, precision). Models were considered clinically acceptable if rBias was between ±20% and 95% confidence intervals included zero. Models were compared with rRMSE; no threshold was used. The influence of clinical factors on model performance was assessed with multiple linear regression.

RESULTS

Data from 82 patients were used to evaluate 12 vancomycin models. The Goti model was the only clinically acceptable model with both a priori (rBias 3.4%) and Bayesian forecasting (rBias 1.5%) approaches. Bayesian forecasting was superior to a priori prediction, improving with the use of more recent concentrations. Four models were clinically acceptable with Bayesian forecasting. Renal replacement therapy status (p < 0.001) and sex (p = 0.007) significantly influenced the performance of the Goti model.

CONCLUSIONS

The Goti, Llopis and Roberts models are clinically appropriate to inform vancomycin dosing in critically ill patients. Implementing the Goti model in dose prediction software could streamline dosing across both ICU and non-ICU patients, considering it is also the most accurate model in non-ICU patients.

Development and comparison of population pharmacokinetic models of vancomycin in neurosurgical patients based on two different renal function markers

WHAT IS KNOWN AND OBJECTIVES

Some previous studies have indicated that serum cystatin C (Cys C) is a better marker than serum creatinine (SCR) for assessing the glomerular filtering rate (GFR). However, in almost all population pharmacokinetic models of vancomycin, the GFR is usually estimated from SCR. Therefore, the aim of this study was to compare the GFR estimated from SCR (sGFR) with the GFR estimated from Cys C (cGFR) and investigate which one can describe the characteristics of vancomycin population pharmacokinetics better in Chinese neurosurgical adult patients.

METHODS

Patients from the Neurosurgery Department aged ≥18 years were enrolled retrospectively. Among these patients, the data from 222 patients were used to establish two population pharmacokinetic models based on sGFR and cGFR, separately. The data from another 95 patients were used for the external validation of these two models. Non-linear mixed-effect modelling (NONMEM) 7.4.3 was used for the population pharmacokinetic analysis.

RESULTS

We developed two one-compartment models with first-order absorption based on Cys C and SCR, separately. In the Cys C model, age, body weight and cGFR were significant covariates on the clearance rate (CL) of vancomycin (typical value, 6.4 L/hour). In the SCR model, age and sGFR were significant covariates on the CL (typical value, 6.46 L/hour). The external validation results showed that the predictive performance of the two models was similar.

WHAT IS NEW AND CONCLUSION

In this study, the predictive performance of two models was similar in neurosurgical patients. We did not find a significant improvement in the predictive performance of the model when GFR was estimated from Cys C.

Towards precision dosing of vancomycin: a systematic evaluation of pharmacometric models for Bayesian forecasting

OBJECTIVES

Vancomycin is a vital treatment option for patients suffering from critical infections, and therapeutic drug monitoring is recommended. Bayesian forecasting is reported to improve trough concentration monitoring for dose adjustment. However, the predictive performance of pharmacokinetic models that are utilized for Bayesian forecasting has not been systematically evaluated.

METHOD

Thirty-one published population pharmacokinetic models for vancomycin were encoded in NONMEM®7.4. Data from 292 hospitalized patients were used to evaluate the predictive performance (forecasting bias and precision, visual predictive checks) of the models to forecast vancomycin concentrations and area under the curve (AUC) by (a) a priori prediction, i.e., solely by patient characteristics, and (b) also including measured vancomycin concentrations from previous dosing occasions using Bayesian forecasting.

RESULTS

A priori prediction varied substantially-relative bias (rBias): -122.7-67.96%, relative root mean squared error (rRMSE) 44.3-136.8%, respectively-and was best for models which included body weight and creatinine clearance as covariates. The model by Goti et al. displayed the best predictive performance with an rBias of -4.41% and an rRMSE of 44.3%, as well as the most accurate visual predictive checks and AUC predictions. Models with less accurate predictive performance provided distorted AUC predictions which may lead to inappropriate dosing decisions.

CONCLUSION

There is a diverse landscape of population pharmacokinetic models for vancomycin with varied predictive performance in Bayesian forecasting. Our study revealed the Goti model as suitable for improving precision dosing in hospitalized patients. Therefore, it should be used to drive vancomycin dosing decisions, and studies to link this finding to clinical outcomes are warranted.