Evaluation of Tobramycin Exposure Predictions in Three Bayesian Forecasting Programmes Compared with Current Clinical Practice in Children and Adults with Cystic Fibrosis

The Significance of Bayesian Pharmacokinetics in Dosing for Critically Ill Patients: A Primer for Clinicians Using Vancomycin as an Example

Antibiotic use is becoming increasingly challenging with the emergence of multidrug-resistant organisms. Pharmacokinetic (PK) alterations result from complex pathophysiologic changes in some patient populations, particularly those with critical illness. Therefore, antibiotic dose individualization in such populations is warranted. Recently, there have been advances in dose optimization strategies to improve the utilization of existing antibiotics. Bayesian-based dosing is one of the novel approaches that could help clinicians achieve target concentrations in a greater percentage of their patients earlier during therapy. This review summarizes the advantages and disadvantages of current approaches to antibiotic dosing, with a focus on critically ill patients, and discusses the use of Bayesian methods to optimize vancomycin dosing. The Bayesian method of antibiotic dosing was developed to provide more precise predictions of drug concentrations and target achievement early in therapy. It has benefits such as the incorporation of personalized PK/PD parameters, improved predictive abilities, and improved patient outcomes. Recent vancomycin dosing guidelines emphasize the importance of using the Bayesian method. The Bayesian method is able to achieve appropriate antibiotic dosing prior to the patient reaching the steady state, allowing the patient to receive the right drug at the right dose earlier in therapy.

Therapeutic drug monitoring-guided dosing for pediatric cystic fibrosis patients: recent advances and future outlooks

INTRODUCTION

Medicine use in children with cystic fibrosis (CF) is complicated by inconsistent pharmacokinetics at variance with the general population, a lack of research into this and its effects on clinical outcomes. In the absence of established dose regimens, therapeutic drug monitoring (TDM) is a clinically relevant tool to optimize drug exposure and maximize therapeutic effect by the bedside. In clinical practice though, use of this is variable and limited by a lack of expert recommendations.

AREAS COVERED

We aimed to review the use of TDM in children with CF to summarize recent developments, current recommendations, and opportunities for future directions. We searched PubMed for relevant publications using the broad search terms "cystic fibrosis" in combination with the specific terms "therapeutic drug monitoring (TDM)" and "children." Further searches were undertaken using the name of identified drugs combined with the term "TDM."

EXPERT OPINION

Further research into the use of Bayesian forecasting and the relationship between exposure and response is required to personalize dosing, with the opportunity for the development of expert recommendations in children with CF. Use of noninvasive methods of TDM has the potential to improve accessibility to TDM in this cohort.

Implementation and Cross-Validation of a Pharmacokinetic Model for Precision Dosing of Busulfan in Hematopoietic Stem Cell Transplanted Children

Busulfan, a drug used in conditioning prior to hematopoietic stem cell transplantation (HSCT) in children, has a narrow therapeutic margin. The model-informed precision dosing (MIPD) of busulfan is desirable, but there is a lack of validated tools. The objective of this study was to implement and cross-validate a population pharmacokinetic (PK) model in the Tucuxi software for busulfan MIPD in HSCT children. A search of the literature was performed to identify candidate population PK models. The goodness of fit of three selected models was assessed in a dataset of 178 children by computing the mean error (ME) and root-mean-squared error of prediction (RMSE). The best model was implemented in Tucuxi. The individual predicted concentrations, the area under the concentration-time curve (AUC), and dosage requirements were compared between the Tucuxi model and a reference model available in the BestDose software in a subset of 61 children. The model from Paci et al. best fitted the data in the full dataset. In a subset of 61 patients, the predictive performance of Tucuxi and BestDose models was comparable with ME values of 6.4% and -2.5% and RMSE values of 11.4% and 13.6%, respectively. The agreement between the estimated AUC and the predicted dose was good, with 6.6% and 4.9% of the values being out of the 95% limits of agreement, respectively. To conclude, a PK model for busulfan MIPD was cross-validated and is now available in the Tucuxi software.

Pharmacokinetic modelling to predict risk of ototoxicity with intravenous tobramycin treatment in cystic fibrosis

INTRODUCTION

Further optimization of therapeutic drug monitoring (TDM) for aminoglycosides (AGs) is urgently needed, especially in special populations such as those with cystic fibrosis (CF), >50% of whom develop ototoxicity if treated with multiple courses of IV AGs. This study aimed to empirically test a pharmacokinetic (PK) model using Bayesian estimation of drug exposure in the deeper body tissues to determine feasibility for prediction of ototoxicity.

MATERIALS AND METHODS

IV doses (n = 3645) of tobramycin and vancomycin were documented with precise timing from 38 patients with CF (aged 8-21 years), including total doses given and total exposure (cumulative AUC). Concentration results were obtained at 3 and 10 h for the central (C1) compartment. These variables were used in Bayesian estimation to predict trough levels in the secondary tissue compartments (C2 trough) and maximum concentrations (C2max). The C1 and C2 measures were then correlated with hearing levels in the extended high-frequency range.

RESULTS

Patients with more severe hearing loss were older and had a higher number of tobramycin C2max concentrations >2 mg/L than patients with normal or lesser degrees of hearing loss. These two factors together significantly predicted average high-frequency hearing level (r = 0.618, P < 0.001). Traditional metrics such as C1 trough concentrations were not predictive. The relative risk for hearing loss was 5.8 times greater with six or more tobramycin courses that exceeded C2max concentrations of 3 mg/L or higher, with sensitivity of 83% and specificity of 86%.

CONCLUSIONS

Advanced PK model-informed analysis predicted ototoxicity risk in patients with CF treated with tobramycin and demonstrated high test prediction.

Individualized precision dosing approaches to optimize antimicrobial therapy in pediatric populations

Introduction:Severe infections continue to impose a major burden on critically ill children and mortality rates remain stagnant. Outcomes rely on accurate and timely delivery of antimicrobials achieving target concentrations in infected tissue. Yet, developmental aspects, disease-related variables, and host factors may severely alter antimicrobial pharmacokinetics in pediatrics. The emergence of antimicrobial resistance increases the need for improved treatment approaches.Areas covered:This narrative review explores why optimization of antimicrobial therapy in neonates, infants, children, and adolescents is crucial and summarizes the possible dosing approaches to achieve antimicrobial individualization. Finally, we outline a roadmap toward scientific evidence informing the development and implementation of precision antimicrobial dosing in critically ill children.The literature search was conducted on PubMed using the following keywords: neonate, infant, child, adolescent, pediatrics, antimicrobial, pharmacokinetic, pharmacodynamic target, Bayes dosing software, optimizing, individualizing, personalizing, precision dosing, drug monitoring, validation, attainment, and software implementation. Further articles were sought from the references of the above searched articles.Expert opinion:Recently, technological innovations have emerged that enabled the development of individualized antimicrobial dosing approaches in adults. More work is required in pediatrics to make individualized antimicrobial dosing approaches widely operationalized in this population.

Bayesian Forecasting for Intravenous Tobramycin Dosing in Adults With Cystic Fibrosis Using One Versus Two Serum Concentrations in a Dosing Interval

BACKGROUND

Intravenous tobramycin treatment requires therapeutic drug monitoring (TDM) to ensure safety and efficacy when used for prolonged treatment, as in infective exacerbations of cystic fibrosis. The 24-hour area under the concentration-time curve (AUC24) is widely used to guide dosing; however, there remains variability in practice around methods for its estimation. The objective of this study was to determine the potential for a sparse-sampling strategy using a single postinfusion tobramycin concentration and Bayesian forecasting to assess the AUC24 in routine practice.

METHODS

Adults with cystic fibrosis receiving once-daily tobramycin had paired concentrations measured 2 hours (c1) and 6 hours (c2) after the end of infusion as routine monitoring. AUC24 exposures were estimated using Tucuxi, a Bayesian forecasting application that incorporates a validated population pharmacokinetic model. Simulations were performed to estimate AUC24 using the full data set using c1 and c2, compared with estimates using depleted data sets (c1 or c2 only), with and without concentration data from earlier in the course. The agreement between each simulation condition and the reference was assessed graphically and numerically using the median difference (∆) AUC24 and (relative) root mean square error (rRMSE) as measures of bias and accuracy, respectively.

RESULTS

A total of 55 patients contributed 512 concentrations from 95 tobramycin courses and 256 TDM episodes. Single concentration methods performed well, with median ∆AUC24 <2 mg·h·L-1 and rRMSE of <15% for sequential c1 and c2 conditions.

CONCLUSIONS

Bayesian forecasting implemented in Tucuxi, using single postinfusion concentrations taken 2-6 hours after tobramycin administration, yield similar exposure estimates to more intensive (two-sample) methods and are suitable for routine TDM practice.

Evaluation of two software using Bayesian methods for monitoring exposure and dosing once-daily intravenous busulfan in paediatric patients receiving haematopoietic stem cell transplantation

AIM

To assess the ability of model-based personalised dosing tools to estimate busulfan exposure (i) in comparison to clinically used intensive sampling exposure estimation procedure, (ii) using limited sampling strategies and (iii) to predict changes in busulfan clearance during busulfan treatment.

METHODS

Data on intravenous busulfan dosing for patients with 4 consecutive days were entered into Bayesian forecasting software, InsightRX and NextDose. Prediction of busulfan cumulative exposure was compared to current clinical practice estimation, aiming for pre-defined individualised target of cumulative exposure. Estimation performance was tested given several limited sampling strategies.

RESULTS

Thirty-two paediatric patients (0.2-16.5 years) provided a total of 103 daily exposure measurements estimated using 7 samples taken per day (full sampling), with 19 patients having sampling following all doses administered. Both software tools utilising Bayesian methods provided acceptable relative bias and precision of cumulative exposure estimations under the tested sampling scenarios. Relative bias ranged from median RE of 0.1-14.6% using InsightRX and from 3.4-7.8% using NextDose. Precision ranged from median RMSE of 0.19-0.32 mg·h·L^-1 for InsightRX and 0.08-0.1 mg·h·L^-1 for NextDose. A median reduction in busulfan clearance from day 1 to day 4 was observed in the clinical data (-10.9%), when using InsightRX (-18.6%) and with NextDose (-14.7%).

CONCLUSION

Bayesian methods were shown to have relatively low bias and precisely estimate busulfan exposure using intensive sampling and several limited sampling strategies, which provides evidence for prospective studies to evaluate these tools in clinical practice. A trend to overestimation of exposure using Bayesian methods was observed compared to clinical practice. Reduction of busulfan clearance from day 1 to 4 of once daily dosing was confirmed and should be considered when adjusting doses.

Key Challenges in Providing Effective Antibiotic Therapy for Critically Ill Patients with Bacterial Sepsis and Septic Shock

Early initiation of effective antibiotic therapy is vitally important for saving the lives of critically ill patients with sepsis or septic shock. The susceptibility of the infecting pathogen and the ability of the selected dosage regimen to safely achieve the required antibiotic exposure need to be carefully considered to achieve a high probability of a successful outcome. Critically ill patients commonly experience substantial pathophysiological changes that impact the functions of various organs, including the kidneys. Many antibiotics are predominantly renally eliminated and thus renal function is a major determinant of the regimen needed to achieve the required antibiotic exposure. However, currently, there is a paucity of guidelines to inform antibiotic dosing in critically ill patients, including those with sepsis or septic shock. This paper briefly reviews methods that are commonly used in critically ill patients to provide a measure of renal function, and approaches that describe the relationship between the exposure to an antibiotic and its antibacterial effects. Two common conditions that very substantially complicate the use of antibiotics in critically ill patients with sepsis, unstable renal function, and augmented renal clearance, are considered in detail and their potential therapeutic implications are explored. Suggestions are provided on how treatment of bacterial infections in critically ill patients with sepsis might be improved. Of high potential are model-informed approaches that aim to individualize initial treatment regimens based on patient and bacterial characteristics, with refinement of regimens during treatment in response to monitoring antibiotic concentrations, responsive measures of renal function, and other important clinical data.

A Critique of Pharmacokinetic Calculators for Drug Dosing Individualization

The 'one-dose-fits-all' approach where drug dosing regimen is prescribed according to recommendations from a summary of product characteristics is not appropriate for many patients whose clinical characteristics significantly differ from the most frequent ones in a population, as it cannot guarantee optimal exposure of target tissues to the drug. Our aim here is to provide a concise review of pharmacokinetic calculators currently available for clinical use and, at the same time, to suggest the minimum standards that they should satisfy to be routinely used in clinical practice. A systematic search of Medline, Ebsco, Scopus, Scindeks, Cochrane Library and Google Scholar was performed to find publications about available pharmacokinetic calculators for drug dose individualization. Theoretically well-founded and mathematically correct calculators for many drugs are available, but only a few calculators for specific drugs have been validated in clinical practice or through clinical trials, and the results published in peer-reviewed journals. The majority of available pharmacokinetic calculators for drug dosing individualization remain unvalidated, i.e., there is no evidence of their efficacy and safety in real-life clinical settings. Pharmacokinetic calculators for drug dose individualization are irreplaceable tools for achieving precision medicine, where dosing regimens are tailored to the needs and personal characteristics of each patient, maximizing efficacy and minimizing toxicity.

Can Population Pharmacokinetics of Antibiotics be Extrapolated? Implications of External Evaluations

BACKGROUND AND OBJECTIVE

External evaluation is an important issue in the population pharmacokinetic analysis of antibiotics. The purpose of this review was to summarize the current approaches and status of external evaluations and discuss the implications of external evaluation results for the future individualization of dosing regimens.

METHODS

We systematically searched the PubMed and EMBASE databases for external evaluation studies of population analysis and extracted the relevant information from these articles. A total of 32 studies were included in this review.

RESULTS

Vancomycin was investigated in 17 (53.1%) articles and was the most studied drug. Other studied drugs included gentamicin, tobramycin, amikacin, amoxicillin, ceftaroline, meropenem, fluconazole, voriconazole, and rifampicin. Nine (28.1%) studies were prospective, and the sample size varied widely between studies. Thirteen (40.6%) studies evaluated the population pharmacokinetic models by systematically searching for previous studies. Seven (21.9%) studies were multicenter studies, and 27 (84.4%) adopted the sparse sampling strategy. Almost all external evaluation studies of antibiotics (93.8%) used metrics for prediction-based diagnostics, while relatively fewer studies were based on simulations (46.9%) and Bayesian forecasting (25.0%).

CONCLUSION

The results of external evaluations in previous studies revealed the poor extrapolation performance of existing models of prediction- and simulation-based diagnostics, whereas the posterior Bayesian method could improve predictive performance. There is an urgent need for the development of standards and guidelines for external evaluation studies.

Reappraisal of Contemporary Pharmacokinetic and Pharmacodynamic Principles for Informing Aminoglycoside Dosing

Therapeutic drug management is regularly performed for aminoglycosides in an effort to maximize their effectiveness and safety. The ratio of maximum plasma drug concentration to minimum inhibitory concentration (Cmax/MIC) has long been regarded as the primary pharmacokinetic/pharmacodynamic (PK/PD) index of clinical efficacy for aminoglycosides due to their concentration-dependent killing. In this review, however, we discuss why the area under the plasma concentration-time curve (AUC)/MIC ratio may be a more reliable indicator of bacterial killing and clinical efficacy for these agents. The definitive AUC/MIC efficacy targets for aminoglycosides are less clear, unlike those that exist for fluoroquinolones. Evaluation of available literature suggests that an AUC/MIC ratio of 30-50 for aminoglycoside therapy may provide optimal outcomes when targeting non-critically ill immunocompetent patients with low-bacterial burden gram-negative infections such as urinary tract infections or in patients receiving additional gram-negative therapy with good source control. However, an AUC/MIC target of 80-100 may be more prudent when treating patients with aminoglycoside monotherapy or in critically ill patients with high-bacterial burden infections, such as nosocomial pneumonia. Reappraisal of current antimicrobial susceptibility breakpoints for aminoglycosides against gram-negative bacteria may also be necessary to achieve these AUC/MIC targets and ensure that current empiric doses are not grossly suboptimal in critically ill patients. Although it has been historically difficult to calculate AUCs in clinical practice, equation-based and Bayesian approaches now can be used to estimate the AUC in clinical practice, with limited PK sampling. Additional research is needed to better define optimal AUC/MIC targets for efficacy, especially when drugs are used in combination, as well as PK/PD targets associated with suppression of resistance. It is also important to determine if AUC can predict nephrotoxicity of these agents or whether trough concentrations should be used instead.

Optimising antimicrobial therapy through the use of Bayesian dosing programs

The optimisation of antibiotic dosing therapy with therapeutic drug monitoring is widely recommended. The aim of therapeutic drug monitoring is to help the clinician to achieve target pharmacokinetic/pharmacodynamic parameters, maximising efficacy and minimising toxicity. Computerised programs, utilising the Bayesian estimation procedures, are able to achieve target concentrations in a greater percentage of patients earlier in the course of therapy compared to linear regression analysis and population methods. This article summarises various methods for dose optimisation of antibiotics with a focus on Bayesian programs.

An evaluation of the user-friendliness of Bayesian forecasting programs in a clinical setting

AIMS

To evaluate 3 Bayesian forecasting (BF) programs-TDMx, InsightRx and DoseMe-on their user-friendliness and common liked and disliked features through a survey of hospital pharmacists.

METHODS

Clinical pharmacists across 3 Australian hospitals that did not use a BF program were invited to a BF workshop and complete a survey on programs they trialled. Participants were given 4 case scenarios to work through and asked to complete a 5-point Likert scale survey evaluating the program's user-friendliness. Liked and disliked features of each program were ascertained through written responses to open-ended questions. Survey results were compared using a χ² test of equal or given proportions to identify significant differences in response.

RESULTS

Twenty-seven pharmacists, from hospitals, participated. BF programs were rated overall as user-friendly with 70%, 41% and 37% (P = .02) of participants recording a Likert score of 4 or 5 for DoseMe, TDMx and InsightRx, respectively. Participants found it easy to access all required information to use the programs, understood dosing recommendations and visualisations given by each program, and thought programs supported decision-making with >50% of participants scoring a 4 or 5 across the programs in these categories. Common liked features across all programs were the graphical displays and ease of data entry, while common disliked features were related to the units, layout and information display.

CONCLUSION

Although differences exist between programs, all 3 programs were most commonly rated as user-friendly across all themes evaluated, which provides useful information for healthcare facilities wanting to implement a BF program.

Handling interoccasion variability in model-based dose individualization using therapeutic drug monitoring data

AIMS

This study aims to assess approaches to handle interoccasion variability (IOV) in a model-based therapeutic drug monitoring (TDM) context, using a population pharmacokinetic model of coagulation factor VIII as example.

METHODS

We assessed 5 model-based TDM approaches: empirical Bayes estimates (EBEs) from a model including IOV, with individualized doses calculated based on individual parameters either (i) including or (ii) excluding variability related to IOV; and EBEs from a model excluding IOV by (iii) setting IOV to zero, (iv) summing variances of interindividual variability (IIV) and IOV into a single IIV term, or (v) re-estimating the model without IOV. The impact of varying IOV magnitudes (0-50%) and number of occasions/observations was explored. The approaches were compared with conventional weight-based dosing. Predictive performance was assessed with the prediction error percentiles.

RESULTS

When IOV was lower than IIV, the accuracy was good for all approaches (50^th percentile of the prediction error [P50] <7.4%), but the precision varied substantially between IOV magnitudes (P97.5 61-528%). Approach (ii) was the most precise forecasting method across a wide range of scenarios, particularly in case of sparse sampling or high magnitudes of IOV. Weight-based dosing led to less precise predictions than the model-based TDM approaches in most scenarios.

CONCLUSIONS

Based on the studied scenarios and theoretical expectations, the best approach to handle IOV in model-based dose individualization is to include IOV in the generation of the EBEs but exclude the portion of unexplained variability related to IOV in the individual parameters used to calculate the future dose.

A clinically relevant decrease in abiraterone exposure associated with carbamazepine use in a patient with castration-resistant metastatic prostate cancer

ADVERSE EVENT

Decreased abiraterone exposure after introducing carbamazepine.

DRUGS IMPLICATED

Abiraterone acetate and carbamazepine.

THE PATIENT

A 65-year-old man with metastatic castration resistant prostate cancer, was treated with abiraterone acetate and prednisolone, and received concomitant carbamazepine for treatment of facial neuropathy.

EVIDENCE THAT LINKS THE DRUG TO THE EVENT

The interaction was confirmed by a decrease in abiraterone exposure >2-fold (area-under-the-curve and trough levels). After discontinuation of carbamazepine therapy, the abiraterone exposure normalized. No alternative causes were found that explain the decrease in abiraterone exposure.

MECHANISM

Induction of CYP3A and potentially phase I metabolism (SULT2A1) by carbamazepine.

IMPLICATIONS FOR THERAPY

Clinicians and pharmacists should be aware of this clinically relevant interaction. The national drug-drug interaction checker does not warn for this interaction, whereas both the Lexicomp® and Micromedex® advice to avoid if possible or to increase the abiraterone dose frequency to twice daily. Carbamazepine should not be combined with abiraterone to avoid underexposure and suboptimal therapy. Therapeutic drug monitoring of abiraterone is useful to guide therapy when drug-drug interactions cannot be avoided.