Ferdinand A. Weinelt (1,2), Miriam Stegemann (3,4), Anja Theloe (4,5), Frieder Pfäfflin (3,4), Stephan Achterberg (3,4), Lisa Ehmann (1,2), Wilhelm Huisinga (6), Robin Michelet (1), Stefanie Hennig* (1,7,8), Charlotte Kloft* (1)
(1) Dept. of Clinical Pharmacy and Biochemistry, Institute of Pharmacy, Freie Universitaet Berlin, Germany, (2) Graduate Research Training program PharMetrX, Germany, (3) Department of Infectious Diseases and Pulmonary Medicine, Charité-Universitätsmedizin Berlin, Corporate Member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health, Berlin, Germany, (4) Antibiotic Stewardship, Charité-Universitätsmedizin Berlin, Corporate Member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health, Berlin, Germany, (5) Pharmacy Department, Charité-Universitätsmedizin Berlin, Corporate Member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health, Berlin, Germany, (6) Institute of Mathematics, Universität Potsdam, Germany, (7) Certara, Inc., Princeton, New Jersey, USA, (8) School of Clinical Sciences, Faculty of Health, Queensland University of Technology, Brisbane, QLD 4000, Australia *shared senior authorship
Objectives: For efficacious meropenem treatment of infections in intensive care unit (ICU) patients maintaining minimum concentrations (Cmin) above the minimal inhibitory concentration (MIC) of the bacterium is recommended [1]. High pharmacokinetic (PK) variability observed in ICU patients increases the risk of suboptimal drug exposures [2]. Recent observations from Charité hospital in Berlin revealed >70% of the patient’s observed Cmin values being outside the target range at Charité (Cmin > 1x – 5x MIC), which urgently requires improvement of existing dosing practices. Model-based dose adjustment tables could be utilised to optimise antibiotic treatment. To produce a relevant and robust model-based dosing tool, the predictability of the underlying PK model in the target population requires evaluation. Data arising from routine therapeutic drug monitoring (TDM) in the applicable population and hospital could be considered for external model evaluation. Unfortunately, those TDM datasets commonly only contain information of the immediate dosing event and the associated TDM observation event. The aim of the present research was to assess whether a routine TDM dataset is sufficient for meropenem model evaluation in ICU patients.
Methods: Routine TDM data of 66 meropenem plasma concentrations from 34 ICU patients were available from Charitè Berlin. Initially only this “routine TDM dataset” was available, containing information on therapy start and dosing and associated sampling events, necessitating the imputation of the prior dosing history. For the imputation, the time between treatment start and the reported dosing event was divided by the number of planned dosing events and an average dosing interval was calculated. Infusion durations and doses were assumed to be consistent for the whole treatment period. For comparison, a “complete TDM dataset” was collected, which comprised the full dosing history of each patient for each sample of the routine TDM dataset. Stochastic simulations using a previously developed meropenem PK model in ICU patients [3] were performed with both data sets using NONMEM 7.4.3. Normalised prediction distribution errors (NPDEs) were analysed for both datasets using the npde package (v. 2.0) in R/Rstudio (v. 3.5.0/v. 1.1.447). The Wilcoxon signed-rank test, Fisher ratio test and Shapiro-Wilks test were used to assess possible deviations (mean≠0, variance≠1 and the normality assumption, respectively) from the standard normal distribution. Additionally, median absolute simulation errors (MASE) and median simulation errors (MSEs) were calculated to assess precision and bias, respectively. Model evaluation was considered successful, if the NPDE distribution did not significantly differ from standard normal distribution, MSEs confirmed low bias and MASEs acceptable precision for ICU patients.
Results: The NPDE distribution based on the stochastically simulated TDM concentrations and the”routine TDM dataset” with imputed prior dosing history differed significantly from the standard normal distribution (adjusted p-value: 0). MAEs of 31% confirmed bias and under prediction of observed concentrations. MASEs of 57% revealed acceptable precision. Here the hypothesis that the model described the newly observed data well was rejected. For the complete TDM dataset, the NPDE distribution did not differ significantly from the standard normal distribution (adjusted p-value: 0.319) and the hypothesis that the model describes the newly observed data well could thus not be rejected. MAEs of -14% confirmed low bias and MASEs of 51% revealed acceptable precision for ICU patients. Comparison between the “routine TDM dataset” and the “complete TDM dataset” revealed frequent changes in doses, dosing times, infusion durations and intervals due to medical examinations and interventions in the ICU patients.
Conclusions: Using the routine TDM dataset including imputed dosing information was not sufficient to perform acceptable external PK model evaluation. For successful evaluation of a PK model clinical data from a setting with high probability of irregular dosing histories requires the availability of the full dosing history. Using such a dataset acceptable predictive performance for the selected PK model was demonstrated and the development of model-based dosing tools for intensive care patients at Charité Berlin and their prospective evaluation in a clinical setting is underway.
References:
[1] R. Guilhaumou et. al. Optimization of the treatment with beta-lactam antibiotics in critically ill patients—guidelines from the French Society of Pharmacology and Therapeutics (Société Française de Pharmacologie et Thérapeutique—SFPT) and the French Society of Anaesthesia and Intensive Care Medicine (Société Française d’Anesthésie et Réanimation—SFAR). Crit. Care 23: 104 (2019).
[2] J. Gonçalves-Pereira, P. Póvoa. Antibiotics in critically ill patients: A systematic review of the pharmacokinetics of β-lactams. Crit. Care 15: R206 (2011).
[3] L. Ehmann, M. Zoller, I. Minichmayr, C. Scharf, W. Huisinga, J. Zander, C. Kloft. Development of a dosing algorithm for meropenem in critically ill patients based on a population pharmacokinetic/pharmacoynamic analysis. Int. J. Antimicrob. Agents 54: 309-317 (2019)
Reference: PAGE () Abstr 9469 [www.page-meeting.org/?abstract=9469]
Poster: Clinical Applications