Antonin PRAET 1–3 (PhD applicant), Florence VETELE 4, Valentine BREANT 4, Oana DUMITRESCU 2,3,5, Anne DOLEANS-JORDHEIM 2,3,6, Philippe REIX 1,7, Laurent BOURGUIGNON 1,2,8, Sylvain GOUTELLE 1,2,8.
(1) LBBE – Laboratoire de Biologie et Biométrie Evolutive, CNRS UMR 5558, Université Claude Bernard Lyon 1, Université Lyon, Villeurbanne, France. (2) ISPB – Faculté de Pharmacie de Lyon, Université Claude Bernard Lyon 1, Université Lyon, France. (3) IAI – Institut des Agents Infectieux, Groupement Hospitalier Nord, Hospices Civils de Lyon, Lyon, France. (4) Service de Pharmacie, Groupement Hospitalier Est, Hospices Civils de Lyon, Bron, France. (5) CIRI – Centre International de Recherche en Infectiologie, CNRS UMR 5308, INSERM U1111, Université Claude Bernard Lyon 1, Université Lyon, Lyon, France. (6) LEM – Laboratoire d’Ecologie Microbienne, CNRS UMR 5557, INRA UMR 1418, Université Claude Bernard Lyon 1, Université Lyon, Villeurbanne, France. (7) CRCM pédiatrique Lyon, Hôpital Femme Mère Enfant, Hospices Civils de Lyon, Bron, France. (8) Service de Pharmacie, Groupement Hospitalier Nord, Hospices Civils de Lyon, Lyon, France.
Objectives:
Children with cystic fibrosis (CF) may benefit from tobramycin therapeutic drug monitoring and pharmacokinetic-based dosage adjustment, which is routine clinical practice in our center. In order to implement a new version of our dosing software BestDoseTM, the development of a new model was necessary. Two pharmacokinetic/pharmacodynamic (PK/PD) indices have been associated with response to aminoglycoside therapy: the ratio of the 24h-area under the concentration-time curve to the bacterial minimal inhibitory concentration – MIC – (AUC24/MIC) and the ratio of maximum concentration (Cmax) to the MIC (Cmax/MIC). The aims of this study were to develop and validate a nonparametric population pharmacokinetic model of tobramycin based on our pediatric CF cohort and to analyze the probability of target attainment (PTA) for those two PK/PD indices.
Methods:
This was a retrospective, single-center study of patients who underwent tobramycin TDM and PK-based dose adjustment from 2010 to 2018. Population PK analysis was conducted with the Non-parametric Adaptive Grid (NPAG) algorithm, implemented in the R package Pmetrics [1]. The influence of six covariates was tested on clearance (CL) and central volume of distribution (V1) of tobramycin: age, total body weight (WT), ideal body weight (IBW, Peck equation [2]), body surface area (BSA, Gehan et George equation [3]), creatinine clearance (CLcr Schwartz equation [4]), and height. In the learning dataset, classical criteria were used to assess model goodness-of-fit (objective-function, Akaike criterion, plots) and predictive performance (bias, imprecision). External validation was performed in a subset of data and predictive performance was computed as well. PK/PD simulations with the final model were performed by computing steady-state concentrations of tobramycin in 1000 virtual patients, for various tobramycin dosage regimens and covariates values. The target values of PK/PD indices were AUC24/MIC ≥ 100 and Cmax/MIC ≥ 10. Maximal MIC was set at 2mg/L, the EUCAST epidemiological cut-off of Pseudomonas aeruginosa. PTA ≥ 90 % was considered as acceptable. Four tobramycin dosing regimens were tested: 5 mg/kg/day, 10 mg/kg/day, 15 mg/kg/day and 20 mg/kg/day.
Results:
We analyzed data from 195 patients who received a total of 754 courses of tobramycin, from 2010 to 2018. A total of 2533 measured tobramycin concentrations were available. The mean ± SD of age, WT and CLcr, were of 11.4 ± 5 years, 35.3 ± 15 kg and 82.6 ± 30 mL/min, respectively. Random selection was used to create a learning set of 146 patients and a validation set of 49 patients. The final model was a two-compartment model. The model included a linear influence of total body weight on V1 and a linear influence of CLcr and BSA on CL. Surprisingly, allometric scaling did not improve the clearance model compared with a linear model. In the learning dataset, predictions correlated well with observations, with low values of bias and imprecision for both populations: R² = 0.829, bias = – 0.40 mg/L, imprecision = 22.71 mg²/L² ; and for individual predictions R² = 0.867, bias = – 0.81 mg/L, imprecision = 17.79 mg²/L². The good predictive performance was confirmed in the validation dataset with R² = 0.824, bias = – 1.52 mg/L, imprecision = 25.7 mg²/L² for individual predictions. Mean tobramycin volume of distribution per unit of total body weight and renal clearance were of 0.23 ± 0.09 L/kg and 2.00 ± 1.84 L/h respectively. There were discrepancies between results of PTA according to which PK/PD target was chosen. Considering a critical scenario for instance (MIC = 2 mg/L), only the 20mg/kg/day dosage was associated with acceptable PTA for the target Cmax/MIC ≥ 10 in patients with low renal function (60 mL/min). None of the tested regimens could achieve acceptable PTA for the target AUC24/MIC ≥ 100, regardless of the renal function tested (60 mL/min, 90 mL/min, 120 mL/min and 150 mL/min).
Conclusions:
The performance of the nonparametric population pharmacokinetic model developed is very satisfactory and this model can be considered for clinical implementation for Bayesian TDM of tobramycin in children with CF. Our results question the dosing and PK/PD targets for tobramycin therapy in CF children, as well as MIC breakpoints.
References:
[1] Neely MN, van Guilder MG, Yamada WM, Schumitzky A, Jelliffe RW. Accurate detection of outliers and subpopulations with Pmetrics, a nonparametric and parametric pharmacometric modeling and simulation package for R. Ther Drug Monit. août 2012;34(4):467-76.
[2] Peck CC, Murphy MG, Conner DP. Bedside estimation of ideal body weight, In?: C. C. Peck and M. G. Murphy (ed.), Bedside clinical pharmacokinetics: simple techniques for individualizing drug therapy. Appl Ther Inc Vanc Wash. 1989;80.
[3] Gehan EA, George SL. Estimation of human body surface area from height and weight. Cancer Chemother Rep. 1970;54(4):225-35.
[4] Schwartz GJ, Muñoz A, Schneider MF, Mak RH, Kaskel F, Warady BA, et al. New equations to estimate GFR in children with CKD. J Am Soc Nephrol JASN. mars 2009;20(3):629-37.
Reference: PAGE () Abstr 9580 [www.page-meeting.org/?abstract=9580]
Poster: Oral: Drug/Disease Modelling