Tobramycin dose individualization using the MonolixSuite
Geraldine Ayral (1), Kaelig Chatel (1), Marc Lavielle (2)(3), Jonathan Chauvin (1)
(1) Lixoft, Antony, France, (2) Center of Applied Mathematics, Ecole Polytechnique, Palaiseau, France, (3) Inria Saclay – Ile-de-France, team Xpop, Palaiseau, France
Objectives: The concept of personalized medicine has long been promoted to bring more benefits to patients, in particular by adapting the administrated dose to the patient’s characteristics. Methodologically and practically, dose individualization is a challenging task. We show how this can be done in an efficient way using the MonolixSuite.
Methods: For the case study, we use the antimicrobial agent Tobramycin, which has a narrow therapeutic index. For efficacy, a sufficiently high serum concentration must be achieved. On the other hand, an excess exposure over a long time period bears the risk of nephrotoxicity and ototoxicity. Using the non-linear mixed-effect model and data published in [1], we performed simulations to determine (i) if the typical treatment is safe and efficient, and (ii) the dose that would be most likely to be safe and efficient.
Results: We first show that the default dosing regimen of 1mg/kg every 8 hours is safe in only 75% of the simulated healthy individuals. This calls for an individualization of the dosing regimen. We thus used the individual covariates to predict via simulations the concentration interval for one specific individual taking into account the residual random effects that remains after consideration for a specific covariate value. We implemented a simple optimization algorithm to determine a priori the dose that has the highest chance of being safe and efficient. For an individual weighting 78kg and with an impaired renal function (creatinine clearance 30mL/min), we propose to adapt the dosing regimen to 1.2mg/kg every 14 hours. If in addition, early drug monitoring permits to measure the drug’s concentration at a few time points after the initial dose, these data can be used to obtain the distribution of the individual’s parameters (given the covariates and the population parameters), using the Markov Chain Monte-Carlo procedure implemented in the Monolix software. We show that 4 measurements permits to reduce the concentration prediction interval width 5-fold for this individual.
Conclusions: This example shows how a personalized treatment can be realized using population PK modeling and simulation, and how the MonolixSuite software (Monolix for parameter estimation and Simulx for simulations) components facilitate an efficient implementation. The modeling/simulation approach permits to precisely assess the trade-off between the prediction precision and the costs of information acquisition.
References:
[1] Aarons, L., Vozeh, S., Wenk, M., Weiss, P., & Follath, F. (1989). Population pharmacokinetics of tobramycin. British Journal of Clinical Pharmacology, 28(3), 305–14.
