POSTER ABSTRACT / DETAILS:
Drug administration is a key element in medicine which is traditionally based on average population pharmacokinetic and/or pharmacodynamic profiles. This consideration is likely to exhibit adverse effects due to violation of toxicity constraints or may fail to retain therapeutic levels. Due to the lack of feedback, there is also the assumption of zero disturbances, which contributes even more to the probability of adverse effects.
Control theory can be used in the field of drug administration in order to face these challenges and design drug dosing strategies, which are individualized for each particular patient. In this work, the Model Predictive Control (MPC) methodology is used for optimal individualized continuous-time drug administration, by exploiting the advantages of this popular control technique: MPC leads to a stable and robust closed-loop system, and it takes into account the state and the actuator constraints, which in some cases are patient-specific. These constraints result from safety considerations against adverse effects and are qualified by means of tissue-specific minimum toxic concentration (MTC) values. Bounds are also imposed on the influx rate when the administration is intravenous.
Physiologically-Based Pharmacokinetic (PBPK) models are used for predicting the dynamic behavior of drug concentrations in plasma, tissue and tumors following drug delivery, using knowledge of the physiology and anatomy of the individual patient. PBPK models rely on fundamental principles such as mass balance equations and reactions kinetics and are mathematically quantified as systems of ordinary differential equations (ODEs). Coupled with a state observer, the overall system can control drug concentration at any organ using only blood measurements.
The overall setting allows the treating physician to modify the desired concentration at each target organ at any time during the therapeutical treatment. Furthermore, there is ability of achieving the desired concentration as well as monitoring in real time the overall distribution of the drug concentration in patient's organism.
A web application is currently under development for implementing the proposed methodology and make it available to physicians and other possible users. The user may use a pre-existing PBPK structure or create his own to describe as accurate as possible a patient and save it for future use. He can also define toxicological bounds or other state and input constraints, having the choice to use default values. Next, he sets the MPC parameters, like a target concentration to a specific organ and the control horizon. The result of the application is the optimal drug dose for the individual patient, as well as the dynamic simulated prediction drug concentration in the organs of interest.
The proposed methodology was tested on a hypothetical PBPK model with 7 compartments (and overall 14 sub-compartments). All constraints are satisfied so the adverse effects are minimized, since the drug doses are kept within the recommended bounds. We demonstrated also that the use of MPC performs well in the presence of modelling errors and measurement noise, which makes it suitable for medical applications.
This work was funded by project 11ΣYN_10_1152, which was co-financed by the European Union and Greece, Operational Program “Competitiveness & Entrepreneurship”, NSFR 2007-2013 in the context of GSRT- National action “Cooperation”.