Monia Guidi (1, 2), Alena Simalatsar (1, 3), Aziz Chaouch (1, 4), Sandro Carrara (5) and Thierry Buclin (1)
(1) Division of Clinical Pharmacology, Lausanne University Hospital (CHUV), Lausanne, Switzerland; (2) School of Pharmaceutical Sciences, University of Geneva, University of Lausanne, Geneva, Switzerland; (3) Institute of Systems Engineering, University of Applied Sciences and Arts - Western Switzerland, Sion, Switzerland; (4) Division of Biostatistics, Institute of Social and Preventive Medicine, Lausanne University Hospital (CHUV), Lausanne, Switzerland; (5) Laboratory of Integrated Systems, Ecole Polytechnique Federale de Lausanne (EPFL), Lausanne, Switzerland
Objectives: Propofol administration through open-loop target controlled infusion (TCI) devices is largely used in clinical practice to induce and maintain both sedation and general anesthesia. The classic TCI algorithm adjusts drug infusion rates to rapidly achieve and preserve stable brain concentrations according to model-predicted plasma and brain levels [1]. The pharmacokinetic (PK) model currently in use in most hospitals was developed on a small set of volunteers [2]. However, an important between-subject variability (BSV) characterizes propofol PK in clinical conditions [3]. This implies that actual propofol levels differ from predicted ones in a significant fraction of patients, so that current TCI pumps deliver inadequate drug dosages with possible important consequences for the patients [4]. Closed-loop automated anesthesia delivery systems might clearly improve the safety and the quality of anesthesia care while giving the anesthesiologists the opportunity to focus on the higher-level clinical tasks required during the surgery [5]. We participate in a project aiming at developing a miniaturized sensor able to monitor circulating propofol during surgery [6]. The aim of the present work was to develop a closed-loop algorithm for propofol administration driven by real-time plasma measurements and to evaluate its performances through a simulation study.
Methods: The suggested algorithm couples the Bayesian minimization approach with the classic open TCI algorithm of Shafer et al. Study population consisted of 1000 simulated female subjects (70 kg, 170 cm, 36 y) with real PK parameters obtained by the comprehensive model of Eleveld et al with BSV [3]. Bayesian minimization of the Eleveld et al model based on all the real-time plasma measurements up to the last available one allowed personalizing the population microconstants of the classic TCI algorithm, so to predict each individual brain concentration-time profile. Plasma measurements were generated using the real individual PK parameters, the model intraindividual variability, and the computed infusion rates. A constant brain target concentration of 5 mg/L (accepted surfaces: target +/-30% (large criterion, LC), 20% (average, AC) or 10% (strict, SC) of target) over a 60 min surgery was chosen. The number of subjects achieving the target surface was calculated. In addition, we defined as problematic patients those reaching the target surface more than 20 min after operation started, spent less than 20 min in the target surface or had average concentration outside the target surface longer than 5 min. Percentages of problematic individuals receiving propofol at infusion rates computed by classic TCI and our closed-loop algorithm were compared. The proposed approach was evaluated against measurement period (step=15 and 60 sec) and delay between real time blood sampling and arrival time for data processing (35 sec) according to the characteristics of the existing sensor prototype for continuous propofol quantification [6]. For the highest step, the role of the proportional component of the intraindividual variability was explored (model: 47%, tested: 24%).
Results: Target was achieved in 99% (LC), 95% (AC) and 86% (SC) of the subjects with 25%, 47% and 78% identified as problematic for classic TCI administration. Conversely, all the patients reached the target when computing infusion rates with the proposed approach independently of the chosen period and delay. Moreover, only problematic according to LC and AC for step=60 sec including or not the delay. The percentage of SC problematic patients increases from 27% to 31% and from 73% to 76% for step=15 and 60 sec, respectively, upon integration of delay in the closed-loop algorithm. As expected, lowering the proportional component of the intraindividual variability decreased the number of problematic patients for step=60 sec to <2% for LC and AC and <44% for SC independently of the delay.
Conclusions: The suggested closed-loop algorithm based on real-time plasma measurement markedly reduces the number of patients significantly under- or over-exposed during anesthesia relatively to the classic TCI devices. Measurement delay minimally affects the performances of the approach, which not surprisingly depends on measurement period and noise.
References:
[1] Shafer, S.L. and K.M. Gregg, Algorithms to rapidly achieve and maintain stable drug concentrations at the site of drug effect with a computer-controlled infusion pump. J Pharmacokinet Biopharm, 1992. 20(2): p. 147-69.
[2] Schnider, T.W., et al., The influence of method of administration and covariates on the pharmacokinetics of propofol in adult volunteers. Anesthesiology, 1998. 88(5): p. 1170-82.
[3] Eleveld, D.J., et al., A general purpose pharmacokinetic model for propofol. Anesth Analg, 2014. 118(6): p. 1221-37.
[4] Guidi, M., et al., Poster: Adequacy of open-loop Target Controlled Infusion devices: is there room for a closed-loop control to imporve automated propofol delivery during anesthesia?: PAGE meeting 2016.
[5] Dumont, G.A. and J.M. Ansermino, Closed-loop control of anesthesia: a primer for anesthesiologists. Anesth Analg, 2013. 117(5): p. 1130-8.
[6] Stradolini, F., et al., Raspberry Pi Based System for Portable and Simultaneous Monitoring of Anesthetics and Therapeutic Compounds. 2017 New Generation in CAS (NGCAS), 2017: p. 101-104.
Reference: PAGE 27 (2018) Abstr 8662 [www.page-meeting.org/?abstract=8662]
Poster: Drug/Disease Modelling - CNS