Bayesian model-based versus Reinforcement learning closed-loop control of BIS guided propofol administration during intra-operative anesthesia.

Propofol is the most frequently applied hypnotic-anesthetic iv drug during
anesthesia. In *open-loop* controlled propofol administration during
anesthesia, initial dosing guidelines are based on the typical subject, without
taking into account the large inter individual variability. To manage this
variability, most clinicians will start by giving a standard dose, observes the
therapeutic effect and will adapt the dose regimen. Although, *open-loop* drug
administration is clinically *standard-of-care*, the efficiency of this
decision process highly depends on the expertise of the clinician, is very time
consuming and might result in a suboptimal therapy. This process of dose
titration might be optimized by applying closed-loop drug administration
techniques. Propofol closed-loop controllers are computer programs designed to
maintain a targeted effect as defined by BIS by adapting the administered
amount of drugs. In closed-loop control, the anesthesiologist only enters the
desired variable to be maintained (*BIS target*). Previously, both groups at
the University Medical Center Groningen (in collaboration with the Ghent
University, Gent, Belgium, Demed Engineering, Temse, Belgium and Aspect
Medical, Norwood, MA, USA) and Stanford University (in collaboration with Mr.
Brett Moore, Department of Computer Science, Texas Tech University, Lubbock,
TX, USA) have developed technologies for computer controlled drug
administration being Bayesian model-based drug administration and Reinforcement
learning (RL) controlled drug administration. For the Bayesian control, various
publication in simulations and clinical practice has shown that safety and
accuracy can be garantueed so far, however, more in depth study is required.
For, the RL, simulation and preliminary volunteer research has been done.

to compare the robustness of two existing closed-loop control technologies for
BIS-guided propofol administration in combination with open-loop
effect-compartment controlled remifentanil infusion. Robustness of the
controllers will be evaluated and compared according to widely accepted
laboratory performance criteria and peri-operative clinical outcome.

Prospective, double-blinded, randomized comparison study

The induction of anesthesia (until start of surgery) will be extended with 5 to
10 minutes for obtaining specific performance endpoints. We do not expect
additional risk for the patient