Bayesian optimized Propofol Target-Controlled Infusion

Population based pharmacokinetic-dynamic (PKPD) models of propofol are used in
daily practice to titrate propofol towards a predicted plasma- and/or
effect-site concentration. 1-4 It has been accepted that the population based
prediction of the propofol plasma concentration may have an error of about 20%
compared to the measured propofol concentrations in the individual patient. 5
This error is considered acceptable in the clinically applied new generation
target controlled infusion (TCI) systems for propofol administration 6-8.

Recently, new technology has been developed to measure propofol concentrations
in plasma with minimal delay of about 10 minutes after sampling. This
technology opens opportunities to decrease the residual error between predicted
(population) and measured (individual) propofol plasma concentrations during
maintenance of anaesthesia. A decrease of the prediction error has several
potential advantages such as less accumulation of drug, faster recovery from
anaesthesia, less overshoot in propofol effect when adjusting the dose etc*

Our study tests whether online adaptation of the population PKPD model (being
used to calculate the infusion rates during maintenance of anaesthesia), based
on differences measured and predicted concentrations, will decrease the
residual errors between subsequent measurements and predicted concentrations.
Such an individualization of the population PKPD should be done in a Bayesian
approach as it has been shown to be a good method of updating pharmacokinetic
models during infusion, when intermittent drug concentration measurements are
performed 9. This method adapts the starting (population) pharmacokinetic
model, on the basis of the measured blood samples to generate a
patient-individualized model.

We hypothesize that online adaptation or individualization of the population
PKPD model of propofol will decrease the residual error between predicted and
measured plasma-concentrations during maintenance of anaesthesia, when the
adapted PKPD models are used to calculate propofol infusion rates required for
the plasma concentrations set by the responsible anaesthetist.

blinded prospective, randomized controlled trial

During the case, the patients will be monitored using standard monitoring, such
as ECG, non-invasive blood pressure (NIBP), pulse-oximetry (SpO2) and
bispectral index (BIS), as per routine clinical practice. Venous access will be
obtained on non-dominant arm or hand for drug and fluid infusion. As per
clinical practice, all patients will receive an arterial line in the radial
artery of the same arm, after local lidocaine injection. This will be used for
blood sampling required for clinical practice and propofol measures, and for
invasive arterial blood pressure (ABP) monitoring.

A maximum amount of 10 additional arterial blood samples (per patients) for
propofol measurement will be taken. The total amount of blood required for the
propofol samples will not exceed 100 ml.

The administration of propofol in the *Bayesian adapted* TCI period is
hypothesized to be optimized towards the individual patient. As such, a benefit
might be expected compared to standard clinical practice (less accumulation of
drug, faster recovery from anaesthesia, less overshoot in propofol effect when
dose adjustments are performed etc)

Safety of the slightly altered propofol titration in the Bayesian adjusted TCI
group is guaranteed by the fact that the output of the pump (expressed in
mg/kg/h) and the resulting clinical effect remains under the control of the
anaesthetist at all times during this study. As such the propofol titration can
be adjusted freely according to the discretion of the anaesthetist.

The risk of participating in this study is identical to the risk of the normal
clinical procedure.