Prediction of Propofol Effect-Site Concentration Associated With Deep Anesthesia

Propofol is the most widely used anesthetic to induce general anesthesia (GA). However, as opposed to volatile anesthesia, propofol concentration monitoring is not directly available in humans. Thus, the use of a pharmacokinetic/pharmacodynamic (PK/PD) model in target-controlled infusion (TCI) is recommended. During infusion, the concentration at effect-site (Ce) is assumed to correlate with the level of hypnosis. However, There is a large variability in propofol requirements in common practice. The variability in propofol requirements is attributed to demographic factors, genetic polymorphism, procedure-related changes, and individual sensitivity. Previous studies have shown that EEG characteristics change with age and cognitive status.

Hypothesis : There is a large interindividual variability in patients' sensitivity to propofol which can be precisely modeled using clinical, demographic, and electroencephalographic (EEG) features available prior to induction of general anesthesia. The investigators also hypothesise that this sensitivity to propofol may identify vulnerable brain phenotype related to poor cognitive performance.

Specific objectives: Primary: to investigate the variability of Ce propofol at which deep anesthesia (DA) occurs during induction of general anesthesia (CeDA). Secondary: to explore the relationship between demographics, cognitive performance, and EEG variables with the independent CeDA variable. Tertiary: to develop and validate a machine learning algorithm to predict CeDA based on clinical, demographic and EEG features obtained prior the induction of general anesthesia.

Methods: This prospective monocentric observational study will include 110 participants of 18 years of age or older scheduled for surgery under general anesthesia. Baseline cognitive performance will be assessed using the Montreal Cognitive Assessment. Induction of GA will be performed using 300ml.h-1 of 1% propofol until DA (Defined as a Patient State Index (PSI) \< 30) is observed. The primary endpoint will be the Ce propofol at which deep anesthesia (CeDA) occurs as calculated by the Eleveld PK/PD model for propofol. Preoperative raw EEG waveforms from the SedLine monitor will be used to extract statistical, entropic, and spectral features. High-density EEG (128 channels) will also be recorded in a subsample of 40 patients to extract brain functional connectivity features. These features will be entered in a generalized additive model to predict CeDA. We will also develop and validate a machine learning algorithm to predict CeDA based on these features.

Significance/Importance: The results of this study could provide a better understanding of the determinants of the inter-individual variability observed in the pharmacodynamic effect of anesthetic agents. Moreover, the prediction of CeDA may help clinicians in setting the right and safe target in target-controlled infusion of propofol during induction of general anesthesia and further limit the occurrence of deep anesthesia during surgery. Finally, the knowledge of the link between EEG characteristics, sensitivity to anesthetics and cognitive performance may lead to more personalized anesthesia delivery.

• Coefficient of determination between statistical, spectral, entropic features extracted from the electroencephalogram and the Concentration at the effect-site propofol associated with deep anesthesia [Time frame: during surgery]
• Coefficient of determination between multivariable model and the Concentration at the effect-site propofol associated with deep anesthesia [Time frame: during surgery]
• Mean absolute error between predicted Concentration at the effect-site propofol associated with deep anesthesia by the machine-learning model and observed Concentration at the effect-site propofol associated with deep anesthesia. [Time frame: during surgery]