Bayesian optimized Propofol Target-Controlled Infusion.

Hooren, H. van (2014) Bayesian optimized Propofol Target-Controlled Infusion. thesis, Medicine.

Text HoorenvanH.pdf Restricted to Registered users only Download (387kB)

Abstract

Introduction: An aim in providing anesthesia is to obtain and maintain the drug effect fast and accurately. Using a TCI (Target-Controlled Infusion) system, the desired plasma concentration can be set and changed. TCI systems calculate the infusion rates required to achieve and maintain the target concentration. The available TCI systems are population based models, meaning that all equilibration parameters are based upon a large group of volunteers and patients. This results in a wide variety of blood concentrations and there still is a typical error in the measurements of the individual patient of 20% between measured and predicted plasma concentration of propofol. This so called “prediction error” is clinically accepted Objective: In this study, the population-based TCI will be compared with an individualized TCI based upon the Bayesian approach. This study investigates whether online adaptation of the population based TCI model by returning measured propofol plasma concentrations to the model leads to a smaller prediction error between measured and predicted plasma concentrations. The purpose of this thesis is to perform on interim analyses in order to evaluate the prospectively calculated sample size and to evaluate possibility to reach significance between groups at the finalization of the study. Materials and methods: Patients scheduled for CABG (Coronary Artery Bypass Graft), meeting the inclusion criteria, were invited to join the study. During this study, all propofol PK (pharmacokinetic) parameters were collected every second. Propofol blood measurements were performed at standardized moments, i.e.: 10, 35, 50, 65 and 75 minutes after start of propofol infusion. After 60 minutes of infusion, in the intervention group, an individualization of the predicted volumes of distribution and clearances in the PKPD-model was implemented by the computer to minimize any predicted error. Results: The difference of prediction error in the post adaptation phase between intervention and control group was made, by using the Wilcoxon rank sum test. A significance of 0,547 indicates that there is no difference between intervention- and control-group. With the data used, the number of patients required to reach statistical significance after adaptation was calculated. Conclusion: The Bayesian optimization of the TCI model was investigated in patients who underwent a CABG. As the study is still including patients, it is too early to make a final conclusion. However, this interim analysis proved that the prospectively calculated power analysis estimated a correct sample size probably leading to a significance between groups favoring TCI with Bayesian optimization. the interim analysis shows that this study might leads to a significant improvement when using Bayesian optimization.

Item Type:	Thesis (Thesis)
Supervisor name:	Struys, Prof. MMRF and Berg, JP van den
Faculty:	Medical Sciences
Date Deposited:	25 Jun 2020 11:06
Last Modified:	25 Jun 2020 11:06
URI:	https://umcg.studenttheses.ub.rug.nl/id/eprint/2565

Actions (login required)

View Item