Model Predictive Controller For Closed-Loop Administration Of Propofol Using Bispectral Index
Yoshihito Sawaguchi M.E.1, Eiko Furutani Ph.D.1, Gotaro Shirakami M.D.2,
Mituhiko Araki Ph.D.1 and Kazuhiko Fukuda M.D.2
1 Department of Electrical Engineering, Kyoto University, Kyoto, Japan,
2 Department of Anesthesia, Kyoto University Hospital, Kyoto, Japan.
Introduction:
For feedback control systems, dead-times included in a controlled object may seriously affect stability and performance of the closed-loop system1. However, in most of previous works on closed-loop administration of propofol using the Bispectral Index (BIS) 2, 3, a dead-time of the BIS response to propofol infusion was not considered explicitly in controller design. Recently, we have developed a closed-loop control system using a model predictive controller1, which can appropriately take the dead-time into account. The system has an identification function of the individual response and a risk control function for preventing drug over-infusion and intra-operative arousal. After obtaining the approval from the Ethics Committee of Kyoto University Hospital, 80 clinical trials were made from July 2002. In this study, we evaluated performance of the system with manually controlled infusion.
Methods:
One hundred seventy adult patients (ASA PS 1-2) scheduled for various kinds of ambulatory surgery were randomly divided into two groups. Anesthesia was induced with propofol IV bolus (2 mg/kg) and subsequent continuous (10 mg/kg/h for 3 min) infusions. In Automatic group (n=80) the controller adjusted propofol infusion rate automatically to keep the BIS value to 50 (40-60) using individual pharmacodynamic parameters calculated during induction. In Manual group (n=90) a clinician adjusted propofol infusion rate manually to keep the BIS value to 50(40-60) according to the ”10, 8, 6” (mg/kg/h) scheme. Supplemental IV fentanyl and vecuronium, local infiltration anesthetic and rectal diclofenac were administered as needed during anesthesia.
Results:
There were no significant differences between the two groups in the backgrounds of patients. In Automatic group, the mean BIS value was nearer to 50 (48 ± 3 and 46 ± 6), the mean infusion rate was lower (8.3 ± 2.0 and 8.9 ± 1.5 mg/kg/h), the settling time (time needed until settling of BIS value in 40-60 for 5 min from start of the infusion) was shorter (26 ± 33 and 44 ± 54 min), and BIS error (BIS<40 or >60) ratio (22 ± 15 and 39 ± 26 %), median performance error2 (-4 ± 6 and -11 ± 14 %) and median absolute performance error2 (11 ± 5 and 17 ± 7 %) were smaller than Manual group (p<0.05).
Conclusion:
The model predictive anesthesia control system can decrease the drug consumption and maintain the BIS value more accurately around the target set point in comparison with the manual control.
References:
1. Chu et al. Annual Reviews in Control 1998; 22: 59-72. 2) Struys et al. Anesthesiology 2001; 95: 6-17. 3) Absalom et al. Anesthesiology 2002; 96: 67-73.