Performance characteristics of TCI devices implementing the Marsh model

R. Adapa, A. Bhatia, A.R. Absalom

University Division of Anaesthesia, Addenbrooke's Hospital, Cambridge

 Introduction

There are several commercially available propofol target-controlled infusion (TCI) devices and pharmacokinetic simulation software packages. We compared the output of different simulation programs with each other, and with the measured output of TCI devices.

Methods
Simulations

We used Stanpump, TIVATrainer and Rugloop to calculate the volumes required by the Marsh model to achieve target blood propofol concentrations of 0.5 and 6.0 mg/ml for 120 minutes.

Infusion device performance

The infusion devices were used to administer 1% propofol by TCI (at target blood propofol concentrations of 0.5 and 6.0 mg/ml for 120 minutes). Devices used were: (1) Alaris Asena PK pump; (2) Graseby 3500 pump; (3) a Fresenius Orchestra Base Primea TCI infusion system; (4) a PC running Stanpump software controlling a Graseby 3500 pump.

For each device, propofol syringe was connected to a 200cm long intravenous infusion administration set, which was in turn connected to a 22G intravenous cannula. The cannula was attached to and its tip directed into, a measuring flask placed on a high precision scale. At 10 min intervals, for each system, the cumulative mass of propofol delivered was recorded and used to calculate the administered volume for each 10 min epoch (Vol_meas). We also recorded from the user interface the cumulative volume that the system estimated it had delivered (Vol_est). For each epoch, j, 10 minute error was defined as: (Vol_{meas, j} - Vol_est,j)/ Vol_est,j x 100. Experiments were conducted under identical conditions; the flask was sealed to avoid evaporation during the experiment.

Results
Simulation programs:
The propofol volumes calculated by the different programs were almost identical.
Infusion devices:
The infusion devices implemented infusion regimens very similar to those calculated by simulation programs. All pumps performed well at the high target concentration. Accuracy at low concentrations is shown in Figure 1.

Conclusion:
All pumps performed sub-optimally at low infusion rates. At low target concentrations, the Fresenius and Graseby pumps significantly overestimated the volume of propofol administered during the early phases of the infusion.

Acknowledgements: Fresenius Kabi kindly loaned the authors a Base Primea system