| JOURNAL HOME | CME HOME | THIS MONTH | PAST ISSUES | ETOC | COLLECTIONS |
| AUTHORS | REVIEWERS | EDITORIAL BOARD | FEEDBACK | RSS | HELP |
 
|
|
|
||
|
|
Anesth Analg 2004;99:1159-1163 © 2004 International Anesthesia Research Society doi: 10.1213/01.ANE.0000131507.65132.7F
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionReferences |
|---|
We have shown that a multicompartment model accurately predicts end-tidal (ET) sevoflurane (sevo) and isoflurane concentrations. The model has been adapted to use real-time fresh gas flow and vaporizer settings to display a 10-min prediction of ET sevo concentrations. In this study, we evaluated the effect of the predictive display on the speed and accuracy of changes in ET sevo by the anesthesiologist. Fifteen patients were studied in whom sevo-based anesthesia was expected to last more than 2 h. Four step changes of target ET concentration (+0.5, +1.0, –1.0, and –0.5 vol%) were made either unaided or with the prediction display. Fresh gas flow was 1 L/min. Response time, maximum overshoot, and stability in the 5 min after the target was achieved were compared by using two-tailed paired Student’s t-tests. Changes were made on average 1.5–2.3 times faster with the predictive display than without it. These differences were statistically significant (P < 0.05) for the +0.5, +1.0, and –0.5 vol% step changes but not for the –1.0 vol% change. There were no differences in the degree of overshoot or stability. These differences are comparable to those seen with an automatic feedback control system. This system may simplify the administration of volatile anesthesia and the use of low-flow anesthesia.
IMPLICATIONS: A model-based system that uses anesthesia machine settings to predict and display future end-tidal (ET) sevoflurane concentrations was tested. Anesthesiologists made step changes in ET sevoflurane 1.5–2.3 times faster when using the system. This display may simplify the administration of volatile anesthesia, especially at low gas flows.
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Pharmacokinetic models are available for a wide range of IV drugs used during anesthesia. These models form the basis for methods of assisting the anesthesiologist in the delivery of these drugs and include computer-based systems such as "StanPump," a manual slide rule for propofol target-controlled infusion (TCI) (1), and a comprehensive display of the previous and projected blood and effect-site levels of all IV drugs (2), but none of these systems is in routine use. Commercial systems for TCI of propofol, such as the Diprifusor®, are available in many countries.
Similar pharmacokinetic models have been described for volatile anesthetics. Although systems for closed-loop control of volatile anesthesia have been described (3–5), these are not in common clinical use. There may be several reasons for this difference, including regulatory hurdles (also affecting TCI of drugs other than propofol), the need for suitable actuators and delivery systems, and perceived reluctance of the marketplace.
We have developed a system to guide the administration of volatile anesthetics. This system displays past values and future predictions for end-tidal (ET) concentrations of volatile anesthetics on the basis of past and current vaporizer and fresh gas flow (FGF) settings. The aim of this article is to describe the system and evaluate its utility in guiding changes in ET sevoflurane (sevo) concentrations.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
We have described the validation of a nine-compartment model of anesthetic uptake and distribution (6) that predicts ET isoflurane and sevo at least as well as various models predict blood propofol concentrations. The model is an adaptation of that described by Heffernan et al. (7), originally developed as a teaching tool. This was based on those described by Mapleson (8,9) and uses nine compartments: circuit, lung/blood, heart, brain, kidney, liver, muscle, fat, and poorly perfused tissues. The model does not include compartments to mimic blood transit time (8,10). Variables of the model, such as the volume of tissue compartment and relative blood flows, are those used by Heffernan et al. (7), and the various partition coefficients are those we have used previously (6,11,12). A detailed description of the model is included in the report of the validation study (6).
In the version of the system used in this study, the anesthesiologist enters the anesthetic in use and the weight of the patient. This information is used to initialize the system by setting compartment sizes, cardiac output, and ventilation based on patient weight, as well as partition coefficients for the chosen volatile anesthetic. The model then receives FGF and vaporizer settings every 10 s from the anesthesia machine (Datex ADU) and uses this information to model the course of the anesthetic.
Every 10 s, after iteration of the main model, the current state of the model is copied to a second instance of the model, which then "looks ahead" 10 min by using the current FGF and vaporizer settings to produce predictions for ET values. These predictions are scaled to bring the value for the current predicted ET in line with the actual ET value measured on the anesthesia monitor by multiplying all predictions by the ratio (current measured ET)/(predicted measured ET).
The system display is updated every 10 s and shows past measured values along with the predictions (Fig. 1). The unscaled "error" in the prediction (current prediction – current measured) is also shown to provide the user with a continuing indication of the performance of the model. For clinical use, the system is run on a Macintosh LCIII computer, and the output is displayed on a screen that incorporates a trend graph of most monitored variables. The system was placed in two operating rooms for 6 mo before this study. A description of the system was distributed to all anesthesiologists in our department (40 staff specialists and 16 residents), and a copy was available with each system.
|
The Canterbury Ethics Committee approved the study. Fifteen ASA physical status I–III patients scheduled to undergo elective surgery expected to last at least 2 h with sevo-based anesthesia agreed to participate in the study and gave written consent. Before starting, use and interpretation of the system display were reviewed with the anesthesiologist in charge of the case. The anesthesiologist was asked to make eight step changes in measured ET sevo values. Each change was to be made as rapidly as possible, and the new level was to be maintained for a further 5 min with minimal variation. The study was designed with two arms. In each arm, 4 changes (an increase of 0.5 and 1 vol% and a decrease of 0.5 and 1 vol%) were made. In one arm, the prediction system display was visible, and in the other arm, the same four changes were made without the display. Random number determined the order of the study arms, and all changes in one arm were completed before proceeding to the other arm. Within each arm, the anesthesiologist chose the time at which each of the four changes in that arm were made and the order in which they occurred and announced each change to the investigator. Each change was preceded by a period of at least 5 min with a constant ET sevo concentration. Total FGF was maintained at 1 L/min throughout the study period.
Data from the anesthetic machine (Datex ADU) and the anesthesia monitor (Datex AS/3) and the output of the model were recorded every 10 s on a Macintosh iBook, which also provided the display used in the study. Data for each subject were transferred to a spreadsheet to allow extraction of the predefined end-points. The primary end-point of the study was the 10%–90% increase time for each change. Secondary end-points were the maximum overshoot and the proportion of time within 0.15 vol% of the target over the 5 min after the target was achieved.
Data are presented as mean ± SD. Results for each change type with and without the prediction system were compared by using a paired two-tailed Student’s t-test; P < 0.05 was taken as significant.
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Data were collected from 15 anesthetics given by 13 anesthesiologists. Ten of the anesthesiologists, including the two who where involved in the study twice, were staff specialists, and the remaining three were residents. All had prior experience with the predictive system. The mean age of the patients was 54 yr (range, 33–72 yr), and mean weight was 83 kg (range, 54–122 kg). Data were incomplete in three cases because not all changes could be completed before the end of surgery; these data were included in the analysis. One set of data was unusable because of equipment problems.
The results are summarized in Table 1. The mean time for a change was between 1.5 and 2.3 times faster in the predictive display group when compared with the control group. These differences were statistically significant for the 1.0 vol% increase (P = 0.02), the 0.5 vol% increase (P = 0.04), and the 0.5 vol% decrease (P = 0.03), but not the 1.0 vol% decrease (P = 0.26). There were no significant differences in the degree of overshoot (P values were all 
0.2).
|
Table 1 also shows the proportion of time spent within 0.15 vol% of the target over the 5 min after the target was first achieved for each type of change. This was taken as an indication of the stability of control of ET sevoflurane levels after the change. There were no significant differences among groups in this measure of stability.
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
This article describes a decision support tool for the administration of volatile anesthesia. Use of this system allowed anesthesiologists to make a number of step changes in ET sevo approximately twice as fast as when the system was not available. There was no difference in the amount of overshoot or variability in the 5 minutes after the change.
Systems that provide closed-loop control of ET volatile concentrations have been described, but with the exception of the Drager Physioflex, which is sold in Europe, these systems have not been available commercially. Anesthesia-delivery systems for total feedback control are under development.
Sieber et al. (3) used a similar study design to assess an automatic feedback control system. In their study, step changes of ±0.3 and ±0.6 vol% isoflurane were made either by anesthesiologists or by the control system. They found that the automatic system was faster than manual control for the ±0.6 vol% changes, but not the ±0.3 vol% step. This is similar to the pattern of results in the present study. In contrast, Sieber et al. (3) found that their automatic system produced less overshoot and more stability than manual control.
In contrast to automated control systems, our predictive display allows the anesthesiologist to administer a manually-driven target-controlled volatile anesthetic by displaying a prediction of ET values over the succeeding 10 minutes on the basis of the current FGF and vaporizer dial settings. Because the predictions are updated every 10 seconds, the user can rapidly adjust the "inputs" to produce a desired effect on ET values. We have shown that this model has minimal bias, with a median performance error (a measure of bias in the individual) of 3.4% and a median absolute performance error (a measure of the inaccuracy of the model) of 10.9% (6). More importantly, the amount of bias stays relatively constant in an individual, as estimated by "divergence" (time-weighted deviation) and "wobble" (variability of the error in an individual). These variables were defined by Varvel et al. (13) and have become the standard method for analyzing the performance of IV infusion systems. Values found with the model used for this study are well within values suggested as acceptable limits for a TCI system: median performance error <10%–20% and median absolute performance error 20%–40% (14).
Varvel et al. (13), in describing their method for assessing the performance of models, suggest that low values of the time-dependent measures of bias are more important than absolute measures of bias. If these factors are low, the user can rely on a system to maintain a given blood or ET level. In the model used in this system, wobble is 3.8%, and divergence is 0.85%/h. Assuming a typical ET target for sevo of 2 vol%, this translates to a wobble in ET of <0.1 vol% and a divergence of <0.02 vol%/h.
The results of this study suggest that the system facilitates more rapid changes in ET values than unaided manual control. This allows the user to determine the current state of the patient on the basis of a large number of factors and then to use the system to help maintain the current state or to help move to another state (more or less volatile) in a controlled manner, as demonstrated by the results of this study. However, these results suggest that the predictive system does not affect the quality of ET control, as estimated by the degree of overshoot and time spent within 0.15 vol% of the target.
We did not see a significant difference in the 1.0 vol% decrease group, although the mean increase time for the change was reduced by 35%. A post hoc power analysis suggests that 30 subjects would be needed to show a significant difference in this group. A statistically significant difference was seen in the other three groups, suggesting that mechanisms other than sample size may have been involved. We suspect that in this study, a given decrease in the ET value was often produced by turning the vaporizer off. This means that, because the FGF was fixed at 1 L/min, the rate of decline was determined by the physical characteristics of the anesthetic. A flow rate of 1 L/min is often used during maintenance in our hospital (15). We have observed that, in routine clinical practice, users of the system manipulate FGF as a secondary means of adjusting future ET values so that FGF may be increased during a change, and once the ET value approaches the target, the FGF is reduced again. Because the system takes the guesswork out of low-flow anesthesia, we often see FGF reduced to quite low levels when the predictive system is used in routine practice. The effect of different FGF rates was outside the scope of this study. However, one would expect a system such as that described in this article to become more useful as flow rates decrease.
There are a number of potential problems with the design of this study. We did not formally assess the learning effect during the study. However, superficial analysis does not suggest any differences in results over the duration of the study. The study design also did not differentiate between the effect of the predictive display and the historical trend display. The standard configuration for anesthesia patient monitors in our hospitals includes the display of a "slow agent" wave form. This gives a display of the recent (five- to seven-minute) history of inspired and expired volatile concentrations and may be part of the reason why there was little difference in variability among groups once the change was completed. In contrast, Sieber et al. (3) found more overshoot and variability in their manual control group.
We chose to study sevo rather than isoflurane because the lower solubility of sevo means that inspired and expired concentrations are more tightly coupled. This should make the step changes required in this study easier to generate unaided with sevo than with isoflurane, and thus a significant result in a study with sevo should carry more weight.
There are a number of advantages to an advisory system such as this. With suitable transducers, it can be adapted to any anesthesia-delivery platform. It overcomes perceived safety issues of closed-loop computer control (16). Although users in this study were asked to make changes as rapidly as possible, the system allows the user to control the rate of change; many control systems are set to achieve the desired level as soon as possible, but this system allows the user to choose the rate of change. The system also illustrates the type of changes a control system might make and may act as a useful introduction to help anesthesiologists become more comfortable with a closed-loop control system. The system also acts as a safety device: if the vaporizer is left on a very high or low setting, the predictive display clearly demonstrates the future effects on ET values and may alert the anesthesiologist before any risk or harm to the patient occurs.
In addition, this system proved to be a valuable teaching tool for illustrating aspects of volatile anesthetic uptake and distribution in a real-time clinical setting. Although only the effect on ET values is displayed, the effects of various changes in both inputs and patient variables are readily illustrated and provide a basis for discussion of the underlying processes. Many useful systems to demonstrate these principles have been described (7,17,18). The model used in our system is, in fact, the descendant of one of these (7). These tools are useful classroom tools. They are often used by those in the early stages of their training but are often difficult to relate to clinical practice.
In a simulation study using a similar type of predictive display of the effect-site concentrations of propofol and remifentanil, Syroid et al. (2) found less variation in effect-site concentrations. Their study participants believed that they performed better and with reduced stress when using the predictive display. These results support the use and value of a predictive display for administering anesthetics and echo comments made informally by users of our system.
The advantages of low-flow and closed-circuit anesthesia are well recognized. Several closed-loop control systems for volatile anesthesia are based on closed-circuit administration (19,20), as are a number of models that could be adapted as controllers (21). The system described in this article facilitates low-flow anesthesia by graphically demonstrating the effects of various flow rates. The user can rapidly determine the minimum FGF that will maintain a given ET value. It also allows users to reduce flows at a rate they are comfortable with. This approach is also useful because it facilitates low-flow techniques rather than being proscriptive.
A system for manual control of target-controlled volatile anesthesia that predicts ET concentrations 10 minutes into the future is described. Use of this system allows faster changes in ET sevo that are comparable to those achieved with an automated system. The system has a number of potential advantages over both current practice and fully automated systems. Development and evaluation of the system are continuing.
|
Acknowledgments |
|---|
Supported by a grant from the Canterbury Medical Research Foundation.
|
Footnotes |
|---|
Presented in part at the Euroanaesthesia Meeting, Glasgow, Scotland, May 2003.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
- Bruhn J, Bouillon TW, Ropcke H, Hoeft A. A manual slide rule for target-controlled infusion of propofol: development and evaluation. Anesth Analg 2003; 96: 142–7.
[Abstract/Free Full Text] - Syroid ND, Agutter J, Drews FA, et al. Development and evaluation of a graphical anesthesia drug display. Anesthesiology 2002; 96: 565–75.[Web of Science][Medline]
- Sieber TJ, Frei CW, Derighetti M, et al. Model-based automatic feedback control versus human control of end-tidal isoflurane concentration using low-flow anaesthesia. Br J Anaesth 2000; 85: 818–25.
[Abstract/Free Full Text] - Ross JA, Wloch RT, White DC, Hawes DW. Servo-controlled closed-circuit anaesthesia: a method for the automatic control of anaesthesia produced by a volatile agent in oxygen. Br J Anaesth 1983; 55: 1053–60.
[Abstract/Free Full Text] - Westenskow DR, Zbinden AM, Thomson DA, Kohler B. Control of end-tidal halothane concentration. A. Anaesthesia breathing system and feedback control of gas delivery. Br J Anaesth 1986; 58: 555–62.
[Abstract/Free Full Text] - Kennedy RR, French RA, Spencer C. Predictive accuracy of a model of volatile anesthetic uptake. Anesth Analg 2002; 95: 1616–21.
[Abstract/Free Full Text] - Heffernan PB, Gibbs JM, McKinnon AE. Teaching halothane uptake and distribution: a computer simulation program. Anaesthesia 1982; 37: 9–17.[Web of Science][Medline]
- Mapleson WW. Circulation-time models of the uptake of inhaled anaesthetics and data for quantifying them. Br J Anaesth 1973; 45: 319–33.
[Abstract/Free Full Text] - Mapleson WW. Mathematical aspects of the uptake, distribution and elimination of inhaled gases and vapours. Br J Anaesth 1963; 36: 129–39.
- Lerou JGC, Dirksen R, Beneken Kolmer HH, Booij LHDJ. A system model for closed-circuit inhalational anesthesia. I. Computer study. Anesthesiology 1991; 75: 345–55.[Web of Science][Medline]
- Kennedy RR, Baker AB. Solubility characteristics of the ideal agent for measurement of cardiac output by soluble gas uptake methods. Br J Anaesth 1993; 71: 398–402.
[Abstract/Free Full Text] - Kennedy RR, Baker AB. The effect of cardiac output changes on end-expired volatile anaesthetic concentrations: a theoretical study. Anaesthesia 2001; 56: 1034–40.[Web of Science][Medline]
- Varvel J, Donoho D, Shafer S. Measuring the predictive performance of computer-controlled infusion pumps. J Pharmacokinet Biopharm 1992; 20: 63–94.[Web of Science][Medline]
- Swinhoe CF, Peacock JE, Glen JB, Reilly CS. Evaluation of the predictive performance of a ‘Diprifusor’ TCI system. Anaesthesia 1998; 53: 61–7.
- Kennedy RR, French RA. An audit of anaesthetic fresh gas flow rates and volatile anaesthetic use in a teaching hospital. N Z Med J 2003; 116: U438.[Medline]
- Egan TD, Shafer SL. Target-controlled infusions for intravenous anesthetics: surfing USA not! Anesthesiology 2003; 99: 1039–41.[Web of Science][Medline]
- Gage JS. Mathematics and judgment: a review of the computer simulation program GAS MAN. MD Comput 1992; 9: 54–6.[Medline]
- Loughlin PJ, Bowes WA, Westenskow DR. An oil-based model of inhalational anesthetic uptake and elimination. Anesthesiology 1989; 71: 278–82.[Medline]
- Ross JAS, Wloch RT, White DC, Hawes DW. Servo-controlled closed-circuit anaesthesia. Br J Anaesth 1983; 55: 1053–9.
- Devchand D, Chakrabarti MK, Lockwood GG. The uptake of enflurane during anaesthesia. Anaesthesia 1995; 50: 491–5.[Web of Science][Medline]
- Lerou JG, Booij LH. Model-based administration of inhalation anaesthesia. I. Developing a system model. Br J Anaesth 2001; 86: 12–28.
[Abstract/Free Full Text]
This article has been cited by other articles:
|
|
 |
|
  L. I. Cortinez, I. F. Troconiz, R. Fuentes, P. Gambus, Y.-W. Hsu, F. Altermatt, and H. R. Munoz The Influence of Age on the Dynamic Relationship Between End-Tidal Sevoflurane Concentrations and Bispectral Index Anesth. Analg., November 1, 2008; 107(5): 1566 - 1572. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. Fuentes, L. I. Cortinez, M. M. R. F. Struys, A. Delfino, and H. Munoz The Dynamic Relationship Between End-Tidal Sevoflurane Concentrations, Bispectral Index, and Cerebral State Index in Children Anesth. Analg., November 1, 2008; 107(5): 1573 - 1578. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. R. Kennedy and R. A. French Changing Patterns in Anesthetic Fresh Gas Flow Rates Over 5 Years in a Teaching Hospital Anesth. Analg., May 1, 2008; 106(5): 1487 - 1490. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 F. A. Drews, N. Syroid, J. Agutter, D. L. Strayer, and D. R. Westenskow Drug Delivery as Control Task: Improving Performance in a Common Anesthetic Task Human Factors: The Journal of the Human Factors and Ergonomics Society, January 1, 2006; 48(1): 85 - 94. [Abstract] [PDF]
|
|
|
|
|
|
R. R. Kennedy The Effect of Using Different Values for the Effect-Site Equilibrium Half-Time on the Prediction of Effect-Site Sevoflurane Concentration: A Simulation Study Anesth. Analg., October 1, 2005; 101(4): 1023 - 1028. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
