| JOURNAL HOME | CME HOME | THIS MONTH | PAST ISSUES | ETOC | COLLECTIONS |
| AUTHORS | REVIEWERS | EDITORIAL BOARD | FEEDBACK | RSS | HELP |
| ||||||||||||||
| ||||||||||||||
|
|
|
|
||||||||||||
|
||||||||||||||
From the *Department of Anesthesiology and Intensive Care, Hospital Clínico Universitario de Valencia, Valencia, Spain; Department of Anesthesiology and Intensive Care, Hospital Clínico Universitario, Valencia, Spain; and Klinik fur Anaesthesiologie, St. Josef-Hospital, Klinikum der Ruhr-Universitat, Bochum, Germany.
Address correspondence and reprint requests to F. Javier Belda, MD, PhD, Anesthesiology and Intensive Care Department, Hospital Clínico Universitario de Valencia, Avenida Blasco Ibáñez, 17. 46010 Valencia, Spain. Address e-mail to fjbelda{at}uv.es.
 |
Abstract |
|---|
 Top Abstract IntroductionMETHODSRESULTSDISCUSSIONAppendix 2REFERENCES |
|---|
METHODS: We studied 50 ICU patients who received sevoflurane via the AnaConDa. They were randomly divided into three groups. A 6-h infusion of liquid anesthetic was adjusted according to the infusion scheme to a target end-tidal sevoflurane concentration of 1% (Group 1%, n = 15) and 1.5% (Group 1.5%, n = 15). The initial rate was adjusted to reach the target concentration in 10 min and then the infusion was reduced to the first hour maintenance rate and readjusted once each hour afterwards. The actual concentrations were measured in the breathing circuit and compared with the target values. In the third group (n = 20) we used the model to increase and decrease the target concentration (±0.3%) for 3 h and evaluated the actual change in concentration achieved. The ability of the infusion scheme to provide the target concentration was quantified by calculating the performance error (PE). Infusion scheme performance was evaluated in terms of accuracy (median absolute PE, MDAPE) and bias (median PE, MDPE).
RESULTS: Performance parameters (mean ± sd, %) were for 1%, 1.5%, increase of concentration by 0.3% and decrease of concentration by 0.3% groups, respectively: MDAPE 5.3 ± 5.5, 2.6 ± 4.0, 5.0 ± 5.6, 5.5 ± 5.4; MDPE –5.3 ± 5.5, –2.3 ± 4.1, –0.1 ± 7.1, 0.2 ± 5.4. No significant differences were found between means of all performance parameters when the 1% and 1.5% groups were compared.
CONCLUSIONS: There is an excellent 6-h predictive performance of a simplified pharmacokinetic model for manually adjusted infusion of liquid sevoflurane when using the AnaConDa to deliver sevoflurane to ICU patients.
|
Introduction |
|---|
|
TopAbstractIntroductionMETHODSRESULTSDISCUSSIONAppendix 2REFERENCES |
|---|
Most commonly, IV drugs are used for ICU sedation.8–18 However, inhaled anesthetics may be ideal sedatives for the ICU14–20 because of their pulmonary elimination, limited amount of metabolism, bronchodilation,20 and cardioprotective effects.21–23
However, inhaled anesthetics are not widely used for sedation in the ICU, since most modern ICU ventilators do not readily accommodate an anesthetic vaporizer. The new anesthetic conserving device (ACD), AnaConDa (Sedana MedicalTM, Sweden),24–26 uses a syringe pump to deliver inhaled anesthetic in liquid form into the breathing circuit of a standard ICU ventilator (Fig. 1).27 A main advantage when compared with an ICU ventilator (open circuit) with a vaporizer attached is that an ACD has been proven to reduce anesthetic consumption24 to a level equivalent to that of a circle system using a fresh gas flow of 1.5 L/min.24,26 The manufacturer of the device recommends an infusion scheme whereby an initial rate is used to achieve the desired concentration followed by a reduced infusion rate to maintain the desired concentration. This recommended approach does not model the known characteristics of uptake and distribution for inhaled anesthetics. A precise infusion scheme for the syringe pump ensuring the desired alveolar concentration of the inhaled anesthetic has not yet been described.
|
We adapted a classical pharmacokinetic model to obtain an infusion scheme for the clinical use of the ACD with sevoflurane. Since the infusion rate of the pump was manually adjusted, a key objective was to make only one infusion rate change per hour, to facilitate its clinical use. Moreover, when a change in concentration was desired, this scheme needed to facilitate increasing or decreasing the target concentration when necessary.
We evaluated the predictive performance in patients of a simple, manually adjusted pump infusion scheme for continuous infusion of liquid sevoflurane to an ACD filter for 6 h with the goal of maintaining a constant targeted alveolar concentration. Performance was analyzed after a standard method described by Varvel et al.,28 which has been extensively used to evaluate the predictive capacity of IV target-controlled infusion systems29 and the accuracy of models of volatile anesthetic uptake.30–32
|
METHODS |
|---|
|
TopAbstractIntroductionMETHODSRESULTSDISCUSSIONAppendix 2REFERENCES |
|---|
In order to study the capacity of the model to increase and decrease (±0.3%) the target concentration, another 20 patients (Group rise/low 0.3%) with similar characteristics were studied. This group received sevoflurane for 3 h.
Patients with respiratory diseases, high ventilatory demands, hemodynamic instability or obesity (20% over their ideal weight) were excluded.
Study Protocol
Patients were sedated and mechanically ventilated with different ICU ventilators. Respiratory patterns were adjusted according to the patients’ needs, to obtain normocapnia (PAco2 ranging from 35 to 45 mm Hg). The Pao2/Fio2 ratio was over 300 mm Hg for all cases and positive end-expiratory pressure was adjusted between 5 and 10 cm H2O. Electrocardiogram leads and hemodynamic variables were monitored by means of a Tramscope monitor (MarquetteTM, Milwakee). A Bispectral Index 2000 monitor (BIS, Aspect Medical SystemsTM, MA) was used to evaluate the sedation level. At the ventilator’s expiratory port a scavenging system was adapted and connected to the central vacuum terminal (Scat, TemelTM, Spain).
At the beginning of the study, all patients were sedated with continuous IV propofol and remifentanil infusions to achieve BIS values between 55 and 65. The ACD was fitted between the Y piece of the respiratory circuit and the endotracheal tube. The sampling line of the gas monitor (Vamos, DragerTM, Lubeck, Germany) was connected to the sampling port of the ACD and the sampling flow (150 mL/min) was redirected to the breathing system on the patient side of the ACD. At that point, propofol administration was stopped. The sevoflurane infusion started when a stable BIS value equal or superior to 80 had been observed for at least 5 min (with an index of signal quality of 80%–100%), which corresponded clinically to a Ramsay score 2–3.33 For sevoflurane infusion to the ACD, a syringe infusion pump Ivac P7000 (Alaris Medical SystemsTM, UK) was used. The pump’s infusion rate was adjusted following the values obtained from the pharmacokinetic model (see model below). The initial rate was designed to reach the alveolar target concentration in 10 min. At that point, infusion rate was reduced to the first hour maintenance rate. The infusion rate was readjusted just once each hour until the study’s 6-H period elapsed.
In the 20 patients for whom the increases and decreases of concentration were performed, the initial target of end-tidal (ET) sevoflurane was 0.7%. After 75 min, following the same model, the infusion rate was adjusted to produce an increase in concentration of 0.3% within 5 min. At minute 120, following the same pharmacokinetic scheme, the infusion rate was stopped for a specified time to decrease the ET concentration by 0.3% and was then restarted to maintain this concentration for another 60 min. Hemodynamic data and ET sevoflurane concentration values were recorded at minute 10 and every 5 min afterwards in all patients.
Model for Sevoflurane Administration
To produce a guide for liquid anesthetic infusion with the ACD, we designed a pharmacokinetic model that included two components: the patient’s anesthetic uptake and the loss of anesthetic through the ACD. Anesthetic uptake and distribution was modeled with a nine-compartment model based on that described by Lowe34 and subsequently modified by Heffernan et al.35,36 and Kennedy and Baker.37 Several schemes for direct injection of liquid anesthetics into a closed circuit following the method of Lowe or the square-root-of-time model have been evaluated.38–41 However, our model is simpler because an open-circuit ventilator is used instead of an anesthetic breathing circular circuit. As a result, the circuit compartment can be disregarded and neither rebreathing nor fresh gas flow need to be considered. Other assumptions regarding cardiac output (CO), lung volumes, ventilation, and gas uptake are described in detail in the Appendix 1 (available at www.anesthesia-analgesia.org). An empirical formula was used to adjust for the loss of anesthetic through the filter (Appendix 1).
The model was implemented using an Excel spreadsheet. When patient weight, sevoflurane target concentration, minute ventilation and induction time are entered into the "patient data and settings" box, both initial and maintenance infusion rate are calculated and displayed in the cells of the bottom right-hand box (Table 1).
|
|
Computation of Performance Parameters
Performance parameters were determined following the methods described by Varvel et al.28 and used in many studies of target-controlled IV anesthesia.29,30,42,43 The basic estimation of the accuracy of each measurement (against the target) is the performance error (PE) calculated as the difference between the measured concentration (Cm) and the predicted (target) concentration (Cp) related to the predicted concentration:
|
|
This way in an ideal system Cm and Cp are equal and PE is zero %.
From the PE at each time-point, two basic parameters are calculated: (1) The median absolute PE (MDAPE, %) which is the median of the absolute values of PE. This MDAPE reflects the precision of the model and is the single best predictor of clinical acceptability of the performance of the model. For example, a MDAPE of 10% indicates that the median of the obtained values will be 10% above or below the target. This way an ideal MDAPE would be zero %. (2) The median PE (MDPE, %) is the median of the sometimes positive and other time negative values of PE. It measures bias which may be above or below the target. For example, a negative MDAPE (i.e., –6%) indicates that the median of the obtained values will be below the target, disregarding its absolute value which is expressed by the MDAPE. Therefore, MDAPE (precision) and MDPE (bias) are the main parameters used to determine the model’s predictive performance in a single patient (intrasubject analysis). For modeling performance in a group of patients, the mean value (and standard deviation) of the individual values of each parameter can be taken ("two-stage approach"). This way, all patients are equally weighed. In addition, a "pooled data approach" was also performed. This method uses all of the measurements from all individuals as if they came from one "average" individual.
Statistical Analysis
All analyses were performed using descriptive statistics from SPSS for Windows (version 11.0, SPSSTM, Chicago, IL). Two-stage analysis values are shown as mean, standard deviation, and confidence interval of 95%. For the pooled data analysis, median, minimum, and maximum values are shown. Box and whisker plots indicate median, 25th and 75th percentile (box), 5th and 95th percentile (whisker), and outlier values. Mann–Whitney’s U test for independent samples was used to test the difference between the means of MDAPE and MDPE at each target concentration for both 6-h as well as the rise and lower results. P < 0.05 was considered statistically significant.
|
RESULTS |
|---|
|
TopAbstractIntroductionMETHODSRESULTSDISCUSSIONAppendix 2REFERENCES |
|---|
|
The performance accuracies of the model with the two-stage approach for the 6 h sedation are shown in Table 3. The results of the pooled data approach can be seen in Figure 2. Figure 3 shows the time course of PE for all cases during the 6-h period. Except for occasional transients, PEs for all cases were between –25% and +25%. No statistically significant differences were found between the means of all performance parameters when compared at the studied concentrations. This means that the predictive performance of the model is the same for the different target sevoflurane concentrations studied.
|
|
|
The performance accuracy of the model and the results of the two-stage approach when it is adjusted to increase or decrease the alveolar concentration are shown in Table 4 and the time course of PE for all these patients is shown in Figure 4. Except for occasional transients, PEs for all cases were between ±20%.
|
|
Changes in mean arterial blood pressure and heart rate were below 10% of control values in all patients, and thus there was no need to modify the infusion rate based upon hemodynamic changes.
|
DISCUSSION |
|---|
|
TopAbstractIntroductionMETHODSRESULTSDISCUSSIONAppendix 2REFERENCES |
|---|
The model was intended as a simple guide to manually adjust liquid sevoflurane infusion rate and as a way to predict the ET concentration within acceptable limits. As shown, the model’s predictive performance has a 3.9% average error in the 6-h analysis, this percentage being slightly higher for increases (5.0%) and decreases (5.5%) of target concentration. No significant differences are found in the MDAPE for the different sevoflurane target concentrations studied.
The model’s predictive performance is greater than other more complex predictive models of anesthetic uptake, which include changes in CO, temperature, etc., designed for liquid anesthetic bolus administration in closed system and for vaporizer adjustment during minimum flow anesthesia. The MDAPE values obtained with these models were 10.9% for sevoflurane,37 13%–17.2% for desflurane,30,32 and 18.7%–19.1% for enflurane.44 The best predictive values published are those obtained by the simpler model by Kennedy et al.30 and Heffernan et al.36 The constraints of the model have already been pointed out by Hendrickx et al.39 Despite the greater simplicity, the better predictive performance of our model may be partially due to its use with an open circuit, which means that all elements not related to uptake by the organs over time are excluded. Only losses through the ACD are to be added, and when the adjusted target concentration and minute ventilation are constant, losses are constant and equivalent to 50%–65% of the administered anesthetic. This fact can be the other main factor that minimizes the error of our model.
Likewise, predictive performance of the model (for any concentration and infusion time studied) is not only within acceptable limits for IV sedative and anesthetic infusion systems,45 but also below values reached in different studies29 using widely accepted clinical models.46,47 These large differences in the predictive performance of continuous IV infusion models among patients stem from the different evaluation methods29 and from inter-individual variations at physiological, genetic, or environmental levels, which have been analyzed by Gepts.46 Again, the better results obtained by our simpler model are explained basically by the high proportion of losses through the filter, which minimizes the error of the patient uptake model. The accurate calculation of these losses, which also includes conditioning factors (minute ventilation and ET sevoflurane), allows for reliable adjustment of the anesthetic administered.
On the other hand, the model demands hemodynamic stability, since the uptake is calculated based on the normal blood flow to the organs. In clinically unstable patients and/or while receiving other drugs such as narcotics, muscle relaxants, and benzodiazepines, a more frequent dose-titration may be needed. In this case a more frequent control of end-expiratory concentration of gas and monitoring of its sedative effects would be recommended. Obviously, the more complex models which consider the variation of CO, of ventilation/perfusion mismatch and other factors, could improve the performance of the system. However, increasing the complexity of a model does not always improve its predictive performance.44
A potential error source may be linked to the gas analyzer. The Vamos monitor’s accuracy is ±0.15% of the volume for sevoflurane, which is most acceptable for clinical use. However, because its display offers just one decimal, the minimum error calculated for each data-point is from 14% for a 0.7% ET sevoflurane to 6.6% for 1.5%. Using an agent monitor with finer resolution may have produced better results than those obtained. Another potential error source is the infusion syringe used. The volumetric accuracy of the Ivac P7000 is ±2%. In general terms, all syringes have similar accuracy and have been largely used in target-controlled infusion systems and for the injection of liquid anesthetics in a circle system,39 given their suitability for these types of applications. Other syringe types are not likely to improve the results. In our patients no problems in dosing associated with the pump, the syringe, or the infusion system were observed. However, at temperatures above 30°C both refill of the syringe and temporary removal of the filter can produce transient increases or decreases respectively in the ET concentration of anesthetic48,49 making the use of an anesthetic agent analyzer mandatory. Finally, although the predictive power of the algorithm is excellent, using caution to prevent errors in dosage is highly recommended. The anesthetic agent analyzer should be provided with alarms for a high-sevoflurane ET concentration, not just to detect predictable fluctuations from syringe refills, but also as a defense against human error and equipment failure. Ideally, this monitor should be programmed to automatically stop the infusion pump and sound an alarm. However, the standard of monitoring requirements when this device is used are yet to be described.
When drawing conclusions about the model, it is worth analyzing the clinical meaning of the results. The 3.9% mean PE (average of each patient’ medians, MDAPE) in the 6-h study shows that by adjusting the infusion rate to reach, for instance, a 1.5% target sevoflurane concentration, 50% of the obtained ET values will range between 1.44% and 1.56%. At the same time, the small negative bias means that in 95% of cases, ET sevoflurane values will be slightly below the target values. This is important for safety reasons, as overdosing is almost impossible and ET-concentration readjustments can be done by simply increasing the infusion rate following the increasing scheme which offers a similar predictive accuracy.
Another important feature is the fact that the infusion rate is readjusted once "induction time" has elapsed and is only changed once every hour. This makes bedside application easier, as one can make it coincide with hourly nursing evaluations in critical care patients. More hourly adjustments, as in circle system models of diverse complexity, e.g., seven liquid enflurane injections per hour,44 would not be clinically feasible in long ICU sedations. With the easy-to-handle model described, a computer-controlled system is no longer essential. A full chart of infusion regimes for patients of different weights and minute volumes is in Appendix 2 (available at www.anesthesia-analgesia.org).
Finally, the manufacturer’s recommendation for sevoflurane infusion may be useful to guide ACD use in clinical practice but many readjustments in infusion rate must be performed in order to avoid frequent PEs. The instructions for device use include a chart for calculating syringe pump rate to reach 0.5% gas concentration at different minute volumes. For 7.0 L/min, recommended rates are 3.0 and 2.2 mL/h for induction (30 min) and maintenance, respectively. Following our model, an initial pump rate of 3 mL/h for 30 min (induction) would produce an ET sevoflurane concentration of 0.54% (which would be displayed as 0.5% in a clinical gas-analyzer). The following reduction to 2.2 mL/h would produce a concentration of about 0.51% at the end of the first hour. This way PE for this period would be very good (<10%). However, from this initial period (1–2 h) the PE would increase progressively if the infusion rate were not reduced each hour in an exponential fashion (2.0-1.8-1.7- 1.6-1.5-1.4 ..... 1.1). That is to say, if 2.2 mL/h were maintained, the ET sevoflurane concentration would be close to 0.8% after 6 h of sedation and 1% at equilibrium (PE of 80%–100%). However, instructions for use of the ACD also stated that for changing the concentration, the relationship between concentration and syringe pump rate is nearly linear. This means that if a concentration of 1% were observed in the gas monitor, halving the infusion rate would reduce the concentration to half. If this rule was applied whenever the observed concentration shifted from the desired concentration, the infusion rate would be adjusted close to that obtained with our scheme.
In conclusion, this study documents the excellent predictive performance of a scheme for a manually adjusted, but pump-driven, infusion of liquid sevoflurane for use with the ACD in postoperative ICU patients with no respiratory pathologies. The predictive features help system safety and hourly adjustment facilitates its clinical use. The easy administration of inhaled sedation drugs in critical care using the ACD offers new approaches to sedation in the ICU,12 not only because of the better pharmacological features of inhaled drugs, but also because of the possibility of monitoring the concentration administered, a key safety aspect in continuous sedative administration. Used with accurate and simple infusion schemes, devices such as the ACD can improve sedation prospects in critical care patients.
|
Appendix 2 |
|---|
|
TopAbstractIntroductionMETHODSRESULTSDISCUSSIONAppendix 2REFERENCES |
|---|
|
|
Footnotes |
|---|
Accepted for publication December 3, 2007.
|
REFERENCES |
|---|
|
TopAbstractIntroductionMETHODSRESULTSDISCUSSIONAppendix 2REFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  K. D. Rohm, A. Mengistu, J. Boldt, J. Mayer, G. Beck, and S. N. Piper Renal Integrity in Sevoflurane Sedation in the Intensive Care Unit with the Anesthetic-Conserving Device: A Comparison with Intravenous Propofol Sedation Anesth. Analg., June 1, 2009; 108(6): 1848 - 1854. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. W. Sturesson, A. Johansson, M. Bodelsson, and G. Malmkvist Wash-in kinetics for sevoflurane using a disposable delivery system (AnaConDa(R)) in cardiac surgery patients Br. J. Anaesth., April 1, 2009; 102(4): 470 - 476. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
