Immediately after open heart surgery, patients may be hypothermic and haemodynamically unstable. Mechanical ventilation is continued in these patients postoperatively until haemostasis is achieved, and a period of haemodynamic stability is established. Hence, sedation has become an integral part of patient care in the postanaesthesia care unit (PACU) in order to minimize patient discomfort.
Recent studies1,2 have demonstrated that using protocols to guide sedation in various groups of critically ill patients decreases the length of mechanical ventilation as well as the stay in ICU and reduces the physician, nursing and other paramedical resource use. Although a wide range of drugs have been used for postoperative sedation in cardiac surgery patients, propofol has gained popularity recently owing to its rapid onset and offset of action.3,4
Conventional tools to assess depth of sedation of patients, such as the Modified Ramsay Scale (MRS),5 are subjective and prone to inconsistencies. In addition, acoustic and tactile stimulation involved in this assessment can interfere with patient care, as these manipulations may cause undesired agitation and arousal resulting in increased requirement for sedation. Hence, an objective, reliable and replicable parameter such as bispectral index (BIS) that can be measured without stimulating the patient is an attractive option to monitor sedation in the critically ill. BIS has been empirically demonstrated to correlate with behavioural measures of sedation.6
Closed-loop control using a BIS-directed closed-loop anaesthesia delivery system (CLADS) has been used in general anesthesia7 and for procedural sedation.8 The algorithm in CLADS has been developed in PGIMER and is not available commercially. Postoperative propofol sedation controlled by a closed-loop system using BIS has not been studied so far. Automated delivery of sedation based on pharmacodynamic feedback from the patient can be an aid to the anaesthesiologist to devote attention to other aspects of postoperative care such as haemodynamic control. The current study was designed to assess the feasibility of BIS-directed CLADS for postoperative sedation of cardiac surgery patients using propofol and to compare it with manual control.
CLADS is a patented (502/DEL/2003) propofol delivery system that uses BIS as the ‘controlled variable’ and a standard infusion pump as ‘actuator’.7 The ‘control algorithm’ is based on the relationship between various rates of propofol infusion (producing different plasma concentrations) and BIS, taking into consideration the pharmacokinetic variables (distribution and clearance) that were established in the developmental stage of CLADS. The algorithm alters the rate of propofol infusion to steer the BIS to the set target, taking into account the existing BIS, time elapsed since the initiation of infusion, pharmacokinetics, time delay factor between sensing and averaging of BIS data, time delay factor between the change in infusion rate and actual change in plasma concentration of propofol as well as the peak effect of propofol.9 An IBM compatible Pentium 4 personal computer (2 GHz) was used to implement the control algorithm as well as to provide a user interface and to control communication with the infusion system (Pilot C; Frasenius Vial, Brezins, France) and BIS and vital signs monitor (AS5; Datex Ohmeda Division, Helsinki, Finland, and GE Healthcare, Singapore) via RS 232 serial ports. CLADS can be operated in two different modes: manual and automatic. In manual mode, the rate of propofol infusion is controlled manually. In automatic mode, the system automatically controls the rate of propofol infusion based on BIS feedback.
After approval from the Institutional Ethics Committee and written informed consent, 41 American Society of Anesthesiologists (ASA) status 2 or 3 patients aged 18–65 years, who underwent elective open heart surgery under general anaesthesia, were included. Anaesthesia was induced using a standardized technique with fentanyl and propofol. Patients were ventilated using an oxygen/air mixture. Neuromuscular blockade was not reversed at the end of surgery, and patients were transferred to the PACU. Patients with hypersensitivity to propofol, weight more than 30% different from ideal body weight, preexisting neurological disorder, ejection fraction of less than 35%, preexisting lung disorder and history of use of any psychoactive medication were excluded. Patients requiring more than two inotropic agents and those who were on intraaortic balloon pump augmentation at the end of surgery were also excluded as they were deemed haemodynamically unstable.
In PACU, the MRS was assessed once patients were haemodynamically stable. The study design began at this point and continued for 4–8 h during mechanical ventilatory support and weaning. Patients were randomly assigned to two groups using computer-generated random number sequence, and the allocation was concealed in an opaque sealed envelope. The allocation sequence was generated by one of the investigators, P.J.M., and the patients were assessed for eligibility and enrolled by another investigator, G.D.P. Data collection was carried out by A.S., who was not involved in the management of either group of patients. In the CLADS group, patients received a continuous infusion of propofol using CLADS, keeping a target BIS of 70. CLADS automatically calculated and titrated the subsequent propofol infusion rate according to the weight of the patient and the target BIS. In the manual group, patients received a continuous infusion of propofol at a rate adjusted by the anaesthesiologist in order to keep a target BIS of 70. The attending anaesthesiologist in the PACU adjusted the infusion to steer the BIS to the target value. If the mean arterial blood pressure (MAP) exceeded 25% of baseline or the heart rate (HR) was 25% above baseline, analgesia was supplemented with 0.05 mg kg−1 morphine. In the event of persistent hypertension or tachycardia, hypovolaemia, fever and other treatable causes were sought and treated appropriately. For intractable hypertension or tachycardia, nitroglycerine infusion or esmolol was administered, respectively. Similarly, if there was a fall in MAP of greater than 25% of baseline in the presence of normovolaemia, inotropic support was increased in order to maintain MAP within 25% of baseline.
Once the specific weaning criteria were met, weaning was commenced. The weaning criteria were
- temperature more than 36°C and less than 37.5°C,
- haemodynamically stable,
- adequate oxygenation (paO2 >70 mmHg at FiO2 <0.35 and positive end-expiratory pressure 5 cmH2O),
- adequate oxygen transport (cardiac index >2.2 l min−1 m−2 with minimal inotropic support, dopamine <10 μg kg−1 min−1, adrenaline <0.1 μg kg−1 min−1),
- adequate recovery of neuromuscular transmission,
- minimal chest tube drainage (<400 ml h−1 in the first hour, <200 ml h−1 in the next 2 h, <100 ml h−1 in the next 4 h),
- adequate urine output (>0.8 ml kg−1 h−1) and
- acceptable chest radiograph and arterial blood gas values.
The propofol infusion was continued during the weaning process and discontinued when the patient was breathing spontaneously at a regular rate and at a pressure support of less than 15 cmH2O with acceptable arterial blood gas values (pH ≥7.35, paO2 >70 mmHg and base deficit <10 mEq l−1). The trachea was extubated when the standard criteria were met (Table 1).
Propofol infusion was not stopped at the time of assessment of the sedation score. The cumulative time during which CLADS did not function properly because of errors in the BIS calculation due to a low signal quality index (SQI) or syringe pump-related defects was excluded from the total duration of sedation to obtain the valid CLADS sedation time. Haemodynamic parameters, such as HR and arterial blood pressure, were recorded continuously during the study period. The percentage of study time that HR and MAP was within ±25% of the baseline was used to assess HR performance and MAP performance. These, along with postoperative use of inotropes and vasoactive medications, were estimated to indicate the haemodynamic effects of propofol sedation.
Physiological data were presented as mean (SD) and time intervals were presented as median (interquartile range). The primary outcome measure was the cumulative duration of time BIS remained within ±10 of the target BIS. The performance of the system was assessed as a secondary outcome measure using the method described by Varvel et al.10
- Offset = Measured BIS − target BIS
- Performance error
- Median performance error (MDPE)
- This is a measure of bias and describes whether the measured values are systematically either above or below the target value.
- Median absolute performance error (MDAPE)
- This is a measure of inaccuracy of the control method.
- This is an index of time-related changes in performance and measures the intraindividual variabilities in the performance errors.
- This is calculated for the ith individual as the slope obtained from linear regression of that individual's |Performance errorij| against time.
The other secondary outcome measures were the duration of sedation and the interval from cessation of sedation to tracheal extubation. Differences between the groups were analysed using the unpaired t-test for parametric data and the Mann–Whitney U test for nonparametric data. Categorical data were analysed using the chi-squared test. All analyses were performed using SPSS version 15.0 for Windows (SPSS Inc., Chicago, Illinois, USA), and a P value of less than 0.05 was considered significant. As there is no published literature on the performance of automated systems for ICU sedation, we designed the study as a pilot trial and estimated post-hoc power.
Forty-one patients were randomized in the study from March 2008 to October 2008 (Fig. 1). One patient was excluded as the intervention was discontinued on account of excessive chest drain and haemodynamic instability leading to re-exploration of the patient. The demographic characteristics of the two study groups were similar (Table 2). Valid CLADS sedation time was 246.4 ± 152.4 min, which amounted to 88.0 ± 14.7% of total sedation time (Table 3). The corresponding duration in the manual group was 84.6 ± 21.9% of the total sedation period. The time from cessation of sedation to tracheal extubation was significantly lower in the CLADS group than in them manual group (P = 0.04). The total duration of mechanical ventilation and sedation was longer in the manual group than in the CLADS group. The duration of stay in PACU was similar in both the groups.
The amount of propofol and morphine used during the study period was similar in the two groups (Table 4). However, the attending clinician made an average of 19 adjustments to the rate of propofol administration in the manual group, whereas none were required in the CLADS group. The percentage of time HR and MAP was within ±25% of the baseline as well as the amount of dopamine and nitroglycerine used postoperatively were comparable. No serious adverse event attributable to the closed-loop system was observed.
CLADS was able to maintain the BIS within ±10 of the target for significantly longer duration than in the manual control group (P = 0.002) (Table 5). The mean offset (P = 0.04), performance error (P = 0.07) and MDAPE (P = 0.02) were lower in the CLADS group (Table 5). Although wobble was similar in the two groups, divergence was significantly lower in the CLADS group (P = 0.007).
Computer-controlled drug delivery in the operating room is an exciting area with promising clinical utility. Automated delivery of sedation based on pharmacodynamic feedback from the patient can be an aid to the intensivist who can devote attention to haemodynamic stability and other aspects of critical care. The current study has described the feasibility of the closed-loop feedback control system using the BIS for postoperative sedation and compared its efficacy with manual human control of propofol administration after cardiac surgery. This clinical trial was conceived following the success of the system to administer propofol anaesthesia in ASA status 2 or 3 patients undergoing noncardiac surgery7 and open heart surgery.9
CLADS provided a clinically adequate depth of sedation in all patients during the period of automatic control, which was valid for 88% of all sedation time. CLADS did not function during the remaining time period either due to poor SQI or syringe pump-related defects. The corresponding time in the manual control group when a valid BIS was available was 84.6% of total sedation time. Characteristically, artefacts, such as electromyography (EMG), have most of their energy in a frequency range that is different from that of electroencephalography (EEG) and influences SQI. The bioelectric amplifier can band-pass filter the signal, passing the EEG and attenuating the EMG.11 However, the effect of this artefact, though reduced, cannot be completely ruled out. Intraoperatively, and under adequate neuromuscular relaxation, EMG artefacts in EEG are not a serious cause for concern. In our study, CLADS was functional for a considerable period of time despite the presence of neuromuscular activity.
The total duration of sedation as well as mechanical ventilation was longer in the manual group than in the CLADS group, though it did not reach statistical significance. The interval between stopping sedation and extubation was significantly shorter in the CLADS group. The longer duration of sedation with a similar rate of propofol infusion may explain the delayed awakening and extubation noted in the manual group.
Alteration of pharmacokinetics, including redistribution, elimination half-lives and clearance of propofol, as a consequence of cardiopulmonary bypass (CPB) and hypothermia has been studied previously.12–14 Although it is expected that propofol pharmacokinetics would return to normal after CPB, the time frame within which it returns and the effect of rewarming and reperfusion on pharmacokinetics during this period has not been studied yet. Therefore, it is difficult to comment on the influence of pharmacokinetics on awakening from propofol sedation after CPB.
In the CLADS group, an MDPE of 1.4 indicates a positive bias, that is, the median measured BIS was 1.4% higher than the target BIS. MDAPE is a measure of inaccuracy, that is, error. An MDAPE of 2.6 in the CLADS group indicates that 50% of the BIS values were within 2.6% of the target BIS. The MDAPE in the manual group was 8.6, indicating that more of the BIS values were away from the target BIS. Wobble was similar in both the groups, indicating the ability of CLADS to maintain a consistent depth of sedation, as it was in the manual group. A positive value of divergence indicates progressive widening of the gap between the target and the measured value, whereas a negative value indicates that the measured values converge on the predicted values (convergence). A mean divergence of −0.16 in the CLADS group indicates that our system has a tendency to decrease performance error with time. These figures are comparable to the performance of the system reported by Leslie et al.8 in patients undergoing colonoscopy under sedation.
There are various methods of automated drug delivery.15 A proportional integral derivative controller calculates the infusion rate using a mathematical formula based on the difference between measured effect value and chosen target and is blind to the effects of metabolism or hysteresis between administration and effect. Model-based adaptive control,16 on the contrary, takes into account knowledge of drug effects, compares the predicted value of the control signal with the actual values obtained from the patient and modifies the model parameters accordingly. Our model of drug delivery is based on the latter concept, and the inbuilt delay in the decision before changing the infusion rate takes into account the hysteresis between drug delivery and effect.
The haemodynamic alterations during the postoperative period reflect the care provided. The haemodynamic control as reflected by the MAP and HR performance was equally as good in the CLADS group as in the manual group. This haemodynamic performance also indirectly reflects the safety of the closed-loop system, as there was no excessive administration of propofol producing haemodynamic instability. This was achieved with lesser sedation in the CLADS group. Moreover, the large number of times propofol infusion rates were changed manually translates to a substantial involvement of human resource in maintaining an appropriate depth of sedation, further increasing the stress of the intensive care specialist. Use of CLADS could free the specialist from this daunting task and enable the intensivist and other caregivers to concentrate on other demanding aspects such as haemodynamics.
The high incidence of propofol rate modifications per hour during the manual control demonstrates that titration in this group was performed actively. The task was probably taken up as a challenge by the anaesthesiologist to maintain BIS within the required limits. This can be explained by the Hawthorne effect, an increase in worker productivity produced by the physiological stimulus of being singled out and made to feel important, thus transforming the manual group to an active control group.17 The number of rate adjustments would probably have been lower in routine care. The investigator bias thus cannot be ruled out as a confounding factor in the current study, though its significance cannot be determined.
The post-hoc power analysis determined the power of the study to be 96% for the primary outcome of duration of time that the BIS remained ±10 of the target BIS. After having demonstrated the feasibility, the next step would be designing an adequately powered randomized controlled trial to demonstrate the clinical advantages of the duration of sedation and mechanical ventilation.
Despite the limitations mentioned, our study does prove the feasibility and efficacy of closed-loop control of sedation in postoperative cardiac surgery patients compared with manual control and supports the use of closed-loop systems in the ICU.
The project received a financial grant from the Department of Information Technology, Ministry of Communication, Government of India.
1 Kress JP, Pohlman AS, Hall JB. Sedation and analgesia in the intensive care unit. Am J Respir Crit Care Med 2002; 166:1024–1028.
2 Kress JP, Pohlman AS, O'Connor MF. Daily interruption of sedative infusion in critically ill patients undergoing mechanical ventilation. N Engl J Med 2000; 342:1471–1477.
3 Beller JP, Pottecher T, Lugnier A, et al
. Prolonged sedation with propofol
infusion in ICU patients: recovery and blood concentration changes during periodic interruptions in infusion. Br J Anaesth 1988; 61:583–588.
4 McMurray TJ, Collier PS, Carson IW. Propofol
sedation after open heart surgery: a clinical and pharmacokinetic study. Anaesthesia 1990; 45:322–326.
5 Ramsay MAE, Savege TM, Simpson BR. Controlled sedation with alphaxolone – alphadolone. BMJ 1974; 280:656–659.
6 Liu J, Singh H, White PF. Electroencephalogram bispectral analysis predicts the depth of midazolam induced sedation. Anesthesiology 1996; 84:64–69.
7 Puri GD, Kumar B, Aveek J. Closed-loop anaesthesia delivery system (CLADS) using bispectral index
: a performance assessment study. Anaesth Intensive Care 2007; 35:357–362.
8 Leslie K, Absalom A, Kenny GNC. Closed loop control of sedation for colonoscopy using the bispectral index
. Anaesthesia 2002; 57:690–709.
9 Agarwal J, Puri GD, Mathew PJ. Comparison of closed loop vs. manual administration of propofol
using the bispectral index
in cardiac surgery
. Acta Anaesthesiol Scand 2009; 53:390–397.
10 Varvel JR, Donoho DL, Shafer SL. Measuring the predictive performance of computer-controlled infusion systems. J Pharmacokinetic Biopharm 1992; 20:63–93.
11 Rampil IJ. A primer for EEG signal processing in anaesthesia. Anesthesiology 1998; 89:980–1002.
12 Russell GN, Wright EL, Fox MA, et al
-fentanyl anesthesia for coronary artery surgery and cardiopulmonary bypass. Anesthesia 1989; 44:205–208.
13 Hammaren E, YLI-Hankala A, Rosenberg PH, Hynynen M. Cardiopulmonary bypass induced changes in plasma concentrations of propofol
and in auditory evoked potentials. Br J Anaesth 1996; 77:360–364.
14 Hynynen M, Hammeren E, Rosenberg PH. Propofol
sequestration within extracorporeal circuit. Can J Anaesth 1994; 41:583–588.
15 Ohara DA, Bogen DK, Noordergraf DD. The use of computers for controlling the delivery of anesthesia. Anesthesiology 1992; 77:563–581.
16 Schwilden H, Stoeckel H, Schuttler J. Closed loop feedback control of propofol
anaesthesia by quantitative EEG analysis in humans. Br J Anaesth 1989; 62:290–296.
17 Franke RH, Kaul JD. The Hawthorne experiments: first statistical interpretation. Am Sociol Rev 1978; 43:623–643.