Anesthesiologists are pioneers in terms of patient safety. In an effort to continue to improve safety, researchers have developed devices incorporating closed-loop systems for anesthetic delivery.1,2 Closed-loop systems consist of an operating system that receives information from a target-control variable. According to the information received, the operating system, based on the built-in algorithms, will automatically adjust the infusion rate of the anesthetic drug. Feedback closes the loops of the system, allowing automatic maintenance of the predetermined targets.1
Most of the tested closed-loop systems are semiautomated anesthetic systems because they include only 1 loop to automatically administer a single pharmacologic agent.3,4 More than 2 decades of research testing semiautomated anesthesia devices using closed-loop administration of propofol have demonstrated that they are efficient, safe, and allow superior steadiness of the control variable compared with manual control in noncardiac surgery.5–8 Increased quality and precision of drug administration are mostly because of frequent sampling of data from the control variable. Consequently, a greater number of drug delivery rate modifications are compared with the standard administration. Finally, it has been reported that a completely automated total IV anesthesia drug delivery system managing all 3 components of anesthesia could outperform manual administration.9
The increased quality of completely automated anesthesia delivery devices encompasses not only the advantage of standardized high-quality anesthesia administration but it also allows for a reduction in workload. Thus, the attending clinician can focus on tasks that require human intelligence such as reasoning, problem solving, making critical decisions, and effective communication. These considerations are paramount because cardiac surgery exposes patients to a higher risk of causalities10 and demands significant multitasking from the anesthesiologist, constantly monitoring hemodynamic changes, including transesophageal echocardiography (TEE). One review of the literature specifically evaluating mortality in anesthesia reports that the highest number of intraoperative deaths occurs during cardiac anesthesia.11 Currently, cardiac anesthesiologists have to cope with an aging population presenting for surgery with severe cardiac diseases and significant comorbidities increasing the risk for complications even more. Thus, it is crucial that cardiac anesthesiologists react promptly to focus their attention on critical intraoperative events so that they can communicate efficiently to solve a potential life-threatening condition.
Therefore, we conducted a pilot study using a previously described pharmacologic robot controlling all 3 components of anesthesia: hypnosis, analgesia, and muscle relaxation.9 This completely automated anesthetic robot allows total IV anesthesia delivery from induction to the end of the surgery. In the present trial, we investigated the hypothesis that a completely automated anesthesia delivery for on-pump cardiac surgery is feasible using a pharmacologic robot. The primary objective of this study was to determine the feasibility of successful automated cardiac anesthesia without manual override by the attending anesthesiologist. The secondary qualitative objectives consisted of assessing clinical performances and controller performances of the anesthesia delivery system.
This was a human research, prospective, observational, single-center study conducted in the Department of Anesthesia and Critical Care of Bordeaux University Hospital (Service d’Anesthésie-Réanimation II, CHU Bordeaux, France). The report includes every item of the STROBE checklist for a prospective observational trial. The current trial was conducted from March 1 to May 1, 2014, after approval from our Institutional Ethics Committee (Comité de Protection des Personnes Sud-Ouest et Outre Mer III, Bordeaux, France/agreement number DC 2014/20) and authorization from the “Commission Nationale de l’Informatique et des Libertés” (registry number 1754349v0). The Ethics Committee considered research focusing on evaluation of an innovative combination of anesthetic drug infusions that are routinely used, which does not vary from common practice, would not require written informed consent from individual study participants. The study was also registered on clinicaltrials.gov (NCT02145585). Inclusion criteria were patients aged from 18 to 90 years undergoing elective cardiac surgery using cardiopulmonary bypass (CPB). Pregnant women, comatose patients, patients with muscle disease, dementia, and those who were allergic to propofol and/or remifentanil were excluded. Twenty patients undergoing cardiac surgery with CPB were enrolled in this prospective observational study.
System Specifications and Description of the Special Cardiac Anesthesia Features Integrated to McSleepy
All patients received anesthesia by a pharmacologic robot, which allows for completely automated IV anesthesia delivery. The pharmacologic robot tested in the present investigation is nicknamed “McSleepy,” which was developed using LabVIEW (LabVIEW 2010, National Instruments Corporation, Austin, TX). This system has been thoroughly described elsewhere.9 This pharmacologic robot combines a proportional integral derivative-controller with a controlling feedback system using self-adaptive algorithms. This expert-based robot integrates the 3 components of general anesthesia: hypnosis, analgesia, and muscle relaxation. The bispectral index (BIS) was used as the control variable for hypnosis to calculate propofol infusion rates to maintain a BIS target of 45. Nociception was controlled through a modified version of the Analgoscore called NociMap. The latter uses exactly the same algorithm of the Analgoscore12 derived from heart rate (HR) and mean arterial blood pressure (MAP) to indirectly determine intraoperative pain. By using the same criteria, it measures the offset percentage between the measured and the target value of both HR and MAP using expert-based rules as described in Table 1. The only difference is the scale of presentation of the score. To facilitate efficient comprehension of the score, the scale ranges were changed. Instead of starting from −9 (very profound analgesia) to +9 (very superficial analgesia), the NociMap score starts from −100 (very profound analgesia) to +100 (very superficial analgesia). The NociMap was used as the control variable to calculate the new infusion of remifentanil, which corresponds to the multiplication of 3 factors: the correction factor, K1 factor, and K2 factor. The first is derived by the NociMap (details are described in Table 1), whereas, the other 2 consider trends in offset from the target values over time. Depth of neuromuscular blockade was determined through phonomyographic signals13 received from the adductor pollicis. Train-of-four ratios were performed automatically every 15 minutes at the adductor pollicis muscle. The results of train-of-four ratios were sent to the anesthesia robotic system that gave a bolus of rocuronium automatically when the ratio was >25 and only if 5 minutes occurred after the previous rocuronium bolus.
A vital signs monitor incorporating a Philips BIS module (IntelliVue MX700 patient monitor, Philips Healthcare, Suresnes, France) was used to obtain control variables, whereas 3 standard infusion pumps (Graseby 3400, Graseby Medical, Watford, Hertfordshire) were used as actuators for the injection of the IV anesthetic drugs. A computer touchscreen all-in-one (Samsung DP700, PC, Intel Core i5-3470T 2.90GHz Processor, UE) applied the algorithm, provided the user interface, and closed the loops between the control variables and the syringe pumps. Communication through RS-232 ports allows the computer to control the syringe pumps. Manual override was possible at any time, either at induction or during maintenance, with a touch of a button in case the anesthesiologist judged it necessary.
On activation of the automated anesthesia system, a dialog box prompted the user to enter patient data (ie, patient identification, weight, height, age, gender, ASA physical status, and type of surgery). New safety features were specifically designed and added into the algorithm to perform cardiac anesthesia. These features were activated when cardiac surgery was selected on the “type of surgery” item of the dialog box. By selecting this item, the setup window offers the possibility of selecting new patients’ characteristics that could interfere with the anesthetic strategy. These characteristics are β-blocker therapy, the presence of atrial fibrillation, and whether patients tolerate volumes (left ventricular ejection fraction [LVEF] <40%). After confirmation that the data entered were correct, the system automatically started the IV anesthesia according to the improved algorithm for cardiac anesthesia installed into the computer. The induction for cardiac anesthesia starts with a remifentanil infusion dose of 0.5 μg kg−1 min−1 for the first 2 minutes, 0.2 μg kg−1 min−1 for the third minute, and 0.1 μg kg−1 min−1 for the rest of the induction. Three minutes after remifentanil infusion, a propofol bolus was delivered with a 1 mg kg−1 dose during a period of 90 seconds. After the first bolus, the system waited for a different length of time depending on patients’ LVEF. If the patient had a LVEF < 40%, the system waited for 90 seconds for the BIS to decrease <60. In contrast, when the LVEF was ≥40%, the system waited for 60 seconds for the BIS to decrease <60. In both cases, if after the first bolus the BIS values were still >60, the system injected a second 0.5 mg kg−1 propofol bolus. After this bolus, the system waited for 60 seconds for the BIS values to decrease <60. Three additional boluses of 0.5 mg kg−1 were allowed to reach BIS scores <60. For safety reasons, further boluses were allowed only if the clinician pressed the bolus button. When BIS values were <60, a 0.9 mg kg−1 rocuronium bolus was delivered automatically. During the induction phase, the expert-based system attempts to imitate the decision-making expertise of a qualified cardiac anesthesiologist. At completion of the induction, the user was transferred to a new window displaying BIS values, NociMap values, continuous infusion rates, drug doses, live TEE video, and vital signs (ie, systolic and diastolic blood pressure, MAP, and HR) automatically appeared. This interface allows the anesthesiologist to indicate the exact time points of surgery and offers the option to administer propofol boluses and rocuronium bolus pressing on a touchscreen button. During maintenance, lowest and highest continuous infusion rates were given to avoid drugs undershooting or overshooting. When lower limits of the BIS are attained, the robotic system stops propofol infusion automatically. This was done to prevent any cardiovascular side effect. When the heart was arrested, CPB was instituted using pulsatile perfusion with a frequency ranging from 60 to 70 beats min−1 allowing the NociMap algorithm to properly deliver remifentanil. However, in case of low arterial blood pressure with a constant HR during CPB, a minimal remifentanil continuous infusion of 0.05 μg kg−1 min−1 was guaranteed to allow optimal analgesia even during this stage. During the emergence phase, when the “15 minutes to the end” time point button was pressed, administration of rocuronium was blocked to avoid profound neuromuscular blockade that could jeopardize early tracheal extubation if targeted. The main algorithm flowchart has been described previously.14 Before clinical use, the cardiac anesthesia version of the robotic system was tested. The test consisted of running cases in a simulation mode to ensure proper execution of the new programming specifically written for cardiac anesthesia, as well as correct communication between the controllers and the syringe pumps.
Premedication with 0.1 mg kg−1 oral midazolam was performed on the surgical ward. On arrival in the operating room, patients were transferred to a water-heated mattress placed on the operating table. Patients were monitored with physiologic variables including a 2-channel electrocardiogram, pulse oximetry, BIS monitor, and arterial line. Once the patient was monitored and a 16-/14-gauge IV line inserted, robotic anesthesia was started. At the end of the induction and after insertion of the endotracheal tube, a triple-lumen central venous catheter was introduced in the right internal jugular vein. Afterward, a TEE probe was placed to perform a systematic comprehensive TEE examination encompassing all cardiac structures. The TEE probe also allowed hemodynamic assessment and goal-directed therapy for optimal fluid filling. This therapy consisted of 250 mL crystalloid fluid challenges prolonged until obtainment of a stroke index steady state. The latter was defined as a condition with a stroke index increment <10% after the last fluid challenge. When this steady state was reached, if the MAP was lower than 65 mm Hg, boluses of ephedrine and/or norepinephrine were administered. If deemed necessary, a continuous infusion of the latter was delivered to maintain the MAP >65 mm Hg. If a low cardiac output, defined as a cardiac index <2.0 L min−1 m−2, was found through TEE assessment during and after separation from CPB, dobutamine was initiated at doses ranging from 5 to 20 μg kg−1 min−1. Discontinuation of catecholamine infusion occurred in the intensive care unit (ICU) and was left to the attending intensivist’s discretion. During surgery, CPB was performed using pulsatile perfusion with a frequency ranging from 60 to 70 beats min−1. When CPB was initiated, body temperature was decreased to 35°C to 36°C and then rewarmed to 36.5°C after completion of surgery and before separation from CPB. Fifteen minutes before the end of surgery, 0.2 mg kg−1 morphine, 100 mg ketoprofen, 1 g acetaminophen, and 0.3 mg kg−1 nefopam were delivered. Total IV anesthesia was discontinued at the last skin suture. Tracheal extubation on the operating table occurred if the patient was deemed eligible by the attending anesthesiologist according to specific cardiac extubation criteria.15 Otherwise, patients were transferred to the cardiac ICU where pressure support ventilation mode with light sedation was administered until extubation criteria were met.
This pilot study determined the success rate of cardiac anesthesia; successful cardiac anesthesia was defined as induction, maintenance, and emergence from anesthesia or until a patient’s transfer to the ICU requiring no attending anesthesiologist’s intervention. Further outcomes were clinical and controller anesthesia performances. Clinical performance of hypnosis is the ability to maintain BIS as close to the target of 45 as possible, which was defined using 4 categories: excellent, good, poor, and inadequate control, when the real BIS values are within 10%, from 11% to 20%, from 21% to 30%, or >30%, respectively, from the BIS target. The clinical performance of analgesia is defined as the ability to maintain the NociMap score as close as possible to 0. We defined 3 ranges of pain control. The range of −33 to +33 represents excellent pain control. The range starting from −33 to −66 and the range beginning from +33 to +66 represent good pain control. Finally, the range starting from −66 to −100 and the one commencing from +66 to +100 represent insufficient pain control. When the patient presented with either “vagal-type reactions” (HR reduction only) or “hypovolemia” (HR augmented only) refers to periods when no score was determined, and we have defined these events as “other.” When these events would occur, a minimal remifentanil continuous infusion 0.05 μg kg−1 min−1 was ensured.
The Varvel parameters were used to determine controller performance.16 These parameters are the performance error (PE), which is the difference between the real and the target values; the median PE, which is a measure of bias to describe whether the real values are either above or below the target values; the median absolute PE, which designates the size of errors; the wobble, which is a measure of the intraindividual variability in PE; and the divergence, which is the evolution of the controller’s performance through time. Finally, our tertiary outcomes consisted of evaluating the dose of anesthetic drug delivered, the number of infusion rate modifications, the number of patients extubated on the operating table, and time of BIS values <40, 35, 30, and >60. In addition, if not reported spontaneously, all patients were interrogated using a previously published questionnaire on postoperative day 1 to determine whether awareness occurred during the intervention.17
Data Collection and Statistical Analysis
Variables of interest, BIS, and hemodynamic variables were automatically recorded on a Excel datasheet every 5 seconds for the duration of anesthesia until tracheal extubation in the operating room or transfer to the ICU. Investigators also recorded the type of vasoconstrictors and catecholamines used and whether patients were tracheally extubated on the operating table. The clinical and controller performance analyses were completed including and excluding the CPB phase; these phases were defined as “with CPB” and “without CPB,” respectively. Sample size was chosen according to the range of 12 to 30 patients suggested for pilot studies.18 We decided to recruit 20 patients expecting a 20% dropout rate caused by loss of recorded signal or communication errors between the syringe pump and the controller. Data were collected for descriptive purposes. The Shapiro–Wilk test was used to test the distribution (all P > 0.17). Quantitative data are presented as mean (SD) or median (interquartile range) according to their distribution and categorical data as frequency (proportion). Confidence intervals (CIs) were also calculated for the outcome data using the Clopper–Pearson method.
Twenty patients were enrolled in the study after satisfactory simulation tests, showing good communication between the controllers and the syringe pumps and efficient application of the new cardiac anesthetic characteristic. Completely automated cardiac anesthesia was successfully performed in 80% (97.5% CI, 53%–95%) of the cases. Four miscommunications between the controller and the syringe pumps because of technical errors occurred during 4 different anesthesia cases causing a period of 15 minutes of manual overdrive necessary to restart the computer and the program and to enter patients’ data into the program. For statistical purposes concerning sample size calculation for further trials, and to avoid bias, these 4 patients were excluded from the final analysis. Patients’ anthropomorphic data, baseline comorbidities, surgery-related data, and the types of surgery patients underwent are described in Table 2. Control performance values of the robotic anesthesia system including and excluding the CPB phase are shown in Table 3. Figures 1 and 2 depict the control of depth of hypnosis and of analgesia, respectively. Overshoot and undershoot of the BIS target are described in Table 4. We recorded the closed-loop system making 106 (standard deviation [SD], 33) modifications per hour of the propofol infusion rate, and the average infusion rate administered was 94 (SD, 25) μg kg−1 min−1. During maintenance, 28 (SD, 9) modifications per hour of the remifentanil infusion rate occurred and the average infusion rate administered was 0.12 (SD, 0.06) μg kg−1 min−1. The total rocuronium dose was 1.21 (SD, 0.3) μg kg−1. Invalid BIS values, defined as a BIS value with concomitant signal quality index score lower than 40 and electromyography value higher than 40, were excluded from analysis. Invalid BIS values occurred in 3.5% (97.5% CI, 1%–7%) of the anesthesia time. Twelve patients received a continuous infusion of norepinephrine, and 3 patients received dobutamine. Three patients met the extubation criteria on the operating table. They were extubated on average 17 (SD, 1) minutes after discontinuation of the anesthetic drugs. None of the 20 patients enrolled reported explicit or implicit intraoperative recall when interrogated.
This study demonstrates that automated anesthesia delivery for cardiac surgery is feasible using a robotic anesthesia system administrating propofol, remifentanil, and rocuronium. The system demonstrates being capable of maintaining adequate hypnosis and analgesia. The closed-loop system maintained anesthesia for procedures in 80% (97.5% CI, 53%–95%) of cases without any human intervention. Four communication errors between the syringe pump and the controller occurred at the beginning of the trial after inappropriate manipulation of the automated system. These errors arose during an incorrect switch management of an empty syringe to a filled syringe. This problem was solved through a teaching session emphasizing the appropriate manipulation of the syringe pump. This session, addressed to the resident and registered anesthesiologist nurse in charge of the study patient, took place before the beginning of each study case.
Four categories of hypnosis performance were given considering that excellent control and inadequate control were the most clinically significant categories in these groups. Excellent control represents BIS values ranging from 41 to 49, whereas inadequate control encompasses BIS values either >58 or <32. These ranges are significant because they are associated with an imminently high risk of awareness or excessively profound hypnosis, respectively. Risk of awareness is central in cardiac surgery because it has been estimated to occur in 1.1% to 1.5% of cases during this type of procedure.19 In our trial, very short periods of BIS > 60 occurred, representing an overall period shorter than 2% of the entire anesthesia time. None of the 20 patients reported intraoperative awareness.
Only 1 investigation has been published using a semiautomated closed-loop system for propofol for patients scheduled to undergo cardiac surgery with CPB.20 In that study, the authors described the hypnosis performance of their system as the percentage of anesthesia duration when the BIS was within 40 and 60. Assuming that this range is equivalent to a BIS value within ±20% of our preset target, their results were within 20% of the predefined target for a length of time 10% longer than our study. This difference could not be attributed to a difference in an invalid BIS definition or length because their results were similar to ours. However, an important difference in our study was our study population, which was on average 32 years older. This characteristic is crucial because advancing age is a factor associated with “relative anesthetic overdose.” The latter could be defined as the cumulative time of low BIS values without increased anesthetic delivery.21 In addition, unlike Agarwal et al’s study,20 we did not exclude patients requiring epicardial pacing after removal of the aortic cross-clamp from our final analysis. Similar to electrocautery, epicardial pacing has been shown to possibly induce BIS electrical signal artifacts. However, newer versions of the BIS, like the one used in our trial, are more resistant to these artifacts.22 In contrast to Agarwal et al’s results,20 we did not find significant differences between the hypnosis performance including the CPB phase and the phase excluding CPB. It is likely that the absence of differences in our investigation is related to our CPB strategy that avoids moderate and deep hypothermia. The latter impacts the electroencephalogram decreasing BIS values.23
Conversely, we found a median PE of −9.9%, including CPB time. It indicates that the median BIS was below the target with a median BIS value of 36 demonstrating that the controller performance tends to overshoot. Such performance and trend are similar to the manual control found in our previous trials for noncardiac surgery.5,9 Compared with clinical performances found in the usual practice of noncardiac surgery, our findings were similar. In fact, the largest trial comparing the present robotic system with common practice found that excellent and good control of hypnosis was achieved for 69% (95% CI, 66%–72%) of the anesthesia time in the control group.9 In this study, excellent and good control of hypnosis with CPB represented 70% (97.5% CI, 63%–76%) of the time. In contrast, the same categories of hypnosis performances of the present automated system for noncardiac surgery were superior with 81% (95% CI, 78%–84%) of the time.9 It could be speculated that the reason for this discrepancy in the hypnosis performance of our robotic system was likely to have been a consequence of the older and more frail population recruited for this investigation compared with the study for noncardiac surgery. Indeed, in this study patients were 72 (7) years old, whereas patients were 54 (20) years9 old in the noncardiac study. In the current investigation, profound anesthesia, defined as a BIS score <30, occurred on average 13% (95% CI, 11%–5%) of the anesthesia time that represents only 32 minutes of the average anesthesia time. Minimal periods of very profound anesthesia are important because the cumulative duration of low BIS values during cardiac anesthesia has been associated with a relative risk for intermediate-term mortality of 1.29.24 There is no evidence of a direct causal association between low BIS and mortality. It has been hypothesized that relative anesthetic overdose leading to a lower BIS score may be secondary to “organ dysfunction” and/or “organ hypoperfusion.”21 Thus, it has been speculated that low BIS values associated with low MAP could be responsible for increased mortality.25,26 In that respect, we followed a protocol to optimize patients’ systemic perfusion through TEE-guided goal-directed fluid management. The latter strategy has 2 advantages. First, goal-directed therapy, although not as fully investigated as in noncardiac surgery, appears to reduce postoperative complications and hospital length of stay after cardiac surgical procedures.27 Second, this approach also allowed avoiding clinical scenarios where the NociMap algorithm runs at the minimal infusion rate because of hypovolemia or low cardiac output syndrome. Overall, the robotic anesthesia system offered satisfactory analgesia 70% (97.5% CI, 63%–76%) of the time. Although systemic blood perfusion optimization was performed, the condition defined as “other” occurred in 27% (97.5% CI, 21%–34%) of the anesthesia time. During this time, a minimal infusion rate of remifentanil was delivered. It is likely that this state defined as other arose because of hypovolemia, during epicardial pacing with low arterial blood pressure or constant pulsatile perfusion rate with low systemic resistance and/or low CPB flow perfusion rate.
Ideally, a closed-loop system for systemic blood perfusion optimization should be combined with robotic anesthesia delivery. An ultimate system such as this should not only automatically administer fluid therapy, because it has already been successful for major noncardiac surgery,28 but should also automatically deliver vasoconstrictors and inotropic agents when low systemic vascular resistance and low cardiac output syndrome are present. A combined optimal robotic anesthesia and perfusion system for cardiac anesthesia that would reduce attending anesthesiologists’ workload, allowing them to concentrate on a higher level of care requiring human intelligence, will make a difference in cardiac patients’ outcomes.10 Closed-loop systems designed for cardiac anesthesia following expert-based principles can help standardize cardiac patients’ care. This is important, as clearly demonstrated in a recent investigation, showing that expert cardiac anesthesiologists can decrease the incidence of postoperative complications and mortality.29
There are several limitations to our study. Because this was a pilot study, the sample size of our investigation was not large enough to speculate on the efficacy of the robotic system to reduce the incidence of awareness compared with manual administration. In addition, our preliminary results cannot be generalized to all cardiac surgery. Although the population enrolled included different cardiac procedures, the system was tested only with mild hypothermia. Hence, the feasibility of the deep hypothermic approach needs to be tested.
Completely automated anesthesia for cardiac surgery is feasible. The system we tested may play a crucial role in increasing patients’ safety, standardizing anesthesia delivery on expert-based principles, and freeing anesthesiologists to focus on higher level tasks that require human intervention. Further trials should be conducted in cardiac patients comparing the present robotic system with manual control administration, as well as trials testing the feasibility and safety of this automatic anesthesia system combined with a closed loop for “blood perfusion optimization.”
Name: Cedrick Zaouter, MD, MSc.
Contribution: This author helped design the study, conduct the study, analyze the data, and write the manuscript.
Name: Thomas M. Hemmerling, MD.
Contribution: This author helped design the study, analyze the data, and write the manuscript.
Name: Romain Lanchon, MD.
Contribution: This author helped conduct the study and analyze the data.
Name: Emanuela Valoti, MD.
Contribution: This author helped conduct the study, analyze the data, and write the manuscript.
Name: Alain Remy, MD.
Contribution: This author helped conduct the study.
Name: Sébastien Leuillet, MSc.
Contribution: This author helped analyze the data and write the manuscript.
Name: Alexandre Ouattara, MD, PhD.
Contribution: This author helped design the study, analyze the data, and write the manuscript.
This manuscript was handled by: Maxime Cannesson, MD, PhD.
Philips Europe™ supplied the monitor device intelliVue MX 700 for measurements during the study period.
1. Hemmerling TM, Terrasini N. Robotic anesthesia: not the realm of science fiction any more. Curr Opin Anaesthesiol 2012;25:73642.
2. Dumont GA, Ansermino JM. Closed-loop control of anesthesia: a primer for anesthesiologists. Anesth Analg 2013;117:11308.
3. Atchabahian A, Hemmerling TM. Robotic anesthesia: how is it going to change our practice? Anesth Pain Med 2014;4:e16468.
4. Hemmerling TM. Miller R. Administration of anesthesia by robots. Miller’s Anesthesia, 2015:8th ed. Philadelphia, PA:Elsevier Health Sciences, 2550.
5. Hemmerling TM, Charabati S, Zaouter C, Minardi C, Mathieu PA. A randomized controlled trial demonstrates that a novel closed-loop propofol system performs better hypnosis control than manual administration. Can J Anaesth 2010;57:72535.
6. Liu N, Chazot T, Genty A, Landais A, Restoux A, McGee K, Laloë PA, Trillat B, Barvais L, Fischler M. Titration of propofol for anesthetic induction and maintenance guided by the bispectral index: closed-loop versus manual control: a prospective, randomized, multicenter study. Anesthesiology 2006;104:68695.
7. Liu N, Chazot T, Hamada S, Landais A, Boichut N, Dussaussoy C, Trillat B, Beydon L, Samain E, Sessler DI, Fischler M. Closed-loop coadministration of propofol and remifentanil guided by bispectral index: a randomized multicenter study. Anesth Analg 2011;112:54657.
8. Puri GD, Kumar B, Aveek J. Closed-loop anaesthesia delivery system (CLADS) using bispectral index: a performance assessment study. Anaesth Intensive Care 2007;35:35762.
9. Hemmerling TM, Arbeid E, Wehbe M, Cyr S, Taddei R, Zaouter C. Evaluation of a novel closed-loop total intravenous anaesthesia drug delivery system: a randomized controlled trial. Br J Anaesth 2013;110:10319.
10. Wahr JA, Prager RL, Abernathy JH III, Martinez EA, Salas E, Seifert PC, Groom RC, Spiess BD, Searles BE, Sundt TM III, Sanchez JA, Shappell SA, Culig MH, Lazzara EH, Fitzgerald DC, Thourani VH, Eghtesady P, Ikonomidis JS, England MR, Sellke FW, Nussmeier NA; American Heart Association Council on Cardiovascular Surgery and Anesthesia, Council on Cardiovascular and Stroke Nursing, and Council on Quality of Care and Outcomes Research. Patient safety in the cardiac operating room: human factors and teamwork: a scientific statement from the American Heart Association. Circulation 2013;128:113969.
11. Braz LG, Braz DG, Cruz DS, Fernandes LA, Módolo NS, Braz JR. Mortality in anesthesia: a systematic review. Clinics 2009;64:9991006.
12. Hemmerling TM, Charabati S, Salhab E, Bracco D, Mathieu PA. The AnalgoscoreTM: a novel score to monitor intraoperative nociception and its use for closed loop application of remifentanil. J Comput 2009;4:3118.
13. Wehbe M, Mathieu PA, Hemmerling TM. Relaxofon: a neuromuscular blockade monitor for patients under general anesthesia. Conf Proc IEEE Eng Med Biol Soc 2012;2012:1503.
14. Wehbe M, Arbeid E, Cyr S, Mathieu PA, Taddei R, Morse J, Hemmerling TM. A technical description of a novel pharmacological anesthesia robot. J Clin Monit Comput 2014;28:2734.
15. Zaouter C, Imbault J, Labrousse L, Abdelmoumen Y, Coiffic A, Colonna G, Jansens JL, Ouattara A. Association of robotic totally endoscopic coronary artery bypass graft surgery associated with a preliminary cardiac enhanced recovery after surgery program: a retrospective analysis. J Cardiothorac Vasc Anesth 2015;29:148997.
16. Varvel JR, Donoho DL, Shafer SL. Measuring the predictive performance of computer-controlled infusion pumps. J Pharmacokinet Biopharm 1992;20:6394.
17. Errando CL, Sigl JC, Robles M, Calabuig E, García J, Arocas F, Higueras R, Del Rosario E, López D, Peiró CM, Soriano JL, Chaves S, Gil F, García-Aguado R. Awareness with recall during general anaesthesia: a prospective observational evaluation of 4001 patients. Br J Anaesth 2008;101:17885.
18. Hemmerling TM, Arbeid E, Wehbe M, Cyr S, Giunta F, Zaouter C. Transcontinental anaesthesia: a pilot study. Br J Anaesth 2013;110:75863.
19. Ghoneim MM, Block RI. Learning and memory during general anesthesia: an update. Anesthesiology 1997;87:387410.
20. 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;53:3907.
21. Leslie K, Short TG. Low bispectral index values and death: the unresolved causality dilemma. Anesth Analg 2011;113:6603.
22. Dahaba AA. Different conditions that could result in the bispectral index indicating an incorrect hypnotic state. Anesth Analg 2005;101:76573.
23. Kertai MD, Whitlock EL, Avidan MS. Brain monitoring with electroencephalography and the electroencephalogram-derived bispectral index during cardiac surgery. Anesth Analg 2012;114:53346.
24. Kertai MD, Pal N, Palanca BJ, Lin N, Searleman SA, Zhang L, Burnside BA, Finkel KJ, Avidan MS; B-Unaware Study Group. Association of perioperative risk factors and cumulative duration of low bispectral index with intermediate-term mortality after cardiac surgery in the B-Unaware Trial. Anesthesiology 2010;112:111627.
25. Sessler DI, Sigl JC, Kelley SD, Chamoun NG, Manberg PJ, Saager L, Kurz A, Greenwald S. Hospital stay and mortality are increased in patients having a “triple low” of low blood pressure, low bispectral index, and low minimum alveolar concentration of volatile anesthesia. Anesthesiology 2012;116:1195203.
26. Leslie K, Short TG. Sedation depth and mortality: a large randomized trial is required. Anesth Analg 2014;118:9035.
27. Aya HD, Cecconi M, Hamilton M, Rhodes A. Goal-directed therapy in cardiac surgery: a systematic review and meta-analysis. Br J Anaesth 2013;110:5107.
28. Rinehart J, Lilot M, Lee C, Joosten A, Huynh T, Canales C, Imagawa D, Demirjian A, Cannesson M. Closed-loop assisted versus manual goal-directed fluid therapy during high-risk abdominal surgery: a case-control study with propensity matching. Crit Care 2015;19:94.
29. Glance LG, Kellermann AL, Hannan EL, Fleisher LA, Eaton MP, Dutton RP, Lustik SJ, Li Y, Dick AW. The impact of anesthesiologists on coronary artery bypass graft surgery outcomes. Anesth Analg 2015;120:52633.