Electroencephalography-Guided Anesthesia and Delirium in Older Adults After Cardiac Surgery

Key Points

Question

Does electroencephalography-guided anesthetic administration to minimize electroencephalography suppression decrease the incidence of postoperative delirium in older patients undergoing cardiac surgery?

Findings

In this multicenter randomized clinical trial involving 1140 patients aged 60 years and older undergoing cardiac surgery, postoperative delirium occurred in 18.15% of participants in the electroencephalography-guided anesthetic group and 18.10% in the usual care group, a difference that was not statistically significant.

Meaning

These findings do not support the use of electroencephalography-guided anesthetic administration for the prevention of postoperative delirium among older adults undergoing cardiac surgery.

Abstract

Importance

Intraoperative electroencephalogram (EEG) waveform suppression, suggesting excessive general anesthesia, has been associated with postoperative delirium.

Objective

To assess whether EEG-guided anesthesia decreases the incidence of delirium after cardiac surgery.

Design, Setting, and Participants

Randomized, parallel-group clinical trial of 1140 adults 60 years or older undergoing cardiac surgery at 4 Canadian hospitals. Recruitment was from December 2016 to February 2022, with follow-up until February 2023.

Interventions

Patients were randomized in a 1:1 ratio (stratified by hospital) to receive EEG-guided anesthesia (n = 567) or usual care (n = 573). Patients and those assessing outcomes were blinded to group assignment.

Main Outcomes and Measures

The primary outcome was delirium during postoperative days 1 through 5. Intraoperative measures included anesthetic concentration and EEG suppression time. Secondary outcomes included intensive care and hospital length of stay. Serious adverse events included intraoperative awareness, medical complications, and 30-day mortality.

Results

Of 1140 randomized patients (median [IQR] age, 70 [65-75] years; 282 [24.7%] women), 1131 (99.2%) were assessed for the primary outcome. Delirium during postoperative days 1 to 5 occurred in 102 of 562 patients (18.15%) in the EEG-guided group and 103 of 569 patients (18.10%) in the usual care group (difference, 0.05% [95% CI, −4.57% to 4.67%]). In the EEG-guided group compared with the usual care group, the median volatile anesthetic minimum alveolar concentration was 0.14 (95% CI, 0.15 to 0.13) lower (0.66 vs 0.80) and there was a 7.7-minute (95% CI, 10.6 to 4.7) decrease in the median total time spent with EEG suppression (4.0 vs 11.7 min). There were no significant differences between groups in median length of intensive care unit (difference, 0 days [95% CI, −0.31 to 0.31]) or hospital stay (difference, 0 days [95% CI, −0.94 to 0.94]). No patients reported intraoperative awareness. Medical complications occurred in 64 of 567 patients (11.3%) in the EEG-guided group and 73 of 573 (12.7%) in the usual care group. Thirty-day mortality occurred in 8 of 567 patients (1.4%) in the EEG-guided group and 13 of 573 (2.3%) in the usual care group.

Conclusions and Relevance

Among older adults undergoing cardiac surgery, EEG-guided anesthetic administration to minimize EEG suppression, compared with usual care, did not decrease the incidence of postoperative delirium. This finding does not support EEG-guided anesthesia for this indication.

Trial Registration

ClinicalTrials.gov Identifier: NCT02692300

This randomized clinical trial examines whether electroencephalogram -guided anesthesia decreases the incidence of delirium after cardiac surgery among adults 60 years or older.

Introduction

Postoperative delirium is associated with poor outcomes, including cognitive decline.¹ Delirium is common in patients older than 60 years after cardiac surgery,² with preventive attempts being largely unsuccessful.^3,4 Avoidance of deep general anesthesia has been advocated for preventing postoperative delirium.^5,6 Electroencephalographic (EEG) suppression is an accepted biomarker of deep anesthesia.⁷ In an observational cohort study conducted in 727 patients undergoing major cardiac and thoracic surgery, every 5 minutes of EEG suppression was associated with a 5% increased adjusted odds of postoperative delirium.⁸ The Electroencephalographic Guidance of Anesthesia to Alleviate Geriatric Syndromes (ENGAGES) trial did not find a decrease in delirium incidence with EEG guidance to minimize volatile anesthetic exposure and EEG suppression in older adults undergoing high-risk surgical procedures.⁹ However, a mediation analysis from that trial found that EEG suppression might still contribute to delirium.¹⁰ Other trials have found that deep anesthesia was associated with delirium.^6,11 The incongruent findings underline the need for further research to clarify whether EEG guidance of anesthesia can prevent postoperative delirium and what patient populations might benefit.

Most of the evidence regarding anesthetic depth and postoperative delirium has been obtained from patients undergoing noncardiac surgical procedures.^5,6 In a trial of 310 patients undergoing cardiac surgery, EEG-guided anesthesia was associated with a nonsignificant absolute reduction in postoperative delirium incidence of 9%.¹² The purpose of the ENGAGES-Canada trial was to evaluate EEG-guided anesthesia in older patients undergoing cardiac surgery.¹³ The hypothesis was that minimizing intraoperative EEG-suppression would decrease incident postoperative delirium and other complications.¹³

Methods

Design, Setting, and Ethics

The multicenter, pragmatic, evaluator- and patient-blinded randomized clinical trial was conducted at 4 Canadian tertiary hospitals and was funded by the Canadian Institutes of Health Research (PJT-159482). A protocol for the trial has been published and is available in Supplement 1.¹³ An independent data and safety monitoring board oversaw the trial, which was registered with ClinicalTrials.gov (NCT02692300). Written informed consent was obtained from every patient. Randomization was stratified by site, and patients were randomized in a 1:1 ratio to undergo usual care or to EEG-guided anesthesia. Random number sequences and group assignments were computer generated and sealed in opaque envelopes that were opened in the operating room on the day of the surgical procedure. Patients, family members, and research team members assessing delirium remained blinded to the group assignment. Chart reviews were conducted by individuals blinded to the group assignment and to delirium assessment outcomes. The statistician was not blinded to group assignment while performing the analyses. The research ethics boards of the Montreal Heart Institute (2017-2164), the University of Manitoba (HS18290), Queen’s University in Kingston (ANAE-298-16), and the University of Toronto (17-5933) approved the study. The trial was conducted per Good Clinical Practice and the Declaration of Helsinki.¹⁴ The 2010 Consolidated Standards of Reporting Trials guidelines for reporting parallel-group randomized trials¹⁵ were followed when reporting results ( Supplement 2).

Trial Population

Patients aged 60 years and older scheduled to undergo a cardiac surgical procedure with cardiopulmonary bypass were eligible for inclusion. Patients were excluded if they had preoperative delirium, hearing impairment, were blind or unable to read or write French or English, did not speak French or English, or if they had previously experienced intraoperative awareness. The first patient was recruited on December 28, 2016, and the last patient was recruited on February 18, 2022. Preoperative assessments included demographics, medical history, quality of life, mental health, cognitive function, and history of falls. Patients self-identified their race, ethnicity, and sex in a questionnaire. Race and ethnicity were included to examine similarities or differences between race or ethnicity groups in terms of risk for postoperative delirium incidence, and self-identification was based on a fixed categories.

Trial Interventions

Anesthetic agents were chosen at the discretion of EEG-trained anesthesiologists.^16,17 A frontal electroencephalogram sensor was applied to the forehead of each patient. In the usual care group, clinicians were blinded to all EEG waveforms and derived values except the signal quality index. Administering general anesthesia without EEG monitoring was consistent with usual care practice at all 4 participating hospitals. The EEG waveforms and derived parameters (suppression ratio, spectral edge frequency, electromyography, signal quality index, and proprietary index) were displayed in the EEG-guided group. In this group, clinicians were encouraged to decrease volatile anesthetic administration to minimize EEG suppression.

Outcomes

Definitions of the outcomes and instruments used for their detection are provided in the published protocol.¹³ The prespecified primary outcome was delirium incidence during postoperative days 1 to 5. Delirium was assessed daily in the afternoon or evening. Trained research personnel evaluated patients with a brief cognitive screen^18,19 followed by the Confusion Assessment Method (CAM)²⁰ rating or, if intubated and unable to speak or refused the full CAM, the CAM for the Intensive Care Unit (CAM-ICU).²¹ A research team member completed a structured chart review using a validated approach^22,23 for evidence of delirium during postoperative days 1 to 5. An independent expert panel adjudicated uncertain delirium assessments. A patient diagnosed with delirium on any 5 postoperative days using any detection method (CAM, CAM-ICU, or chart review) was considered to have the primary outcome of incident delirium.¹³ Similarly, a comparison of time to delirium onset between groups was also planned, as was covariate adjustment incorporating preoperative characteristics. Prespecified secondary outcomes included ICU and hospital length of stay.¹³ Prespecified exploratory delirium-related outcomes were duration and severity of delirium (CAM-S long-form measure or CAM-ICU-7 delirium severity instrument) and delirium incidence on the day of the surgical procedure (immediate postoperative period). The subsyndromal delirium incidence^24,25 (having a positive feature in the CAM or CAM-ICU but not meeting diagnostic criteria for delirium) was a nonprespecified exploratory outcome.

Other prespecified exploratory outcomes included fall incidence, depressive symptoms (using the Patient Health Questionnaire-4 [range, 0-12; score of 3-12 indicates increasing severity of depression]), alcohol use (using the Alcohol Use Disorders Identification Test-Concise [range, 0-12; score ≥3 for women and ≥4 for men indicates alcohol use disorder), posttraumatic stress disorder symptoms (Posttraumatic Stress Disorder Checklist for DSM-5 [range, 0-80; score >33 indicates risk of PTSD]), cognitive abilities (using the Short Blessed Test [range, 0-28; score ≥10 consistent with dementia]), and quality of life at postoperative day 30.¹³ Prespecified adverse events and serious adverse events included 30-day and 1-year postoperative mortality, undesirable intraoperative movement, intraoperative awareness, postoperative nausea and vomiting, and complications such as major blood loss and transfusions, stroke, sternal wound infection, sepsis, need for dialysis, and prolonged intubation.¹³ Several sensitivity analyses regarding the primary outcome were planned.

Sample Size Calculation

A statistical analysis plan is provided in Supplement 3. The incidence of postoperative delirium after cardiac surgery is estimated to be 25% to 50%.^26,27 A prior trial of 921 older adults undergoing noncardiac surgical procedures found that EEG-guided anesthesia was associated with an 8% decrease in postoperative delirium incidence.¹¹ A trial in 310 patients undergoing cardiac surgery found that EEG guidance was associated with a nonstatistically significant reduction of 9% in delirium incidence.¹² These trials and a meta-analysis¹² lent support for a large trial to provide stronger evidence to the postulated effectiveness of EEG guidance in preventing delirium in cardiac surgery and other high-risk patients.¹² Assuming an incidence of delirium in the usual care group of 25%, to detect a clinically meaningful difference of 8%^11,12,28 in the delirium incidence between the 2 groups with α = .05, 90% power, and 2-sided Fisher exact test, a total sample of 1132 patients was required; the recruitment target sample was set at 1200 patients.

Statistical Methods

Statistical Analyses for the Primary Outcome

Patients were assessed based on their randomization group. For the primary outcome, the proportions of patients with incident postoperative delirium were compared between the 2 groups using the χ² test. We constructed Kaplan-Meier curves for each group to compare time to delirium onset and conducted a log-rank test. We conducted covariate adjustment including prespecified patient characteristics (age, sex, alcohol use, number of comorbidities, Euroscore II, history of delirium, history of depression, preoperative cognitive impairment [Eight-item Informant Interview to Differentiate Aging and Dementia baseline score {range, 0-8; score ≥2 indicates cognitive impairment likely}], and history of falls during the past 6 months) and group assignment using 2 methods: (1) logistic regression and (2) standardized estimator combined with bootstrapped 95% CIs.

Statistical Analyses for Secondary and Exploratory Outcomes and Adverse Events

The secondary outcomes of median length of ICU and hospital stay were compared between the treatment groups using the Wilcoxon rank sum test with a 95% CI. For the prespecified delirium-related exploratory outcomes, we compared the duration of delirium between groups with the Wilcoxon rank sum test with a 95% CI and the difference between groups in severe delirium and delirium incidence on the day of the surgical procedure with the χ² test. To assess risk factors for delirium, a Delphi process^27,29,30 was conducted among investigators to select 14 candidate perioperative variables (trial group assignment, age, sex, alcohol use, number of comorbidities, Euroscore II, history of delirium, history of depression, cognitive impairment, falls within 6 months, valvular surgery, duration of general anesthesia, time with EEG suppression, median inhaled anesthetic concentration). We used a logistic regression analysis to evaluate variables associated with postoperative delirium incidence. Study site was included post hoc in the model. For the nonprespecified exploratory outcome of subsyndromal delirium, proportions of patients were compared between groups with the χ² test. We compared postoperative 30-day patient-reported outcomes measures (survey instruments for falls, depressive symptoms, posttraumatic stress disorder symptoms, cognitive abilities, and quality of life) between the groups. χ² or Fisher exact tests were used to compare discrete data. Mann-Whitney U or unpaired t tests were used to compare continuous data, depending on their distributions. We also conducted logistic regression analyses with prespecified variables to evaluate candidate associations with 30-day and 1-year falls and 1-year mortality. Mortality was compared between groups at 30 days and 1 year using Cox proportional hazard and Kaplan-Meier plots. Other adverse and serious adverse events were compared between groups using the χ² test. No planned sensitivity analyses or imputations were based on missing values for secondary or other outcomes or adverse events.

Sensitivity Analyses

The 3 primary end point analyses excluded patients who could not be assessed for delirium. In 2 sensitivity analyses, these patients were all assumed to have had incident delirium or not. Two prespecified additional sensitivity analyses were conducted: 25% of cases within the guided group, stratified by site, were excluded with the longest cumulative EEG suppression time and highest median volatile anesthetic concentrations. We conducted an additional sensitivity analysis in which groups were based on the treatment received. For these sensitivity analyses, delirium incidence between groups was compared with the χ² test.

Continuous variables are presented as mean (SD) or median (IQR), depending on their distributions and discrete variables are presented as number (percent). In addition to analyses for outcomes, we also performed statistical analyses for comparisons between groups of perioperative measures (EEG suppression times, median volatile anesthetic concentrations, and median blood pressure), intraoperative medications, surgical procedure type, and number of surgical procedures. Unpaired t or Mann-Whitney U tests were used to compare continuous data, depending on their distributions. χ² or Fisher exact tests were used for comparisons of discrete data. CIs for median differences were calculated using Hodges-Lehmann estimates, and Newcombe method with continuity correction was used for differences between proportions. All results are presented with 95% CIs. All significance testing was 2-sided, with P values <.05 considered statistically significant and P values <.005 as more compelling evidence. The statistical analyses were performed with SAS version 9.4 (SAS Institute) and STATA version 17.0 SE (StataCorp LP).

Results

During the study period, 1225 patients were enrolled and 1140 were randomized. Of these patients, 573 were randomized to receive usual anesthesia care and 567 to receive EEG-guided care. The primary outcome analysis was performed on 569 patients in the usual care group and 562 patients in the EEG-guided group (Figure 1; eTables 1-3 in Supplement 4). There were no meaningful differences in patient baseline characteristics between groups or by study site (Table 1; eTables 4-8 in Supplement 4). There were no statistically significant differences between groups in intraoperative administration of intravenous hypnotic anesthetics, amnesic agents, opioids, neuromuscular blocking drugs, or other perioperative medications (Table 2; eTable 9 in Supplement 4). There were no meaningful differences between groups in median duration of the surgical procedure, anesthesia, mechanical ventilation, cardiopulmonary bypass, and aortic cross-clamp. There were also no significant differences between groups in median blood pressure or other perioperative measures (Table 2). In the EEG-guided group, there was a decrease of 0.14 (95% CI, −0.15 to −0.13) in the median minimum alveolar concentration, from 0.80 to 0.66, and a 7.7-minute (95% CI, −10.6 to −4.7) decrease in the median total time spent with EEG suppression (from 11.7 to 4.0 minutes) (Table 2; eFigure 1 in Supplement 4). There were meaningful differences among study sites in median volatile anesthetic administration, duration of EEG suppression, and other perioperative measures (eTables 10 and 11 in Supplement 4).

Figure 1. Flow of Participants in a Trial of Electroencephalography (EEG)-Guided Anesthesia in Cardiac Surgery.

^aRandomization was stratified by site and patients were randomized in a 1:1 ratio to receive usual care or to EEG-guided anesthesia.

^bThese deviations are due to EEG malfunctioning, “off-pump” coronary artery bypass graft, patient taken back to the operating room on postoperative day 1, and use of circulatory arrest.

^cDeviation of anesthesia protocol includes EEG guidance in a patient randomized to the blinded protocol or no EEG in a patient randomized to the guided protocol.

Table 1. Preoperative Patient Characteristics.

Characteristics	No. (%)
	EEG guided (n = 567)	Usual care (n = 573)
Demographics
Age, median (IQR), y	70.0 (65.0-75.0)	71.0 (67.0-76.0)
Sex
Male	443 (78.1)	415 (72.4)
Female	124 (21.9)	158 (27.6)
Racea
American Indian or Alaska Native	5 (0.9)	5 (0.9)
Asian	2 (0.4)	4 (0.7)
Black or African American	2 (0.4)	3 (0.5)
Native Hawaiian or Other Pacific Islander	0	2 (0.3)
White	379 (66.8)	373 (65.1)
Other	6 (1.1)	6 (1.0)
Not reported	173 (30.5)	180 (31.4)
Attended college	253 (44.6)	228 (39.8)
Residency
Montreal	232 (40.9)	233 (40.7)
Kingston	173 (30.5)	175 (30.5)
Winnipeg	158 (27.9)	160 (27.9)
Toronto	4 (0.7)	5 (0.9)
Living alone	115 (20.3)	139 (24.3)
Wearing hearing aid	72 (12.7)	69 (12.0)
Wearing glasses/contact lenses	452 (79.7)	459 (80.1)
Daily alcohol use ≥5 drinks	76 (13.4)	71 (12.4)
Body mass index, median (IQR)	28.4 (25.3-32.5)	28.7 (25.6-32.1)
Medical history
Comorbidities
Hypertension	441 (77.8)	430 (75.0)
Coronary artery disease	383 (67.6)	392 (68.4)
Diabetes	200 (35.3)	208 (36.3)
Myocardial infarction	160 (28.2)	173 (30.2)
Atrial fibrillation	153 (27.0)	135 (23.6)
History of falls in the last 6 mob	108 (19.0)	100 (17.4)
Congestive heart failure	96 (16.9)	102 (17.8)
History of obstructive sleep apnea	80 (14.1)	74 (12.9)
Chronic obstructive pulmonary disease	65 (11.5)	55 (9.6)
Cerebrovascular disease	51 (9.0)	50 (8.7)
Solid tumor	51 (9.0)	51 (8.9)
Chronic kidney disease	44 (7.8)	61 (10.6)
History of deliriumc	34 (6.0)	43 (7.5)
Transient ischemic attack	28 (4.9)	25 (4.4)
Left ventricular ejection fraction <40%	25 (4.4)	24 (4.2)
Stroke	22 (3.9)	25 (4.4)
History of major depression	21 (3.7)	23 (4.0)
Liver disease	4 (0.7)	7 (1.2)
No. of comorbidities, median (IQR)	3.0 (3.0-4.0)	4.0 (3.0-5.0)
Physical status
ASA physical status ≥4d	527 (92.9)	524 (91.4)
EuroSCORE II, median (IQR)e	1.4 (1.0-2.5)	1.6 (1.1-2.8)
NYHA functional classification ≥3f	56 (9.9)	52 (9.1)
Preoperative assessments
AUDIT-C score ≥4 for males or ≥3 for femalesg	180 (31.7)	176 (30.7)
STOP-BANG score ≥3h	69 (12.2)	67 (11.7)
AD8 score ≥2i	54 (9.5)	48 (8.4)
PHQ-4 score ≥6j	44 (7.7)	41 (7.2)
SBT score ≥10k	27 (4.8)	15 (2.6)
PCL-5 score ≥33l	5 (0.9)	3 (0.5)
Cognitive Abilities T-scores, mean (SD)m	58.1 (7.75)	58.6 (7.25)
Cognitive Concerns T-scores, mean (SD)m	29.2 (5.28)	28.8 (4.85)
PROMIS Global-10 T-scores, mean (SD)m
Physical health–related quality of life	45.7 (7.47)	45.9 (7.56)
Mental health–related quality of lifen	51.1 (7.03)	52.0 (6.74)

Abbreviation: EEG, electroencephalography.

^a Participants were asked to identify their race from a list of categories. “Other” was listed as an option. Multiple options were allowed to be chosen. Three patients selected 2 options each, with one being “Unknown/Not reported;” for simplicity, the latter was omitted because a specific race was also reported.

^b History of accidental falls was obtained by questionnaire and defined as necessitating medical evaluation or medical treatment.

^c History of delirium was defined as the patient reporting having postoperative delirium after a previous surgery.

^d American Society of Anesthesiologists (ASA) physical status scores range from 1 to 6. A score of 1 indicates healthy; 2, mild systemic disease; 3, severe systemic disease; 4, severe systemic disease that is a constant threat to life; 5, not expected to survive without procedure; and 6, neurologically deceased organ donor. In this study, the ASA was dichotomized with a threshold set at 4 or higher, indicating severe systemic disease that is a constant threat to life.

^e European System for Cardiac Operative Risk Evaluation 2 (EuroSCORE II) is used to predict the likely risk of in-hospital mortality after cardiac surgery. Scores are categorized as low-risk patients (≤2 points) with a predicted mortality less than 1%, patients at moderate risk (3-5 points) with mortality around 3%, and a high-risk group (≥5 points) with predicted mortality of 10%-11%.

^f New York Heart Association (NYHA) functional classification of heart failure ranges from 1 to 4. Class I indicates no symptoms, no limitation of physical activity; II, mild symptoms, slight limitation during ordinary activity; III, marked limitation even during less than ordinary activity; and IV, severe limitations even while at rest.

^g Alcohol Use Disorders Identification Test-Concise (AUDIT-C) score ranges 0 to 12. A score of 0 indicates no alcohol use; a score of 4 or higher is considered positive for men, while a score of 3 or higher is considered positive for women.

^h STOP-BANG (snoring history, tired during the day, observed apnea while sleep, high blood pressure, body mass index more than 35, age more than 50 years, neck circumference more than 40 cm, and male gender) score screens for obstructive sleep apnea (OSA). Range, 0 to 8; a score of 2 or less is considered low risk, 3 or 4 is considered likely to have OSA, and a score of 5 or more is considered high risk for having either moderate or severe OSA.

ⁱ Ascertain Dementia 8-Item questionnaire (Ad8) is used to detect early signs of cognitive changes associated with dementia. Range 0 to 8; a score of 0 or 1 indicates unimpaired cognition and a score of 2 or greater indicates that cognitive impairment is likely to be present.

^j Patient Health Questionnaire-4 (PHQ-4) screens for depression symptoms. Range, 0 to 12; a score of 2 or less indicates no symptoms of depression; 3 to 5, mild depression; 6 to 8, moderate depression; and 9 or above, severe depression.

^k Short Blessed Test (SBT) screens for early cognitive changes associated with Alzheimer disease. Range, 0 to 28; a score of 0-4 indicates unimpaired cognition; 5-9, questionable impairment; 10 or more, impairment consistent with dementia.

^l Posttraumatic Stress Disorder Checklist for DSM-5 (PCL-5) is used to assess DSM-5 symptoms for posttraumatic stress disorder (PTSD). Range, 0 to 80; a total score lower than 33 may indicate the patient does not meet criteria for PTSD and a score of 33 or higher suggests the patient needs further assessment to confirm a diagnosis of PTSD.

^m Cognitive Abilities and Cognitive Concerns are 8-item Patient Reported Outcomes Measurement Information System (PROMIS) instruments each used to assess cognitive impairment. The PROMIS Global-10 is a 10-item questionnaire used to assess the health care–related quality of life (including social, mental, and physical) measures for the general population. The total raw score of each of these instruments is converted to a T-score metric, using item-level calibrations. A mean T-score of 50 with an SD 10 is set as the threshold. Higher T-scores (>50) indicate better perceived levels.

ⁿ See eTable 4-6 in Supplement 4 for additional baseline characteristics.

Table 2. Perioperative Measures, Intraoperative Medications, and Types of Surgical Procedures.

Measures	EEG guided (n = 567)	Usual care (n = 573)	Difference (95% CI)a
Perioperative measures, median (IQR)
Cumulative time with EEG suppression, min	4.04 (0.66 to 16.71)	11.74 (2.15 to 36.89)	−7.70 (−10.61 to −4.67)
Cumulative time with EEG suppression ratio >1%, minb	25.0 (6.0 to 71.0)	60.0 (15.0 to 123.0)	−35.0 (−45.0 to −25.0)
Volatile anesthetic concentration (MAC)c	0.66 (0.45 to 0.90)	0.80 (0.59 to 0.95)	−0.14 (−0.15 to −0.13)
Age-adjusted volatile anesthetic concentrationd	0.79 (0.54 to 1.04)	0.93 (0.70 to 1.16)	−0.13 (−0.13 to −0.12)
Mean arterial pressure, mm Hg	68.0 (60.0 to 76.0)	67.0 (60.0 to 75.0)	1.0 (0.97 to 1.04)
Duration of anesthesia, min	275.5 (230.0 to 343.0)	268.5 (225.0 to 343.0)	7.0 (−6.2 to 18.2)
Duration of operation, min	205.0 (162.0 to 259.0)	199.0 (158.0 to 258.0)	6.0 (−5.4 to 17.4)
Duration of cardiopulmonary bypass, min	91.5 (66.0 to 126.0)	86.5 (64.0 to 125.0)	4.5 (−2.5 to10.5)
Duration of aortic cross-clamping, min	70.0 (50.0 to 99.0)	68.0 (46.0 to 96.0)	2.0 (−3.5 to 7.5)
Estimated blood loss, mL	350.0 (300.0 to 550.0)	350.0 (300.0 to 500.0)	0 (−28.5 to 28.5)
Duration of mechanical ventilation, min	347.0 (204.0 to 545.0)	352.0 (190.0 to 600.0)	−5.0 (−44.4 to 34.4)
Intraoperative medications, median (IQR)
Morphine equivalents, mge	50.0 (39.5 to 63.0)	50.0 (40.0 to 63.0)	0 (−1.7 to 1.7)
Midazolam, mg	3.0 (2.0 to 4.0)	2.0 (2.0 to 4.0)	1.0 (0.45 to 1.55)
Propofol, mg	466.0 (242.0 to 677.7)	500.0 (210.0 to 714.0)	−34.0 (−83.0 to 15.0)
Rocuronium, mg	100.0 (90.0 to 100.0)	100.0 (90.0 to 100.0)	0 (0 to 0)
Type of surgical procedure, No. (%)f
On-pump coronary artery bypass graft	399 (70.4)	394 (68.8)	1.6% (−3.7 to 7.1)
Aortic valve replacement	191 (33.7)	202 (35.2)	−1.6% (−7.2 to 4.1)
Other surgical procedure	67 (11.8)	53 (9.2)	2.6% (−1.1 to 6.3)
Mitral valve repair	43 (7.6)	34 (5.9)	1.6% (−1.4 to 4.7)
Mitral valve replacement	37 (6.5)	35 (6.1)	0.4% (−2.6 to 3.4)
Maze procedure	12 (2.1)	7 (1.2)	0.9% (−0.8 to 2.7)
Off-pump coronary artery bypass graft	3 (0.5)	4 (0.7)	−0.2% (−1.4 to 1.1)
Surgical procedures, No. (%)
1	405 (71.4)	428 (74.7)	−3.3% (−8.5 to 2.0)
≥2	162 (28.6)	145 (25.3)	3.3% (−2.0 to 8.5)

^a All 95% CIs for difference between medians were computed using Hodges-Lehmann estimator with asymptotic standard error. All 95% CIs for difference between 2 independent percentages were computed using Wilson procedure with correction for continuity.

^b The Entropy, Bispectral Index_, and SEDLine EEG monitors display the suppression ratio, which indicates the percent of suppressed EEG activity in approximately a 1-minute period.

^c Volatile anesthetic concentration is measured in multiples of minimum alveolar concentration (MAC), where MAC of 1.0 is the concentration at which patient movement is prevented with surgical stimulation in 50% of patients.

^d Age-adjusted MAC is calculated using the following equation: MAC₄₀ × 10^{(age − 40) × (−0.00269)} where MAC₄₀ is the MAC for a healthy person 40 years of age.

^e All opioids were converted into morphine equivalents using the following conversion factors: 1 mg morphine = 2 µg sufentanil; 1 mg morphine = 0.2 mg hydromorphone; 1 mg morphine = 10 µg fentanyl (https://clincalc.com/opioids/).

^f Patients who underwent combined procedures are reported under multiple surgical procedure types.

Prespecified Primary Outcome

Delirium assessments were missing for 8 patients. The main reason for ineligibility of delirium assessment was a withdrawal of consent. The main reason for nonassessment of delirium by CAM or CAM-ICU (approximately 15% of assessments) was patient’s refusal on at least 1 of the postoperative days (eTable 12 in Supplement 4), whereas 100% of chart reviews were completed for all eligible patients for postoperative days 1 to 5 (eTable 13 in Supplement 4). Missing CAM and CAM-ICU assessments were balanced between the EEG-guided and usual care groups over postoperative days 1 to 5 (eTable 13 in Supplement 4). Nonadherence with intervention assignment occurred with 5 patients (3 in the EEG-guided group and 2 in the usual care group), none of whom had delirium.

Delirium occurred in 102 of 562 patients (18.15%) in the EEG-guided group and 103 of 569 (18.10%) in the usual care group (difference, 0.05% [95% CI, −4.57% to 4.67%]) (Table 3). In both groups, the proportion of patients with delirium peaked on postoperative day 1 and then decreased daily over the following 4 days (eTable 14 in Supplement 4). By the log-rank test, time to delirium onset did not differ significantly between the groups (hazard ratio, 1.00 [95% CI, 0.76-1.32]; P > .99; Figure 2). After covariate adjustment for baseline characteristics, EEG guidance was not associated with decreased odds of delirium (adjusted odds ratio, 1.12 [95% CI, 0.80-1.55]; P = .51) (eTables 15A and 15B in Supplement 4). In this multivariable logistic regression analysis of unstandardized estimates, older age, higher EuroScore II score, history of delirium, and history of depression were all associated with significantly increased adjusted odds of postoperative delirium (eTables 15C, 16A, and 16B in Supplement 4).

Table 3. Primary, Secondary, and Exploratory Outcomes and Adverse Events.

Outcome	EEG guided (n = 562)	Usual care (n = 569)	Difference (95% CI)a	P value
Prespecified primary outcome, No./total No. (%)
Delirium incidence on postoperative day 1-5b	102/562 (18.15)	103/569 (18.10)	0.05 (−4.57 to 4.67)	>.99
Prespecified secondary outcomes, median (IQR)
Time spent in the intensive care unit (ICU), d	3.0 (2.0 to 4.0)	3.0 (2.0 to 4.0)	0 (−0.31 to 0.31)	.61
Hospital length of stay, d	9 (7.0 to 13.0)	9 (7.0 to 14.0)	0 (−0.94 to 0.94)	.76
Prespecified exploratory delirium-related outcomes on postoperative d 1-5
Incidence of severe delirium, No./total No. (%)c	36/544 (6.6)	30/542 (5.5)	1.08 (−1.94 to 4.12)	.54
Delirium on the day of the operation, No./total No. (%)	16/476 (3.4)	21/479 (4.4)	−1.02 (−3.74 to 1.65)	.52
Delirium duration, median (IQR), d	1 (1.0 to 3.0)	1 (1.0 to 2.0)	0 (−0.42 to 0.42)	.55
Nonprespecified exploratory delirium-related outcomes on postoperative d 1-5, No./total No. (%)d
Subsyndromal delirium incidence	286/562 (50.9)	311/569 (54.7)	−3.8 (−9.67 to 2.17)	.20
Delirium plus subsyndromal delirium incidence	388/562 (69.0)	414/569 (72.8)	−3.7 (−9.11 to 1.70)	.19
Other prespecified exploratory outcomes (30-d postoperative)e
Accidental falls	25/395 (6.3)	20/407 (4.9)	1.4 (−2.0 to 4.9)	.51
Patient Health Questionnaires score ≥9, No./total No. (%)f	6/395 (1.5)	5/407 (1.2)	0.3 (−1.71 to 2.37)	.97
AUDIT-C score ≥4 for males or ≥3 for females, No./total No. (%)g	77/395 (19.5)	75/407 (18.4)	1.1 (−4.5 to 6.7)	.75
PDEQ score, median (IQR)h	12 (10.0 to 15.0)	11 (10.0 to 15.0)	1.0 (0.47 to 1.53)	.44
PCL-5 score, median (IQR)i	3 (1.0 to 8.0)	3 (1.0 to 7.0)	0 (−0.85 to 0.85)	.93
PROMIS Cognitive abilities T-score, mean (SD)j	16.8 (4.01)	16.6 (4.11)	0.09 (−0.48 to 0.66)	.75
PROMIS Cognitive concerns T-score, Mean (SD)j	18.3 (2.77)	18.1 (2.81)	0.17 (−0.22 to 0.56)	.40
PROMIS Global-10 T-scores, mean (SD)j
Physical health–related quality of life	44.4 (5.54)	44.4 (5.81)	−0.001 (−0.79 to 0.79)	.998
Mental health–related quality of life	49.7 (7.06)	50.1 (6.95)	−0.41 (−1.39 to 0.57)	.41
Prespecified adverse events, No./total No. (%)
30-d postoperative mortality	8/567 (1.4)	13/573 (2.3)	−0.9 (−2.7 to 0.9)	.39
1-y postoperative mortality	19/567 (3.3)	19/573 (3.3)	0.04 (−2.26 to 2.33)	>.99
Undesirable intraoperative movementk	41/567 (7.2)	26/573 (4.5)	2.7 (−0.18 to 5.63)	.12
Intraoperative awareness	0/567	0/573	0 (−0.83 to 0.83)	NA
Postoperative nausea and vomiting	65/567 (11.5)	61/573 (10.6)	0.8 (−2.97 to 4.61)	.97
Patients with serious adverse eventsl	64/567 (11.3)	73/573 (12.7)	−1.4 (−5.37 to 2.47)	.78
Sensitivity analyses in relation to the primary outcome, No./total No. (%)
Missing delirium assessments assumed positivem	107/567 (18.87)	106/572 (18.53)	0.34 (−0.04 to 0.05)	.94
Missing delirium assessments assumed negativem	102/567 (17.99)	103/572 (18.01)	−0.02 (−0.05 to 0.5)	>.99
Excluding quartile of patients with highest MACm	72/407 (17.7)	103/569 (18.1)	−0.41 (−5.35 to 4.71)	.94
Excluding quartile of patients with most electroencephalogram (EEG) suppression timen	59/420 (14.1)	103/569 (18.1)	−4.05 (−8.71 to 0.79)	>.99
Delirium incidence per treatment receivedo	102/561 (18.18)	103/570 (18.07)	0.11 (−4.51 to 4.74)	>.99

^a 95% CIs computed using Hodges-Lehmann estimator with asymptotic standard error for medians and Wilson procedure with correction for continuity for percentages.

^b Patients were analyzed in the groups they were randomized.

^c With the Confusion Assessment Method (CAM) long-form; score ≥10 indicates severe delirium.

^d CAM diagnosis was reached when a chart review indicated acute onset or a fluctuating course and inattention and disorganized thinking or an altered level of consciousness. Subsyndromal delirium was present when the patient was not considered to have delirium by CAM, CAM-ICU, or the chart review, but met any of the individual criteria for delirium on the CAM-ICU or CAM.

^e Sample sizes at 30 days were 395 for EEG guided and 407 for usual care.

^f Patient Health Questionnaire-4 range, 0-12; ≤2 indicates no symptoms; 3-5, mild depression; 6-8, moderate depression; ≥9, severe depression.

^g Alcohol Use Disorders Identification Test-Concise (AUDIT-C) range, 0-12; 0, no alcohol use; ≥4 for men and ≥3 for women, alcohol use disorder.

^h Peritraumatic Dissociative Experiences Questionnaire (PDEQ) range, 10-50; higher scores indicate increased dissociation.

ⁱ Posttraumatic Stress Disorder (PTSD) Checklist for DSM-5 (PCL-5) range, 0-80; <33 generally does not meet criteria for PTSD and ≥33 suggests further assessment to confirm a diagnosis.

^j Patient Reported Outcomes Measurement Information System (PROMIS) Global-10 is a 10-items questionnaire used to assess the health care related quality of life. Raw scores were converted to a T-score using item-level calibrations. Higher T-scores (>50) indicate better perceived levels.

^k Mild: undesired spontaneous breathing or non-purposeful movement with no impact on the surgery; moderate: movement that mildly impacted the operation and required deepening of anesthesia, deepening of analgesia, or increased muscle relaxant; severe: movement with a marked negative impact on the operation.

^l See eTable 26 in Supplement 4 for details.

^m The 8 missing delirium assessments were assumed to all have delirium or not.

ⁿ Patients were excluded in the guided group with the highest quartile of median minimum alveolar concentration (MAC) of volatile anesthetic and cumulative suppression time.

^o If clinicians in the usual care group viewed EEG data, patients were included in the EEG-guided group; if EEG data were not able to be viewed in the guided group, patients were included in the usual care group.

Figure 2. Cumulative Incidence of Delirium Through Postoperative Day 5.

The shaded areas indicate 95% CIs. EEG indicates electroencephalography.

Prespecified Secondary Outcomes

There was no significant difference between groups in median length of ICU stay (difference, 0 [95% CI, −0.31 to 0.31] days; P = .61) or hospital stay (difference, 0 [95% CI, −0.94 to 0.94] days; P = .76) (Table 3). The median length of ICU and hospital stay differed among study sites (eTable 10 in Supplement 4).

Prespecified and Nonprespecified Delirium-Related Exploratory Outcomes

There was no significant difference between the groups in delirium duration, severe delirium incidence, delirium on the day of the surgical procedure, or subsyndromal delirium (Table 3). In both groups, patients with delirium had significantly more time with EEG suppression (eFigure 2 in Supplement 4). In a multivariable logistic regression analysis including preselected variables potentially linked to delirium (eTables 15A-15C and 17 in Supplement 4), older age, history of delirium, higher Euroscore II, duration of anesthesia, and duration of EEG suppression were associated with increased odds of delirium. The estimated increased odds of postoperative delirium associated with age was 4% for every additional year, 10% for every unit of Euroscore II score, and 19% for every additional hour of anesthesia. Study site was not associated with adjusted odds of delirium, and alcohol use was associated with decreased odds of delirium (eTable 17 in Supplement 4). The estimated increased odds of postoperative delirium associated with EEG suppression was 4% (adjusted odds ratio, 1.04; [95% CI, 1.04-1.07]; P < .01) for every additional 5 minutes with EEG suppression (eTable 17 in Supplement 4). There were statistically significant differences among study sites in delirium incidence, severe delirium, delirium on the day of the surgical procedure, and subsyndromal delirium (eTables 10 and 18 in Supplement 4). However, there were no significant differences between study groups at any site in postoperative or subsyndromal delirium incidence (eTable 18 in Supplement 4), including the sites with the largest percentage decreases in volatile anesthesia administration and EEG suppression time, there was still no statistically significant decrease in delirium incidence in the EEG-guided group (eTable 11 in Supplement 4).

30-Day and 1-Year Prespecified Exploratory Outcomes

There were no meaningful differences between groups in fall incidence or measures reflecting depressive symptoms, harmful alcohol use, posttraumatic stress disorder symptoms, cognitive abilities, and quality of life at postoperative day 30 (Table 3; eTables 19-21 in Supplement 4). Falls before the surgical procedure, but not delirium, were associated with falls within 30 day of the operation (eTable 22A in Supplement 4). Falls before the surgical procedure and delirium were associated with falls within 1 year of the operation (eTable 22B in Supplement 4). Euroscore II, length of hospital stay, and EEG suppression were associated with 1-year mortality (eTable 23 in Supplement 4).

Adverse and Serious Adverse Events

For the EEG-guided group, the hazard ratio was 0.62 (95% CI, 0.26-1.46; P = .28) for 30-day mortality and 1.06 (95% CI, 0.56-2.03; P = .85) for 1-year mortality (eFigure 3A and 3B in Supplement 4). There were no significant differences between the groups in relation to other adverse or safety outcomes. (Table 3; eTables 24-26 in Supplement 4). Undesirable intraoperative movement occurred in 41 patients (7.2%) in the EEG-guided group and 26 patients (4.5%) in the usual care group. No patients reported intraoperative awareness (eTable 24 in Supplement 4). Complications occurred in 137 of 1140 patients (12.0%): 64 of 567 (11.3%) in the EEG-guided group and 73 of 573 (12.7%) in the usual care group (Table 3; eTables 25-26 in Supplement 4).

Sensitivity Analyses

In the sensitivity analyses concerning the primary outcome, counting patients with all delirium assessments missing (n = 8) as all having delirium or as all not having delirium , and in the analyses excluding the quartile of patients with longest EEG suppression times and highest median volatile anesthetic concentrations, there was no statistically significant difference in delirium incidence between groups. There was also no significant difference between groups in delirium incidence in the sensitivity analysis, in which groups were analyzed according to treatment received (Table 3).

Discussion

This trial shows that using EEG to guide volatile anesthetic administration, thereby minimizing EEG suppression time, did not result in a significant reduction in the incidence, duration, or severity of postoperative delirium among older adults undergoing cardiac surgery. Other outcomes, including ICU and hospital stay and postoperative mortality, were also not significantly reduced by EEG guidance of anesthesia.

The results of this study are consistent with the Regional vs General Anesthesia for Promoting Independence after Hip Fracture (REGAIN) trial³¹ and the Effect of Regional vs General Anesthesia on Incidence of Postoperative Delirium in Older Patients Undergoing Hip Fracture Surgery (RAGA) trial.³² These trials found no significant difference in postoperative delirium incidence in older adults with fractured femurs, regardless of whether general anesthesia or spinal anesthesia, with minimal or no sedation, was administered. Conversely, the findings of ENGAGES-Canada are incongruent with the Cognitive Dysfunction after Anesthesia (CODA) trial¹¹ and the Balanced Anesthesia Delirium substudy.⁶ These trials found that EEG-guided anesthesia was associated with significantly decreased delirium incidence. One hypothetical explanation for the contradictory findings is that the control groups in the CODA¹¹ and Balanced Delirium substudy⁶ were exposed to deeper anesthesia than control group patients in the other trials. The ENGAGES-Canada trial differs from all these trials in that only patients undergoing cardiac surgery were included, meaning its findings are most relevant and generalizable to these patients.

The current trial had methodological strengths. First, the methods used were akin to those used in the ENGAGES trial.⁹ Regarding reproducibility, most of the findings of the current trial are congruent with the ENGAGES trial, which found that in older adults undergoing major surgical procedures, EEG guidance to decrease anesthetic exposure and EEG suppression was not associated with significantly decreased delirium incidence, duration, or severity.⁹ Second, the current trial addresses several limitations identified with the ENGAGES trial.^33,34 The current trial was multicenter, whereas ENGAGES was a single-center trial. Anesthetic management differed across sites in the current trial; however, at each site, EEG guidance resulted in a substantial separation between groups in EEG suppression time and inhaled anesthetic concentration without a significant difference in delirium incidence. Also, 3 different EEG monitors were used in the current study, suggesting that the nonsuperior findings with EEG guidance are unlikely to be attributable to the unreliability of a single proprietary EEG device or algorithm. Lastly, in the ENGAGES trial, there was an 11% decrease in anesthetic concentration and a 46% decrease in EEG suppression time in the guided group. In the current trial trial, these decreases were 17.5% and 66%, respectively, suggesting that EEG guidance had an even more profound impact on anesthetic management.

Limitations

This trial has several limitations. First, despite the 66% reduction in EEG suppression time, a greater reduction might be required to impact postoperative delirium incidence. However, this seems unlikely because there was only a median of 4 minutes of cumulative EEG suppression time in the EEG-guided group. Second, the 17.5% decrease in volatile anesthetic administration in the EEG-guided group might have been insufficient to prevent delirium. Third, delirium might have been underdetected. However, this is unlikely because the delirium incidence in the current trial was similar to that reported in other recent trials of patients undergoing cardiac surgery, including those with more frequent postoperative delirium assessments.^3,35 Fourth, the results of this trial remain consistent with the possibility that EEG suppression minimization has a smaller effect on delirium reduction than the 8% that was estimated when designing the trial. A trial of 2700 patients would be required to detect a decrease in delirium incidence from 18% to 14% and a trial of 11 000 patients would be required to detect a decrease from 18% to 16%.

Conclusions

Among older adults undergoing cardiac surgery, EEG-guided anesthetic administration to minimize EEG suppression, compared with usual care, did not decrease the incidence of postoperative delirium. This finding does not support EEG-guided anesthesia for this indication.

Supplement 1.

Trial Protocol

Supplement 2.

Consort Checklist

Supplement 3.

Statistical Analysis Plan

Supplement 4.

eTables and eFigures

Supplement 5.

Nonauthor Collaborators

Supplement 6.

Data Sharing Statement