Title

Author(s)

Citation

Issued Date

URL

Rights

Patient outcome and brain state monitoring during general anesthesia

Chan, Tak-vai; 陳德威 Chan, T. [陳德威]. (2015). Patient outcome and brain state monitoring during general anesthesia. (Thesis). University of Hong Kong, Pokfulam, Hong Kong SAR. Retrieved from http://dx.doi.org/10.5353/th_b5570773 2015

http://hdl.handle.net/10722/226886

Creative Commons: Attribution 3.0 Hong Kong License

Abstract of thesis entitled

Patient Outcome and Brain State Monitoring During General Anesthesia Submitted by

Tak Vai CHAN for the degree of Doctor of Philosophy at The University of Hong Kong in July 2015

General anesthesia is the reversible state of drug-induced unconsciousness. Conventional tracking of depth of anesthesia relied heavily on the changes in automatic and somatic responses to surgical stimulations. Recent development of intraoperative neurophysiologic recordings, such as the bispectral electroencephalographic index and the auditory evoked potential, has provided real-time information on anesthetic drug effects, and may facilitate adjustment of anesthetic dosing. This thesis aimed to evaluate the impact of brain state monitoring on patient outcomes after general anesthesia.

Firstly, a randomized controlled trial was performed in 1,063 patients using auditory evoked potential recordings or routine clinical signs to guide anesthetic administration. Auditory evoked potential monitoring reduced anesthetic dosage by 15-30%. This was associated with quicker recovery, shorter hospital stay and superior quality of recovery. In addition, there was a lower incidence of postoperative nausea

and vomiting (19.0% versus 47.9%, p < 0.001) and other infective complications (8.6% versus 11.3%, p = 0.028).

Secondly, in 921 elderly patients having major noncardiac surgery, a randomized controlled trial was conducted to compare the effect of bispectral index monitoring or routine care on postoperative cognitive outcome. Anesthetic dosage was reduced by 20-30% with bispectral index monitoring. Following surgery, the incidence of delirium was reduced in the monitoring group compared with routine anesthetic care (15.6% versus 24.1%, p = 0.010). The risk for postoperative cognitive dysfunction, measured by detailed neuropsychology test battery, was also decreased with bispectral index monitoring at 3 months after surgery (10.2% versus 14.7%, adjusted odds ratio (95% confidence intervals): 0.67 (0.32-0.98), p = 0.025).

Thirdly, a systematic review and meta-analysis was performed on 6 awareness trials involving 34,211 patients. Bispectral index monitoring was associated with a decreased risk of unintentional awareness during general anesthesia (0.13% in monitoring group versus 0.25% in controls), relative risk (95% confidential intervals): 0.54 (0.31-0.94), p = 0.028. The effect was most obvious in patients receiving intravenous anesthesia and those undergoing noncardiac surgery.

Finally, a long-term follow-up study was performed in 536 high-risk surgical patients participated in an awareness trial. By adopting bispectral index monitoring to avoid episodes of deep anesthesia, long-term mortality was decreased, adjusted hazard

ratio (95% confidence intervals) 0.85 (0.74-0.98), p = 0.042. In addition, the rates of myocardial infarction and stroke were reduced when anesthetic was adjusted to maintain a bispectral index value within the therapeutic range.

The findings highlighted the utility of brain state monitoring in optimizing anesthetic administration. When unexpected periods of inadequate anesthesia was avoided, brain state monitoring was effective in the prevention of awareness during general anesthesia. Similarly, as unnecessarily deep level of anesthesia was prevented with monitoring, there was a substantial decrease in anesthetic exposure. Consequently, an improvement in the quality of recovery was observed with superior cognitive outcome and a decrease in long-term mortality and morbidity. The series of studies imply a greater role for anesthesia in determining patient outcome than previously acknowledged. The practice of brain state monitoring during general anesthesia should be encouraged to improve perioperative outcome, especially in high-risk patients undergoing complex major surgery.

_______________________________ An abstract of exactly 499 words

Patient Outcome and Brain State Monitoring During General Anesthesia by

Tak Vai CHAN (陳德威) M.B. B.S. (Hons) University of New South Wales, Australia

A thesis submitted in partial fulfillment of requirements for the Degree of Doctor of Philosophy at The University of Hong Kong July 2015

Clinical trials reported in this thesis

DECLARATION

I declare that this thesis entitled “Patient Outcome and Brain State Monitoring During General Anesthesia” represents my own original work, except where due acknowledgement is made, and that it has not been previously included in a thesis, dissertation or report submitted to the University or to any other institution for a degree, diploma or other qualifications.

________________________________ Tak Vai CHAN

i

ACKNOWLEDGEMENTS

I wish to thank the Department of Anaesthesiology, The University of Hong Kong and Professor KF Ng (supervisor) for giving me the opportunity to complete this work. I am most grateful to my collaborators for their advices and contributions to the trials. Special thanks to Professors Paul Myles and Andrew Forbes (Monash University, Australia), Professor Kate Leslie (Royal Melbourne Hospital, Australia), Professor Tatia Lee (The University of Hong Kong), Professor Tony Gin (The Chinese University of Hong Kong), Dr Erik Jensen (Universitat Politècnica de Catalunya, BarcelonaTech, Spain), Drs Anita Chan, PT Chiu, Leo Chu, MC Chu, Patricia Kan, Emily Koo, (Prince of Wales Hospital), Dr Bassanio Law (Kwong Wah Hospital), Drs Benny Cheng, YL Ho, KK Lam, CW Lau (Tuen Mun Hospital), and Drs KK Liu, Wendy Wong (North District Hospital).

I am grateful to our trial coordinators and research assistants who helped with patient recruitment and data collection - Lydia Chow, Beaker Fung, Angel Ip, Annie Lok, Candy Lok, Angela Mou, Zoe Sun, Matthew Tsang, Ming Wai Tsang, Louise Wong, Kevin Yau and Joy Yip. I would like to thank our computer officers, Thomas Lo and Alex Lee, for maintaining the online randomization services and trial database. I am also grateful to all anesthesiologists, surgical and nursing staffs who allowed us to conduct clinical trials on their patients in the operating rooms and also in the surgical wards and clinics.

I wish to acknowledge the funding support from a number of grant administrative institutions - Research Grants Council, Hong Kong SAR: General ii

Research Fund, #450402, #440006; Food and Health Bureau, Government of Hong Kong SAR: Health and Health Services Research Fund #04060271 and The Chinese University of Hong Kong: Direct Grant for Research, Project codes: #2041096, #4450149.

Last but not the least, I must thank the heavenly Father, my parents, my wife Candy and the two children - Sam and Winnie for their unconditional support and encouragement throughout my study.

iii

TABLE of CONTENTS Declaration

.......................................................................................................... i

Acknowledgements..................................................................................................... ii Table of Contents ....................................................................................................... iv List of Tables

........................................................................................................ ix

List of Figures

........................................................................................................ xi

List of Abbreviations ............................................................................................... xiii

CHAPTER 1

..........................................................................................................1

1.1 Literature Review ...............................................................................................1 1.2 Clinical Measures of Depth of Anesthesia .........................................................1 1.3 Autonomic Response to indicate Depth of Anesthesia.......................................4 1.4 Isolated Forearm Technique ...............................................................................8 1.5 Monitoring Anesthetic Drug Delivery ..............................................................11 1.6 Processed Electroencephalographic Monitoring ..............................................13 1.6.1 Time Domain Analysis ...............................................................................14 1.6.2 Frequency Domain (Spectral) Analysis ......................................................15 1.6.3 Bispectral Analysis .....................................................................................17 1.6.4 Spectral Entropy Analysis ..........................................................................17 1.6.5 Topographic (Spatial) Analysis ..................................................................18 1.6.6 Other EEG Processing Algorithms .............................................................19 1.7 Auditory Evoked Potential Monitoring ............................................................20 1.8 Monitoring Brain State during General Anesthesia..........................................23 1.9 Clinical Utility of Brain State Monitoring during Anesthesia ..........................27 1.9.1 Recovery Times ..........................................................................................27 1.9.2 Awareness ...................................................................................................28 1.9.3 Other Outcomes ..........................................................................................28 iv

1.10 Conclusions.......................................................................................................28 CHAPTER 2

........................................................................................................30

2.1 Study Designs, Hypotheses and Objectives......................................................30 2.2 Rate and Quality of Recovery...........................................................................30 2.3 Awareness during Anesthesia ...........................................................................31 2.4 Postoperative Cognitive Dysfunction ...............................................................31 2.5 Long-Term Mortality and Morbidity ................................................................31 CHAPTER 3

........................................................................................................33

3.1 Quality of Recovery after AEP-Guided Anesthesia .........................................33 3.2 Methods

........................................................................................................34

3.2.1 Study Design ...............................................................................................34 3.2.2 Patient Population .......................................................................................34 3.2.3 Procedures...................................................................................................35 3.2.4 Trial Outcome .............................................................................................37 3.2.5 Statistical Analysis and Sample Size Calculation ......................................39 3.3 Results

........................................................................................................39

3.3.1 Baseline Characteristics ..............................................................................39 3.3.2 Anesthetic Delivery with AEP Monitoring ................................................41 3.3.3 Outcomes after AEP Monitoring ................................................................45 3.4 Discussion ........................................................................................................50 3.4.1 Comparison to other Studies .......................................................................50 3.4.2 AEP Monitoring and Complications ..........................................................51 3.4.3 Intravenous versus Inhalational Anesthesia................................................52 3.4.4 Strengths and Limitations ...........................................................................53 3.4.5 Interpretation and Conclusions ...................................................................53 CHAPTER 4

........................................................................................................54

4.1 Brain State Monitoring to Prevent Awareness during General Anesthesia ......54 v

4.1.1 Psychological Consequences of Awareness ...............................................54 4.1.2 Monitoring Awareness................................................................................55 4.1.3 The B-Aware Trial ......................................................................................55 4.1.4 Meta-Analysis of Randomized Controlled Trials .......................................56 4.2 Methods

........................................................................................................56

4.2.1 Study Eligibility ..........................................................................................56 4.2.2 Study Identification ....................................................................................56 4.2.3 Eligibility Assessment ................................................................................58 4.2.4 Data Collection and Quality Assessment ...................................................58 4.2.5 Statistical Analysis......................................................................................58 4.3 Results

........................................................................................................59

4.3.1 Included Trials ............................................................................................59 4.3.2 Effect of BIS Monitoring ............................................................................63 4.3.3 Subgroup Analyses .....................................................................................65 4.4 Discussion ........................................................................................................69 4.4.1 BIS Monitoring to Prevent Awareness .......................................................70 4.4.2 Strengths and Limitations ...........................................................................71 4.4.3 Conclusions.................................................................................................71 CHAPTER 5

........................................................................................................73

5.1 Cognitive Outcome after BIS-Guided Anesthesia ............................................73 5.1.1 Brain State Monitoring and Anesthetic Administration .............................73 5.1 Methods

........................................................................................................74

5.2.1 Study Design ...............................................................................................74 5.2.2 Study Population .........................................................................................74 5.2.3 Randomization and Blinding ......................................................................75 5.2.4 Study Procedure ..........................................................................................75 5.2.5 Cognitive Measurements ............................................................................77 vi

5.2.6 Outcome ......................................................................................................79 5.2.7 Statistical Analyses .....................................................................................79 5.3 Results

........................................................................................................80

5.3.1 Primary and Secondary Outcomes..............................................................83 5.3.2 Risk Factors of Postoperative Cognitive Dysfunction and Delirium .........90 5.4 Discussion ........................................................................................................93 5.4.1 Principal Findings .......................................................................................93 5.4.2 Postoperative Cognitive Dysfunction .........................................................93 5.4.3 Postoperative Delirium ...............................................................................97 5.4.4 Strengths and Weakness ...........................................................................102 5.4.5 Interpretation and Conclusions .................................................................102 CHAPTER 6

......................................................................................................104

6.1 Long-Term Morbidity and Mortality ..............................................................104 6.1.1 Deep Anesthesia and Postoperative Mortality..........................................104 6.1.2 Long-Term Survival in the B-Aware Trial ...............................................105 6.2 Methods

......................................................................................................106

6.2.1 Long-Term Follow-Up Process ................................................................106 6.2.2 Study Endpoints ........................................................................................107 6.2.3 Statistical Analyses ...................................................................................108 6.3 Results

......................................................................................................109

6.3.1 Predictors of Long-Term Mortality ..........................................................111 6.3.2 Predictors of Myocardial Infarction and Stroke .......................................114 6.4 Discussion ......................................................................................................117 6.4.1 Comparison to other Studies .....................................................................117 6.4.2 Strengths and Limitations .........................................................................120 6.4.3 Conclusions...............................................................................................121 CHAPTER 7

......................................................................................................122 vii

7.1 Conclusions and Future Perspectives .............................................................122 7.2 Benefits of Brain State Monitoring.................................................................122 7.3 Limitations with Brain State Monitoring ........................................................126 7.4 Research Agenda ............................................................................................126 7.4.1 Optimal Depth of Anesthesia....................................................................126 7.4.2 Composite Measure of Anesthetic Depth .................................................130 7.5 Conclusion ......................................................................................................130 APPENDIX 1: Psychometric Properties of Chinese Quality of Recovery Score .131 A1.1 Introduction...............................................................................................131 A1.2 Methods ....................................................................................................131 A1.3 Results.......................................................................................................131 A1.4 Conclusions...............................................................................................134 APPENDIX 2: Post-Traumatic Stress Disorder in the B-Aware Trial ..................138 A3.1 Introduction...............................................................................................138 A3.2 Methods: Clinician Administered PTSD Scale ........................................138 A3.3 Results.......................................................................................................139 A3.4 Conclusions...............................................................................................141 APPENDIX 3: Neuropsychology Test Battery .....................................................142 A3.1 Verbal Fluency Test ..................................................................................142 A3.2 Chinese Auditory Verbal Learning Test ...................................................144 A3.3 Color Trails Test .......................................................................................148 A3.4 Cognitive Failure Questionnaire ...............................................................153 References

......................................................................................................156

List of Publications ..................................................................................................204 Conference Presentations.........................................................................................205

viii

LIST of TABLES Chapter 1 Table 1.1. Pressure, pulse Rate, Sweating and Tear (PRST) score .............................5 Table 1.2. Monitoring of autonomic response as a measure of depth of anesthesia ...6 Table 1.3. Isolated forearm responses during anesthesia and surgery .........................9 Table 1.4. Minimum Alveolar Concentration (MAC) and MAC-awake ..................12 Table 1.5. Characteristics of the currently available monitors of anesthetic depth ...25 Chapter 3 Table 3.1. Patient characteristics of the AEP trial .....................................................42 Table 3.2. Drug delivery and anesthetic depth in AEP trial ......................................43 Table 3.3. Secondary outcomes after AEP-guided and routine care anesthesia ........49 Chapter 4 Table 4.1. Search terms and selection process for systematic review .......................57 Table 4.2. Design characteristics of included trials ...................................................61 Table 4.3. Risk of bias table for included trials .........................................................62 Chapter 5 Table 5.1. Patient characteristics at entry of CODA trial ..........................................82 Table 5.2. Comparison of anesthetic techniques .......................................................84 Table 5.3. Performance in neuropsychology tests .....................................................85 Table 5.4. Postoperative cognitive outcomes ............................................................86 Table 5.5. Recovery profiles and postoperative complications .................................88 Table 5.6. Risk factors of cognitive dysfunction at 3 months after surgery ..............91 Table 5.7. Risk factors of postoperative delirium ......................................................92 Table 5.8. Characteristics of studies evaluating postoperative delirium .................100

ix

Chapter 6 Table 6.1. Patient characteristics in long-term follow-up study ..............................110 Table 6.2. Hazard ratios for death ...........................................................................112 Table 6.3. Odds ratios for myocardial infarction .....................................................115 Table 6.4. Odds ratios for stroke .............................................................................116 Table 6.5. Relationship between anesthetic depth and long-term mortality............118 Chapter 7 Table 7.1. Benefits with brain state monitoring during general anesthesia .............123 Appendix 1 Table A1.1. Agreement between the Chinese and English QoR scores ..................132 Table A1.2. The inter-item polychoric correlation matrix .......................................133 Table A1.3. Test-retest and inter-observer agreement of the Chinese quality of recovery score ......................................................................................133 Table A1.4. Concordance and weighted kappa of QoR scores after major and minor surgery ..................................................................................................135 Appendix 2 Table A2.1. Characteristics of patients with and without awareness ......................139 Table A2.2. Diagnosis for post-traumatic stress disorder........................................140

x

LIST of FIGURES Chapter 1 Figure 1.1. Schematic diagram of the Guedel chart ....................................................3 Figure 1.2. Electroencephalogram (EEG) spectral analysis ......................................16 Figure 1.3. Algorithm for autoregressive modelling of auditory evoked potential ...22 Chapter 3 Figure 3.1. Electrode montage and screen display for AEP monitoring ...................36 Figure 3.2. Enrollments in the AEP trial ...................................................................40 Figure 3.3. Emergence time from anesthesia.............................................................44 Figure 3.4. Quality of recovery after AEP-guided or routine care anesthesia ...........46 Figure 3.5. Late recovery times from surgery ...........................................................47 Chapter 4 Figure 4.1. Selection of trials flow chart ...................................................................60 Figure 4.2. Relative risk of awareness in all included trials ......................................64 Figure 4.3. Relative risk of awareness in total intravenous or volatile-based anesthesia .................................................................................................................66 Figure 4.4. Relative risk of awareness in cardiac and noncardiac surgery ................67 Figure 4.5. Relative risk of awareness in trials with and without active controls .....68 Chapter 5 Figure 5.1. Flow chart of trial enrollment..................................................................81 Figure 5.2. Kaplan-Meier estimates of intensive care unit stay and hospital discharge .................................................................................................................89 Figure 5.3. Relative risk of postoperative cognitive dysfunction ..............................96 Figure 5.4. Relative risk of postoperative delirium .................................................101

xi

Chapter 6 Figure 6.1. Patient flow chart of B-Aware long-term follow-up study ...................109 Figure 6.2. Kaplan–Meier estimates of mortality during the follow-up period.......113 Chapter 7 Figure 7.1. Model of low EEG index and perioperative morbidity and mortality ..128 Figure 7.2. Range of BIS in patients randomized to receive BIS 35 or 50 .............129 Appendix 2 Figure A2.1. Difference in CAPS scores between case (awareness) and control (not aware) patients ....................................................................................141

xii

LIST of ABBREVIATIONS AAI

ARX-derived AEP Index

ABM

Anesthesia and brain monitor

AEP

Auditory evoked potential

AMIC

Anesthesia Multimodal Index of Consciousness

ANI

Analgesia nociception index

ARX-model

Autoregressive model with exogenous input

ASA

American Society of Anesthesiologists

B-Aware Trial

Bispectral Index Monitoring to Prevent AWAREness During Anesthesia Trial

BIS

Bispectral index

BAG-RECALL Trial

BIS or Anesthetic Gas to Reduce Explicit Recall Trial

CAPS

Clinician administered post-traumatic stress disorder scale

CCIP

Computer Control Infusion Pump program

CFM

Cerebral function monitor

CI

Confidence intervals

CMS

Clinical Management System

CODA Trial

COgnitive Dysfunction after Anesthesia Trial

CFQ

Cognitive failure questionnaire

CSI

Cerebral state index

ECG

Electrocardiography

EEG

Electroencephalography

EMG

Electromyography

ENIGMA Trial

Evaluation of Nitrous oxide in the Gas Mixture of Anesthesia Trial

xiii

ETAG

End-tidal anesthetic gas

HR

Hazard ratio

HRV

Heart rate variability

ICU

Intensive care unit

IFT

Isolated forearm technique

IoC

Index of consciousness

ITT

Intention-to-treat

κw

Weighted kappa statistic

LOC

Lower oesophageal contractility

MAC

Minimum alveolar concentration

MACS Trial

Michigan Awareness Control Study

MMSE

Mini-mental state examination

OR

Odds ratio

POCD

Postoperative cognitive dysfunction

PONV

Postoperative nausea and vomiting

PRST score

Pressure, pulse rate, sweating and tear score

PSI

Patient state index

PTSD

Post-traumatic stress disorder

QoR

Quality of Recovery

RCT

Randomized controlled trial

RE

Response entropy

RR

Relative risk

SAFE Study

Swedish Awareness Follow-up Evaluation Study

SE

State entropy

SF-36

Short-form health survey

xiv

TESPAR

Time encoded signal processing and recognition algorithm

SuDoCo Trial

Surgery Depth of anesthesia Cognitive outcome Trial

xv

CHAPTER 1

1.1

Literature Review Suppression of clinical responses, such as arousal, limb movements and

autonomic changes are desirable features of general anesthesia.1,2 As the dose of anesthetic is gradually increased, there is progressive loss of consciousness. When sufficient doses of anesthetics are given, somatic and autonomic responses to noxious stimuli, such as laryngoscopy, tracheal intubation and surgical incision, can be abolished. Therefore, patient response to surgery depends on the amount of anesthetics received and the intensity of stimulus applied.1 Inadequate anesthesia may lead to exaggerated response and unintentional awareness during surgery. On the contrary, excessive anesthesia may result in cardiovascular, respiratory and possibly neurological side effects. In order to facilitate titration of anesthetic delivery, it seems logical to develop a measure that could be used to quantify the effect of anesthetics, commonly known as the “depth of anesthesia”.1,3,4 The purpose of this chapter is to review the development of clinical and neurophysiological measures to inform the underlying depth of anesthesia.

1.2

Clinical Measures of Depth of Anesthesia The need to evaluate anesthetic depth during surgery was evident soon after

the introduction of ether anesthesia. In a letter published in the Lancet, Plomley described three stages of ether anesthesia in 1847.5 In his letter, the first stage of anesthesia was associated with a “pleasure feeling of half intoxication”. This was followed by “extreme pleasure” but “not exactly insensible to pain”. Surgery performed during these stages were “unsatisfactory”. It was only during the third 1

stage, when the patient was “completely lost to pain, and to external impressions”, surgery could then be successfully performed.5 Later in the same year, Snow independently described the effects of anesthesia in detail. In the book “On the inhalation of the vapor of ether in surgical operations”, he reported five stages of anesthesia.6 During the first two stages, patient remained awake and responded to verbal commands. In the third stage, patient lost consciousness but might respond to “external stimulus”. At a deeper level of anesthesia (the fourth stage), surgery could be conducted and patient would not be “influenced by external impressions”. It should be noted that there was a fifth stage of anesthesia when excessive amount of ether was administered. In this stage “respiratory movements are more or less paralyzed, and become difficult, feeble, or irregular”.6

The next landmark development occurred during the Great War.7 As one of the few anesthesiologists available in the Western Front, Guedel developed a chart on the stages of anesthesia to guide physicians, nurses and ward assistants in giving ether to the large number of wounded soldiers. The original Guedel chart described the changes in respiration, pupil size, eye and laryngeal movements during ether anesthesia.8,9 This was subsequently expanded to include eye (lid, corneal and pupillary light) reflexes, lacrimation, muscle tone, pharyngeal, laryngeal and respiratory responses to surgery.10 Figure 1.1 shows the extended Guedel chart that divides the anesthetic continuum into four stages. Stage 1 is the start of anesthetic administration to the loss of consciousness. Stage 2 is a period of delirium (excitement). Stage 3 is known as the surgical stage and is divided into 4 planes according to presence of intercostal paralysis.11 Stage 4 is the result of anesthetic overdose, where there is respiratory paralysis and severe hypotension.

2

Figure 1.1. Schematic diagram of the Guedel chart

.

The columns on eye (lid, cornea, pupillary light) reflex, pharyngo-laryngeal movement, tear production, muscle tone and respiratory response to incision were added by Gillespie.10 The wedges indicate the progressive disappearance of signs and reflexes. The horizontal bars shows the period when the events (e.g. swallowing, retching and vomiting) may occur.9,12

3

These early reports highlight the importance of monitoring during general anesthesia. It would be important to ensure an appropriate level of anesthesia is achieved and to avoid anesthetic overdose.

1.3

Autonomic Response to indicate Depth of Anesthesia The Guedel chart was popular for gauging anesthetic administration,

especially when inhalational anesthetic was used as the sole agent. However, clinical assessment using the Guedel chart became invalid with the introduction of muscle relaxants in 1942.13 In this respect, somatic responses cannot be elicited following neuromuscular blockade.14 Therefore, anesthesiologists have to rely on autonomic signs such as hypertension, tachycardia, pupillary dilatation, lacrimation and sweating to indicate inadequate anesthesia. A number of systems have been developed to quantify the autonomic changes during surgery.

The PRST score integrates changes in arterial Pressure, pulse Rate, Sweating and Tear production to reflect the depth of anesthesia (Table 1.1).15 A score of 0 or 1 (range between 0 and 8) indicates adequate and possibly deep anesthesia, whereas a score ≥ 6 indicates inadequate anesthesia. The PRST score is easy to measure. However, in patients with confirmed awareness during anesthesia, hypertension and tachycardia occurred in < 15% of cases.16 In another study, intraoperative hemodynamic changes in patients with confirmed awareness during anesthesia were indistinguishable from matched controls.17 These findings challenge the validity of the PRST score. Despite the limitations, clinical assessment remains as the main tool for monitoring anesthetic depth.

4

Table 1.1. Pressure, pulse Rate, Sweating and Tear (PRST) score Index

Condition

Systolic blood pressure (mmHg)

< control + 15 < control + 30 > control + 15 < control + 15 < control + 30 > control + 15 Nil Skin moist to touch Visible beads of sweat No excess tears when eyelids opened Excess tears when eyelids opened Tears overflow from closed eyelids

Heart rate (beats/min)

Sweat

Tears or lacrimation

Score 0 1 2 0 1 2 0 1 2 0 1 2

The failure of PRST score could be related to the fact that subtle changes in autonomic responses are easily missed by clinical assessment. Therefore, a number of investigators have developed objective measures to quantify autonomic perturbations during anesthesia.18-25 Common modalities include lower oesophageal contractility, heart rate variability, skin conductance, reflex pupillary dilatation, surface electromyography and ocular microtremor (Table 1.2). Unfortunately, these measurements are nonspecific and are subjected to the influence of concomitant medications such as β blockers, anticholinergics and opioids. Furthermore, it has been suggested that movement and autonomic responses are primarily spinal reflexes indicating the analgesic, rather than the hypnotic state of the patient.26-32 Interestingly, three of the devices - the ANI monitor (MDoloris Medical Systems, Lille Cedex, France),19 the Algesimeter (Medstorm Innovations, Oslo, Norway)27 and the AlgiScan (Equip Medikey, Gouda, Netherlands),28 are currently marketed as monitors of intraoperative analgesia instead of hypnosis. 5

Table 1.2. Monitoring of autonomic response as a measure of depth of anesthesia Measurement / Principle Lower oesophageal contractility (LOC)18,33 LOC measures spontaneous and provoked (in response to balloon inflation) contractions of the distal oesophageal smooth muscle.

Heart rate variability (HRV)19 This is a measure of heart rate changes (expressed as ECG RR intervals) during anesthesia. Using frequency domain analysis with fast Fourier transformation, HRV can be broadly classified into 3 components:34,35 (1) Very low frequency component (0.004 - 0.04 Hz) is a measure of thermoregulatory activity; (2) Low frequency component (0.04 - 0.15 Hz) is related to sympathetic discharges and (3) High frequency component (0.15 - 0.4 Hz) is a measure of parasympathetic activity. Skin conductance20 Skin conductance measures the current that passes through the palmer surface of the hand for any given voltage applied. During stress, the sweat glands are filled, leading to a decrease in skin conductance. Therefore, skin conductance is a measure of sympathetic activity.20

Interpretation during anesthesia Light anesthesia or awakening from anesthesia results in frequent (> 4/min) spontaneous, irregular, non-peristaltic contractions (duration 5 s, amplitude 30 - 50 mmHg). Adequate anesthesia is reflected by a suppression of provoked oesophageal contraction (within 5 s of oesophageal balloon inflation). During anesthesia, HRV is dominated by the high frequency component under the influence of the parasympathetic activity. As anesthesia becomes light or in the presence of intense noxious stimulation, HRV is replaced by low frequency spectrum. HRV is incorporated in the “Analgesia Nociception Index (ANI)” (ANI monitor, MDoloris Medical Systems, Lille Cedex, France), and is marketed as an indicator of analgesia during surgery.36

During inadequate anesthesia, there is an increase in sympathetic activity and the number of skin conductance fluctuations per second is increased.37 Skin conductance monitoring is commercially available (Skin Conductance Algesimeter, Medstorm Innovations, Oslo, Norway).

6

Reflex pupillary dilatation Pupillary dilatation in response to noxious stimuli is a measure of sympathetic reserve and is independent to the type of anesthetic used.

Surface electromyography In partially paralyzed patients, an increase in electromyographic (EMG) activity is an indirect measure of stress during general anesthesia and surgery.

Ocular microtremor24,38 This is similar to facial EMG monitoring during general anesthesia, but extraocular EMG activity cannot be measured using surface electrodes. A piezoelectric strain gauge device is mounted so that it is in contact with the scleral surface. Eyeball displacement is recorded.

Reflex pupillary dilatation indicates inadequate depth of anesthesia or level of analgesia.21 AlgiScan (Equip Medikey, Gouda, Netherlands) is a portable infrared pupillometer that is able to measure real-time pupillary diameter. The extent of pupillary dilatation in response to tetanic stimulation over the ulnar nerve produces a “pain pupillary index”, ranging from 0 (no pain) to 10 (severe pain). The FACE monitor (Patient Comfort, Inc., Nevada City, CA) measures EMG activities in 4 facial muscles (frontalis, orbicularis occuli, corrugator and zygomaticus).23 The Anesthesia and Brain Monitor (ABM Datex-Ohmeda, Madison, WI) measures the frontalis EMG alone.22,25 In both monitors, an increase in the amplitude of EMG indicates inadequate anesthesia. An increase in ocular microtremor is an indication of inadequate anesthesia.

7

1.4

Isolated Forearm Technique In order to overcome the effect of muscle relaxants, Tunstall devised the

isolated forearm technique (IFT) to assess anesthetic depth.39 In this method, an arterial tourniquet is placed over the forearm and is inflated immediately before the administration of muscle relaxant. This technique allows patients to respond with the “isolated forearm”. Purposeful movements to specific verbal commands or surgical stimuli are considered as positive responses. The tourniquet is deflated 5-15 min later to prevent ischemic paralysis. The cuff is inflated again before further bolus dose of muscle relaxant is administered.40

Table 1.3 summarizes the responses of IFT monitoring from 26 studies involving a total of 1,153 patients. Typically, these studies reported a large proportion of patients who respond to verbal command (average 36.8%, range 0-100%), indicating that these patients were awake at some points during surgery. In contrast, only few patients were able to recall these experiences (average 2.34%, range 041.7%). The pooled sensitivity and specificity for IFT to detect intraoperative awareness were 77.8% [95% confidence intervals (CI): 57.7-91.3%] and 64.2% (61.367.0%), respectively. In addition, IFT response in all the reported series did not correlated with other measures of anesthetic depth. There is ongoing debate on the validity of IFT. It is now believed that the biologic mechanisms underlying amnesia and motor responsiveness may be different and could contribute to the discrepancy between recall and movement.41-43 Although IFT does not require expensive equipment or proprietary software, it will depend on the dedicated attention of a vigilant anesthesiologist to detect and to act upon subtle response in the isolated forearm.44,45

8

Table 1.3. Isolated forearm responses during anesthesia and surgery

Author, Year

Surgery

Anesthetic technique

Tunstall, 197739 Russell, 198646

Cesarean section Major gynecology

Watanabe wt al, 198847 Tunsatll & Sheikh, 198948 Baraka et al, 199049 Bodgod et al, 1990

Cardiac surgery Cesarean section

NR Nitrous oxide or etomidate infusion Fentanyl 25-100 µg/kg Isoflurane 1.25% or enflurane 1.5% Ketamine Enflurane/nitrous oxide

Russell, 199350

Major gynecology

King et al, 199351 Gaitini et al, 199552

Cesarean section Cesarean section

Byers et al, 1997 Russell & Wang, 199753 Flaishon et al, 199754 St. Pierre et al, 200055 Russell & Wang, 200156

Tonsillectomy Major gynecology Noncardiac surgery Noncardiac surgery Major gynecology

Cesarean section Cesarean section

No. of patients with Number of patients Unverified Verified Recall response response 12 11 4 0 55 31 13 1 10 113

NR NR

8 47

0 0

20 74

NR 27

3 32

0 2*

Midazolam/alfentanilpropofol Halothane-nitrous oxide Ketamine-nitrous oxide

33

23

26

3**

30 50

NR NR

29 18

0 0

Halothane Halothane-nitrous oxide Propofol/thiopentone Etomidate 0.2-0.4 mg/kg Propofol-alfentanil infusion

41 35 40 30 40

NR 10 NR NR NR

8 0 40 17 7

0 4 0 1 0

Alternative depth assessment -

Lower oesophageal contractility PRST score Processed EEG Spectral edge frequency BIS -

9

Loveman et al, 200157

Cardiac surgery

Propofol infusion

14

NR

6

0

Schneider et al, 200258

Non-neurosurgical surgery Peripheral surgery Ambulatory surgery Orthopedic or general surgery Major gynecology Cardiac surgery

Propofol infusion

20

NR

8

0

NR Propofol infusion Propofol infusion

41 65 56

NR NR NR

10 37 37

Propofol infusion Propofol infusion

12 10

NR NR

12 4

Major gynecology

Propofol-remifentanil infusion Isoflurane or sevoflurane

51

NR

7

0

Narcotrend Middle latency response BIS

184

NR

2

0

BIS

34 22 61

NR NR 7

11 16 22

0 2 0

BIS BIS BIS

Slavov et al, 200259 Kerssens et al, 2002 Kerssens et al, 200360 Russell, 200661 Bell et al, 200662 Kocaman Akbay et al, 200763 Andrade et al, 200864 Russell, 201345 Russell, 201345 Zand et al, 201465

Pediatric noncardiac surgery Major gynecology Major gynecology Cesarean section

Isoflurane/epidural Propofol/epidural Sevoflurane / nitrous oxide

0 0 9** * 5 NR

Auditory evoked response BIS BIS BIS BIS

NR = not reported; PRST = arterial Pressure, pulse Rate, Sweating and Tear score; EEG = electroencephalogram; BIS = bispectral index; *Both patients did not respond to verbal command during surgery; **All 3 patients had unverified response only; ***One of the 9 patients with awareness did not have response to command during surgery.

10

1.5

Monitoring Anesthetic Drug Delivery Given the limitations of clinical assessment, it has been proposed to monitor

anesthetic drug delivery as a surrogate measure for depth of anesthesia. The utility of this measure is based on two important assumptions. Firstly, it assumes that the dose of drug administered directly reflects on the amount of drug apply to the site of action (i.e. the brain). Secondly, the dose-response relationship is consistent between individuals and among different patient populations.

Much of the work has been performed on volatile agents. Eger and colleagues introduced the term “minimum alveolar concentration (MAC)” as an unified definition of anesthetic potency.66 In this regard, one MAC of volatile anesthetics is expected to prevent movement in 50% of patients after skin incision.66,67 Using a similar approach, Stoelting and co-workers defined MAC-awake as the minimum alveolar concentration required to prevent purposeful response to verbal command in 50% of patients.68 Based on these concepts, anesthesiologists are able to predict anesthetic depth with end-tidal alveolar gas measurements. The biggest advantage of MAC monitoring is that anesthetic gas analyzers are widely available for real-time feedback of anesthetic administration.

It should be noted that there are discrepancies between end-tidal and brain concentrations of volatile anesthetics. In soluble volatile anesthetics, such as isoflurane, equilibration may take > 10 min to complete. Therefore, during a step change in anesthetic delivery, the end-tidal concentration would be considerably different from that in the brain.69,70 There are also substantial variations in the measurements of MAC-awake. In studies of healthy patients or volunteers receiving

11

isoflurane and sevoflurane, the coefficients of variation for the MAC-awake values were 20.9% and 14.6%, respectively (Table 1.4).69-83

Table 1.4. Minimum Alveolar Concentration (MAC) and MAC-awake Anesthetic

Author, year

Desflurane

Chortkoff et al, 199581 Jones et al, 199076

Sevoflurane Katoh et al, 199369 Katoh et al, 199475 Katoh & Ikeda, 199883 Suzuki et al, 199873 Inomata et al, 199984 Goto et al, 200077 Kihara et al, 200072 Inomata et al, 200271 Davidson et al, 200878

Chen et al, 201479 Liang et al, 201474 Isoflurane

Dwyer et al, 199280 Gaumann et al, 199270 Katoh et al, 199369 Zbinden et al, 199482 Goto et al, 200077

MAC-awake to age adjusted MAC ratio 0.36 0.53*

No. of patients

Age (year)

MACawake

22 10

21-30 18-36

2.60 2.42

12 16 10 18 26 52 18 24 40 20 20 20 29 24

18-55 19-60 22-62 21-60 20-55 30-48 21-58 4-8 3-8 2.1-4.8 5.0-7.9 8.1-12.7 0.1-0.6 18-40

0.63 0.41 0.67 0.62 0.70 0.66 0.59 0.78 0.81 0.81 0.47 0.47 1.00 0.65

0.34 0.22† 0.36 0.34 0.37 0.34 0.35 0.39 0.36 0.30 0.20 0.20 0.29** 0.29

17 8 8 12 9 20 18

18-34 18-49 20-52 18-57 18-59 18-55 20-55

0.38 0.31 0.23 0.36 0.25 0.39 0.40

0.59 0.25 0.19† 0.31 0.22† 0.31 0.35

MAC was determined by *tetanic stimuli to the ulnar nerve or **tracheal intubation; †MACAwake determined by slow washout method.

12

In addition, a number of factors are known to affect anesthetic potency. For instances, advanced age, pregnancy and anemia reduce MAC to different extent, whereas chronic drug abuse increases MAC.85-89 Taken together, these factors contribute to the pharmacokinetic and pharmacodynamic variability, and may alter the effects for any given dose of anesthetics. Nevertheless, isoflurane, nitrous oxide and desflurane at concentrations of 1.5-2 times MAC-awake (or 0.7 MAC) are effective in preventing memory and recall of emotionally charged information during anesthesia.80,90,91 A protocol aiming to achieve end-tidal volatile concentration ≥ 0.7 MAC has been implemented in large clinical trials of awareness.92-95 Others have developed software, based on complex models, to display the predicted anesthetic concentration in the brain. Two of these displays - Navigator Applications Suite (GE Healthcare, Helsinki, Finland) and SmartPilot View (Dräger Medical, Lubeck, Germany), have been incorporated in anesthesia monitoring system.96 However, the performance of these displays to facilitate titration of anesthetic drug delivery has not been evaluated systematically.

1.6

Processed Electroencephalographic Monitoring Since anesthetic works primarily on the brain to produce loss of

consciousness, it seems logical to monitor brain activity as a surrogate measures of anesthetic depth. However, despite the introduction of human electroencephalographic (EEG) recording in 1929,97 little progress was made for the subsequent 50 years. It was not until the introduction of fast and powerful microprocessors, EEG recording was then becoming available for intraoperative monitoring during the past decade.98-100 In this respect, ongoing research has focused on the optimal method that

13

permits automated EEG analysis and coverts the complex waveform into a real-time descriptor of depth of anesthesia.

1.6.1 Time Domain Analysis Anesthetics produce suppressive effects on EEG.97,101-108 With the exception of ketamine and nitrous oxide, increasing dose of anesthetics slows the EEG signal from β wave (13-30 Hz) during the awake state to the much slower θ (4-7 Hz) and δ (0.5-3 Hz) waves while asleep. It is therefore not surprising that earlier algorithms analyzed EEG based on time domain. Zero crossing analysis was one of the earliest methods developed. It calculates the average time intervals between adjacent EEG waves of different polarities per segment (epoch).109 However, not all EEG signals cross the zero line, therefore zero crossing analysis generally underestimates EEG frequency. Other investigators have attempted to improve time domain analysis by adding the amplitude data to the calculations. Aperoidic analysis determines the average amplitudes of EEG signals in the fast (8-29.9 Hz) and slow (0.5-7.9 Hz) frequency bands.110 Therefore, the contributions of fast and slow waves in a series of EEG epochs could be displayed in a 3-D parallelogram (x axis - frequency, y axis amplitude and z axis - time). This technique has been incorporated in the Lifescan EEG monitor (Neurometrics, San Diego, CA).51 The Cerebral Function Monitor (CFM) displayed the amplitudes of EEG in the θ and δ range in a logarithm scale,111 and the Cerebral Function Analysing Motor (CFAM) extended the display to all frequency bands.112 Despite the simplicity, time domain analysis provides important measures for EEG signals. For instance, burst suppression ratio measures the percentage of electrical silence per EEG epoch and is an indicator of very deep anesthesia.107,113

14

1.6.2 Frequency Domain (Spectral) Analysis This analysis decomposes an epoch of EEG signals into a number of periodic functions using the fast Fourier transformation algorithm.104,105,107,114 This is displayed as a histogram comparing the power (i.e. square of amplitude) over a range of frequency, and is commonly known as the power spectrum of the EEG epoch (Figure 1.2). The EEG signal can then be summarized by a number of indices derived from the power spectrum. Common derivatives include the peak frequency (frequency with the largest power), median frequency (frequency that split the total power in half) and 95% spectral edge frequency (the frequency below which contains 95% of the total power). In addition, one can defined the power spectrum by comparing the relative power in each of the EEG frequency bands with the total power (e.g. relative δ power).114 It is unclear which of the EEG spectral derivatives would be superior to indicate anesthetic depth. Nonetheless, it is possible to superimpose arrays of power spectra from consecutive epochs and display the results in a waterfall plot (compressed spectral array, Figure 1.2), so that changes in the spectra over time could be observed. Alternatively, the amplitude (or power) data could be displayed as density of a dot (gray or color scale), and a contour map of changing power is generated (density spectral array).107 The major disadvantage of EEG spectral derivatives is that it undergoes a biphasic response to increasing dose of anesthetics (initial β activation followed by θ and δ slowing), therefore a single processed EEG value may indicate two distinct depths of anesthesia.83

15

Figure 1.2. Electroencephalogram (EEG) spectral analysis

An epoch of EEG is collected from leads Fp1 and Fz (upper panel). The power (i.e. amplitude2) of EEG waveform is plotted over frequency bands, using fast Fourier transformation algorithm (middle panel). Power spectra from consecutive EEG epochs are displayed as compressed spectral array (lower panel).

16

1.6.3 Bispectral Analysis The bispectral analysis is a higher order signal processing technique that quantify phase locking (quadratic coupling) of different frequency bands in the EEG signal.115 The analysis was first introduced in field of geophysics to explain the nonlinear characteristics of wave motion in nature.116 It is believed that brain waves may behave in similar fashion, and therefore bispectral analysis was applied to describe EEG signals in 1971.117 It should be clear that bispectral analysis is distinct from bispectral index (BIS, Coviden, Boulder, CO). In this regard, the bisepctrum expressed as the synchrony of fast and slow EEG waves (SyncFastSlow = log(B0.547Hz/B40-47Hz))

is only one of the three components (β ratio and burst suppression) in

the construction of the BIS number.99,100,118 More importantly, the BIS algorithm is based on a probability model that correlates between three derived components from frontal EEG and clinical status using a board range of anesthetic agents in >1,500 patients.118 Therefore, BIS is a linear index of likelihood to indicate a specific depth of anesthesia. Although the proprietary algorithm has not been disclosed, it has been shown that a BIS value < 30 is primarily a function of burst suppression in volunteers receiving propofol infusion.119

1.6.4 Spectral Entropy Analysis Spectral entropy is a measure of “predictability” of EEG signal based on the amplitude distribution (Shannon entropy), normalized to the total power, in any EEG epoch.120,121 Therefore, a regular sinusoidal wave with perfectly predictable amplitude in the subsequent wavelets will have an entropy value of zero. In contrast, the EEG amplitude of an awake individual is asynchronous and disorganized, the entropy value would be unity because the future waveform cannot be predicted. The spectral entropy

17

has been implemented in the GE Datex-Ohmeda Entropy™ Module (GE Healthcare, Milwaukee, WI). In this module, EEG within the frequency band of 0.8-32 Hz is captured for the calculation of state entropy (SE) to indicate the depth of anesthesia. The module also reports response entropy (RE) using a frequency band that covers electromyographic activity (0.8-47 Hz). An increase in RE indicates patient arousal with increased muscle activity.122 Spectral entropy is also used in Narcotrend (MonitorTechnik, Bad Bramstedt, Germany) as one of the components to classify EEG into different stages of anesthetic depth.123

Other methods to compute EEG entropy have been proposed. The HilbertHuang spectral entropy estimates instantaneous amplitude of component frequency where the underlying function of the wavelet has not been assumed.124 Approximate entropy accounts for the nonlinear nature of the EEG signals, but it requires a long segment of clean EEG signal for accurate calculation.125 The computational algorithm has been simplified to produce the permutation entropy.126 Currently, there are only limited data on the validity of these algorithms to detect consciousness during anesthesia.124-127

1.6.5 Topographic (Spatial) Analysis Topographic analysis evaluates EEG patterns between different regions of the brain. With increasing dose of anesthetics, there is an anterior shift of α and δ power from vertex to the frontal regions.128-131 There is also an increase in coupling in the frontopolar regions between hemispheres.131 Topographic difference is the primary component of the patient state index (PSI, Physiometrix, Billerica, MA).132,133 In the calculation of PSI, EEG spectral data are collected and analyzed from leads (in the

18

original PSArray sensor) along the vertex (Fpz, Cz, Pz) and at both frontal areas (Fp1, Fp2). Similar to BIS, these data are correlated with clinical states in a large pool of patients to derive a probability index of hypnosis during anesthesia. PSI is currently incorporated in the SedLine brain function monitor (Masimo, Irvine, CA). Interestingly, the new PSI sensor (PSArray2) has removed the electrodes over the vertex and posterior regions of the brain. It remains unclear how this change might affect the validity of PSI.

1.6.6 Other EEG Processing Algorithms Symbolic dynamic method assigns sequence of symbols to EEG segments. The alteration of symbols therefore describes the dynamic behavior and represents the EEG waveform. This algorithm, together with β ratio and burst suppression are the primary elements of the Index of Consciousness (IoC).134,135 Other algorithms that have been tried to describe EEG signals include detrended fluctuation analysis, cumulative power spectrum and auto-mutual information system. These algorithms evaluate the randomness of EEG signals.136-138 On the other hand, recurrence quantification analysis is a measure of “repeatability” of a dynamic sequence.139,140 Finally, the time encoded signal processing and recognition (TESPAR) algorithm aims to map the EEG waveform by 2 dimensional analysis. The time domain is quantified by zero crossing analysis. The features of the EEG signals within the epoch is translated into TESPAR alphabet based on the positions of the wavelets.141 In should be noted that the validity of these algorithms to describe the EEG signals remains limited and further investigations are required before they can be used to report anesthetic depth.

19

1.7

Auditory Evoked Potential Monitoring Apart from spontaneous EEG recordings, auditory evoked potential (AEP) has

been proposed as a measure of anesthetic depth. AEP measures the progress of auditory input as it travels from the cochlea into the cortical centers known to be associated with conscious perception. Following repeated clicks, a series of characteristic brain waves arising from the brainstem, the auditory radiation, the auditory cortex and its association areas can be recorded.142-145 The early brainstem (1-10 ms) and the late cortical (50-500 ms) components are resistant to anesthetics. However, the amplitude of the middle latency (10-50 ms) AEP typically decreases and its latency increases with anesthetic depth during various physiological conditions.99,146,147

Compared with EEG recordings, there are distinct advantages of AEP monitoring. Given that unintentional intraoperative recall is usually auditory in nature, it is intuitive to monitor awareness with AEP.16,17,41,148-150 Furthermore, surgical stimulation partly reverses the AEP changes associated with anesthetic administration, suggesting that it may be useful to indicate the balance between hypnosis and the intensity of stimulation.151 Finally, there is less overlap in the ranges of AEP associated with conscious and unconscious states.146 Hence, it is possible to assign a value that can ensure hypnosis in all patients but without causing hazardous side effects due to deep anesthesia.

The clinical application of AEP monitoring has however, been hindered by the long delay in signal acquisition and difficulty in complex waveform analysis. Using 256 sweeps, at least 36 seconds is required to produce a reliable AEP signal with a

20

click rate of 7.1 Hz. Kenny and co-workers used the moving average technique to expedite AEP signal analysis. In this method, AEP waveform is updated every 3 seconds.152,153 AEP morphology is reported by summing up the square roots of the amplitude difference between successive time points (0.56 ms apart). This is then normalized to produce the AEPIndex as a measure of anesthetic depth.153 Alternatively, an autoregressive model with exogenous input (ARX-model) has been proposed to shorten the delay (Figure 1.3).154 In this model, the number of sweeps required to produce a reliable AEP waveform is reduced from 256 sweeps in the conventional moving average technique to 18 sweeps. Therefore, it is possible to follow the rapid transition from hypnosis to consciousness. The resultant waveform is then transformed into a numeric index that describes the shape of the AEP signal, i.e. ARXderived AEP Index (AAI version 4.1, Danmeter, Odense, Demark).155,156 In clinical studies, AAI values between 15 and 25 indicates adequate hypnosis for surgery.155,156 In order to minimize interruption of AEP measurements with electrical and other sources of interference, a hybrid index has been proposed.157 When signal-to-noise ratio falls below 1.3, AAI values (version 4.2) are replaced by β ratio in the EEG power spectrum. The revised AAI values range from 0 to 100 and a value between 40 and 60 is recommended for adequate anesthesia.

21

Figure 1.3. Algorithm for autoregressive modelling of auditory evoked potential

Signal-to-noise ratio for AEP input is estimated by averaging an epoch of signal that is synchronized with auditory clicks (Ampsync) and then asynchronously using the same sweeps (Ampasync). Large asynchronous amplitude indicates noisy signal. Typical awake AEP signal is shown in the dotted box. A/D = analog to digital; AEP = auditory evoked potential; MTA = moving time average; ARX = autoregressive model with exogenous input; EMG = electromyography; BS = burst suppression; AAI = ARX-derived AEP Index

22

1.8

Monitoring Brain State during General Anesthesia Given the growing interests to monitor consciousness during anesthesia, a

number of manufacturers have produced their devices to measure depth of anesthesia. Table 1.5 summarizes the features of commercially available devices for monitoring brain state during general anesthesia. These devices are designed to collect EEG signals (spontaneous or evoked) from a specific electrode montage (typically frontal). EEG data are extracted using the abovementioned algorithms. These are then correlated with clinical status (e.g. response to verbal command) in a reference population using a statistical model to generate a dimensionless index of anesthetic depth. Therefore, these indices are statistical functions that represent the likelihood of present depth of anesthesia.118 For instance, a patient with a BIS value of 50 is extremely unlikely to be conscious during anesthesia, whereas return of consciousness is expected when BIS is well over 90. In the majority of studies, these devices are able to track the anesthetic effect during induction, when the depth of anesthesia is very deep and upon emergence of anesthesia.158 Obviously, the accuracy of these indices will depend on the quality of the calibration dataset. It should be noted that measurements of entropy and auditory evoked potential have not undergone probability modeling but the scales have been readjusted so that the values are comparable to other indices.

The advantage of the commercially available monitor is that it simplifies complex EGG signals into a single accessible number. When alarm settings are appropriately adjusted, it allows anesthesiologists to concentrate on the overall patient care and not to be distracted by spending time to interpret the constantly changing raw EEG waveforms.108 Using the depth of anesthesia index, anesthesiologists are able to

23

adjust anesthetic administration according to the prevailing anesthetic depth and the intensity of surgical stimulus.

As with all neurophysiologic monitoring devices, the monitors are subjected to the influences of environmental and physiologic factors.159-162 In this regard, electrical mains, movement and electrocautery artifacts produce high frequency signals and are common sources of interference to signal recordings.159,162-164 These factors should be carefully considered when interpreting the indices.163,165 Nevertheless, despite the inter-individual variability, brain state monitoring could provide useful information for tracking anesthetic drug effect.166,167

24

Table 1.5. Characteristics of the currently available monitors of anesthetic depth Parameters

Machine / Manufacturer

Consumable

Physiologic signals

Recommended range of values for anesthesia

Principles of measurement

Bispectral index (BIS)115

BIS Complete 2channel (or 4channel) monitor (Coviden, Boulder, CO)

BIS (4-electrode, bilateral, pediatric and extend) sensor

1-2 channel frontal EEG

40-60

BIS is derived from the weighted sum of three EEG parameters: (1) relative α/ß ratio (2) biocherence of the EEG waves (SyncFastSlow) and (3) burst suppression The relative contribution of these parameters has been tuned to correlate with the degree of sedation produced by various sedative agents. BIS ranges from 0 (asleep)-100 (awake).

Patient state index (PSI)132,133

SedLine brain state monitor (Masimo, Irvine, CA)

PSArray2 sensor

4 channels frontal EEG

25-50

PSI is derived from progressive discriminant analysis of several quantitative EEG variables that are sensitive to changes in the level of anesthesia, but insensitive to the specific agents producing such changes. It includes changes in: (1) power spectrum in various EEG frequency bands (2) hemispheric symmetry, and synchronization between brain regions and the inhibition of regions of the frontal cortex. PSI ranges from 0 (asleep)-100 (awake).

Narcotrend stage Narcotrend index123,168

Nacrotrend monitor (MonitorTechnik, Bad Bramstedt, Germany)

Ordinary ECG electrode

1-2 channels frontal EEG

35-65 (corresponds to stage D0-2 to C1)

The Narcotrend monitor classifies EEG signals into 15 stages of anesthesia (A=awake; B0–2=sedated; C0–2=light anesthesia; D0–2= general anesthesia; E0,1=general anesthesia with deep hypnosis; F0,1=burst suppression). The classification algorithm is based on a discriminant analysis of entropy measures and EEG spectral variables. More recently the monitor converts the Narcotrend stages into a dimensionless number from 0 (asleep) to 100 (awake) by nonlinear regression.

State and Response Entropy122

GE Datex-Ohmeda Entropy Module (GE Healthcare, Milwaukee, WI)

Entropy sensor

1 channel frontal EEG

40-60

Entropy described the “irregularity” of EEG signal. Entropy module calculates spectral entropy of the EEG spectrum. Two spectral parameters are calculated: (1) State entropy (SE, frequency band 0-32 Hz) and (2) Response entropy (RE, 0-47 Hz) also includes muscle activity SE have been re-scaled, so that 0 is asleep and 91 is awake, while the range for RE is 0-100.

25

AEPIndex153

aepEX PLUS (Medical Device Management, Braintree, Essex, UK)

Ordinary ECG electrode

AEP (frontomastoid)

40-60

aepEX PLUS measures middle latency AEP (0-144 ms). The waveform is updated by moving average technique. AEPIndex has been scaled and ranges from 0-100.

Aline Autoregressive Index (AAI version 4.1)154,155

AAI monitor (Danmeter A/S, Odense, Demark)

Ordinary ECG electrode

AEP (frontomastoid)

15-25

(AAI version 4.2)157

AEP/2 monitor (Danmeter A/S, Odense, Demark)

Ordinary ECG electrode

AEP (frontomastoid)

40-60

AAI is derived from the middle latency AEP (20-80 ms). AAI is extracted from an autoregressive model with exogenous input (ARX–model) so that only 18 sweeps are required to reproduce the AEP waveform in 2-6 s. The resultant waveform is then transformed into a numeric index (0-100) that describes the shape of the AEP. AAI > 60 is awake, AAI of 0 is deep anesthesia. Incorporating EEG data (β ratio) when signal-to-noise ratio for AEP measurement is < 1.3. The revised AAI has been rescaled, so that the range of values indicating adequate anesthesia are comparable to other devices.

Cerebral state index (CSI)169

Cerebral state monitor (CSM), Danmeter A/S, Odense, Demark

Ordinary ECG electrode

1 channel frontomastoid EEG

40-60

CSI is a weighted sum of (1) α ratio, (2) β ratio, (3) difference between the two and (4) burst suppression. It correlates with the degree of sedation by adaptive neuro-fuzzy inference system. CSI ranges from 0 (asleep) to 100 (awake).

Index of Consciousness (IoC)135

IoC-View monitor (Morpheus Medical, C/ Llacuna, Spain)

Ordinary ECG electrode

1 channel frontal EEG

40-60

Symbolic dynamic method is used to encode EEG signals. Other components include β ratio and burst suppression to indicate light and deep anesthesia. The IoC value is obtained by correlating these EEG parameters with clinical level of consciousness using a discriminatory function. IoC ranges from 0 (asleep) to 99 (awake).

qCON170

qCON 2000 monitor (Quantium Medical, Mataró, Barcelona, Spain) SNAP II monitor (Stryker Instruments, Kalamazoo, MI)

Ordinary ECG electrode

1 channel frontal EEG

40-60

qCON calculates the spectral ratios at 4-8 Hz, 8-13 Hz, 11-22 Hz and 33-44 Hz with the total spectrum and correlates with the clinical states in 1,110 subjects using adaptive neuro-fuzzy inference system. qCON ranges from 0 (asleep) to 99 (awake).

Ordinary ECG electrode

1 channel frontal EEG

40-60

SNAP index compares spectral parameters at 80-120 Hz with that in 0.1-18 Hz. SNAP index ranges from 0-100.

SNAP Index171

EEG = Electroencephalogram; AEP = Auditory evoked potential

26

1.9

Clinical Utility of Brain State Monitoring during Anesthesia The primary purpose of brain state monitoring is to guide anesthetic dosing.

Based on the derived indices, anesthesiologist aims to titrate anesthetic administration so as to prevent periods of under-dosing and to avoid autonomic stimulation and unintentional awareness. Similarly, monitoring brain state facilitates anesthetic delivery to avoid periods of overdose and to minimize anesthetic exposure. This should expedite recovery and may contribute to better outcome. Being the first in the market, most research has evaluated the impact of BIS monitoring on patient outcome. The following discussion is largely limited to studies using BIS only.

1.9.1 Recovery Times In a meta-analysis of 20 studies involving a total of 2,557 patients, BIS-guided anesthesia reduced propofol infusion rate by 17.8% (95% CI: 9.9-25.8%) and decreased volatile administration by 18.4% (95% CI: 3.3-30.7%), compared with standard patient care.172,173 This was associated with a decrease in time to tracheal extubation and recovery room discharge by 2.6 (95% CI: 1.8-3.5) min and 6.8 (95% CI: 2.3-11.2) min, respectively.172,173 Although similar results were shown for other devices (entropy: 6 trials, 695 patients) and (Narcotrend: 2 trials, 124 patients), the sample size was relatively small and the data were limited.173 More importantly, it is currently unclear whether the enhanced recovery times could lead to better patient satisfaction or improved quality of recovery.

27

1.9.2 Awareness Other studies have evaluated the effect of BIS monitoring on awareness during anesthesia. In an international multicenter study, the Bispectral Index Monitoring to Prevent AWAREness During Anesthesia (B-Aware) trial randomized 2,463 patients at high-risk of awareness to receive either BIS monitoring or routine care.174 BISguided anesthesia reduced the incidence of awareness compared with routine care (2/1,225 versus 11/1,238), the relative risk reduction was 82% (95% CI: 17-98%) and the number needed to treat was 138 (95% CI: 77-641). There is however ongoing debate as to whether other techniques, such as monitoring of anesthetic delivery, targeting an end-tidal volatile concentration of MAC 0.7-1.394,95 or ≥ 0.593 would produce comparable protection.

1.9.3 Other Outcomes Emerging evidence suggests that anesthetic may produce toxic effect to the brain leading to postoperative delirium and cognitive dysfunction.175-177 Given that BIS monitoring reduces anesthetic exposure,172,173 it is plausible that BIS-guided anesthesia would improve cognitive outcomes. Others have explored the association between deep anesthesia and long-term morbidity and mortality.178-184 However, it remains unclear whether the poor outcome is related to deep anesthesia, or the low BIS value is simply a marker of underlying fragility.

1.10

Conclusions In this chapter, measures to monitor anesthetic depth are reviewed. A large

body of evidence suggests that monitoring brain state using processed EEG or AEP could be used as a surrogate measure of anesthetic depth. However, the incorporation

28

of these devices into routine anesthetic care would require clear demonstration of clinical utility and benefits.185 In two advisory statements published by the American Society of Anesthesiologists and National Institute for Health and Clinical Excellence,186,187 it was clearly noted that there were insufficient data to produce definitive recommendations for the routine use of brain state monitors. In this thesis, clinical trials, systematic reviews and meta-analyses were conducted to determine the impact of brain state monitoring on patient recovery. It is hoped that new knowledge could be generated to guide the use of these monitors for better patient outcomes.

29

CHAPTER 2

2.1

Study Designs, Hypotheses and Objectives Traditional anesthetic practice relies heavily on indirect measures, such as

somatic and autonomic responses, to titrate anesthetic delivery. Recent development in neurophysiologic monitoring allows anesthesiologists to measure the underlying depth of anesthesia using commercially available brain state monitors.100 In this thesis, four studies were carried out to determine the effectiveness of these monitors in contemporary clinical anesthesia. Using different study designs and patient populations, this thesis examined two monitoring devices, viz. AEP and BIS. Specifically, the impact of brain state monitoring on quality of recovery, awareness during anesthesia, postoperative cognitive dysfunction, delirium and long-term mortality and morbidity (i.e. myocardial infarction and stroke) were studied. The series of studies provide a comprehensive assessment of on the utility of brain state monitoring during contemporary general anesthesia. The purpose of this chapter is to outline the design, hypothesis and objective of each of these studies.

2.2

Rate and Quality of Recovery The AEP trial was a prospective randomized, 2-by-2 factorial trial on the

impact of AEP monitoring and choice of anesthetic agent (propofol versus sevoflurane) in 1,063 patients undergoing noncardiac surgery that lasted > 1 hour. The primary outcome was the quality of recovery score recorded regularly up to 30 days after surgery. The hypothesis was that anesthesia guided by AEP monitoring would permit precise titration of anesthetic drugs. This should result in a decrease in drug consumption, allowing better quality of recovery, shorter recovery times and fewer 30

perioperative complications. The objectives were to determine the incidence of anesthetic complications, rate and quality of recovery after AEP-guided anesthesia compared with routine care. We also compared outcome from propofol or sevoflurane anesthesia.

2.3

Awareness during Anesthesia This was a systematic review and meta-analysis on the effect of brain state

monitoring to prevent awareness during anesthesia in 6 randomized controlled trials (RCT), involving over 34,000 patients undergoing a variety of surgical procedures. The meta-analysis also explored whether the use intravenous compared with volatilebased anesthesia, cardiac or noncardiac operations and the use of active controls would modify the risk of awareness.

2.4

Postoperative Cognitive Dysfunction This RCT evaluated the impact of BIS-guided anesthesia on cognitive function

and postoperative delirium in 921 elderly (≥ 60 years) patients having major noncardiac surgery. The alternative hypothesis was that BIS monitoring avoids excessively deep anesthesia, and would therefore improve postoperative cognitive performance. The trial also determined the risk factors for cognitive dysfunction and delirium after major surgery.

2.5

Long-term Mortality and Morbidity In this prospectively-planned, follow-up study of 536 high-risk patients having

major surgery in the B-Aware trial, the impact of BIS-guided anesthesia on long-term

31

mortality and serious morbidity was specifically sought. In a post hoc analysis, the risk of deep anesthesia on myocardial infarction and stroke was also evaluated.

32

CHAPTER 3

3.1

Quality of Recovery After AEP-Guided Anesthesia With rapidly increasing healthcare expenditure, anesthesiologists aim to

facilitate fast track surgery by expediting emergence, improving functional status and minimizing complications after anesthesia.188-192 Brain state monitoring has been reported to assist anesthetic drug titration. In two recent systematic reviews, the use of brain state monitoring reduced anesthetic drug consumption by 17.8-18.4%, resulting in an accelerated recovery.172,173 These findings were however limited because the sample size of these studies were relatively small, ranging from 10 to 125 patients per group. These studies were largely confined to ambulatory surgery or other homogenous patient populations, and were restricted to the study of early recovery times, such as time to awakening, tracheal extubation or discharge from recovery room. More importantly, these studies have focused on the evaluation of one specific brain state monitor - BIS (Coviden, Boulder, CO).

AEP measures the summated EEG signals after delivery of auditory clicks. The middle latency AEP undergoes characteristic changes with increasing doses of anesthetics.144 Compared with BIS, there is less overlap in the changes of AEP associated with wakefulness and unconscious state.146 It is therefore plausible to assign an AEP value that can ensure hypnosis but without causing hazardous side effects related to deep anesthesia.

The purpose of this trial was to determine the impact of intraoperative AEP monitoring, as compared with routine anesthetic care, on the rate and quality of 33

recovery after non-ambulatory elective surgery. In addition, the influence of propofol infusion versus sevoflurane-based anesthesia on perioperative outcome was also evaluated.

3.2

Methods

3.2.1 Study design The AEP Monitoring to Improve Recovery from General Anesthesia trial (The AEP trial) was a prospective, randomized, controlled trial using a 2-by-2 factorial design. The trial objectives, study design and study methods described in this chapter are consistent with those reported at the Chinese Clinical Trial Registry (ChiCTRTRC-05000657). The candidate was the principal investigator of the trial, responsible for the design of the study, funding application, data collection, analysis and interpretation.

3.2.2 Patient Population Three centers (North District Hospital, Prince of Wales Hospital, and Tuen Mun Hospital) participated in the AEP trial. Patients were eligible for the study if they were aged 18 years or older, and were scheduled to undergo noncardiac, nonneurosurgical surgery that was anticipated to last for at least 1 hour. Patients were excluded if they were planned to receive postoperative ventilation, had severe hearing deficit, major psychosis, language difficulties, memory problems, known hypersensitivity or allergy to propofol or sevoflurane. Patients who, for other reasons, were not expected to be available for, or to cooperate with, postoperative interview were also excluded.

34

3.2.3 Procedures The protocol was approved by the Clinical Research Ethics Committee of the Chinese University of Hong Kong (CRE-2001.327) and the Hospital Authority (PD/08/20/01). After obtaining written informed consents, patients were randomly allocated to receive either AEP monitoring or routine anesthetic care. In addition, patients were also randomly assigned to receive either propofol infusion or sevoflurane as their primary anesthetic regimen. Group allocation was distributed in a 1:1:1:1 ratio. Randomization codes were obtained from an interactive web-based system, stratified by centers.

Preoperative preparation was performed in all patients according to local hospital guidelines. In the operating room, following establishment of standard monitoring, silver/silver chloride electrodes were placed over the middle of forehead (Fpz) and the left mastoid (A1), according to the international 10-20 system, for AEP recording in all patients. Another electrode was placed over left side of forehead (Fp1) as reference (Figure 3.1). Immediately prior to induction of anesthesia, a headphone was applied to deliver clicks (70 dB) of 2 ms duration, at 9 Hz using the AAI monitor (Danmeter, Odense, Demark). The resultant AEP was extracted by an autoregressive model with exogenous input (ARX) to produce an ARX-derived AEP index (AAI, Danmeter, Odense, Demark).154-156 AAI was updated every 2-6 seconds. In patients allocated to the AEP-guided anesthesia group, AEP waveform and the AAI trend were displayed, so that anesthetic administration (propofol or sevoflurane) was adjusted to maintain an AAI value between 15 and 25. Alarm limits were set to alert attending anesthesiologists when the AAI value was deviated from the protocol. In patients allocated to the routine care group, AEP was also measured and downloaded for

35

subsequent analysis, but the AEP waveform, AAI value and its trend plot were blinded on the monitor screen through a specific software, designed for the purpose of this trial (Figure 3.1).

Figure 3.1. Electrode montage and screen display for AEP monitoring

Electrode montage [Fpz (forehead), Fp1 (left of forehead) and A1 (left mastoid)] for AEP monitoring (upper panel). Screen displays for AEP (left panels) and routine care (right panels) groups. Note the absence of AEP waveform in the routine care group.

36

Anesthetic delivery in the routine care group was adjusted according to clinical response, aiming to maintain systolic arterial pressure within 15 mmHg of baseline and heart rate between 40 and 90 beats/min range. Dose of anesthetics were also increased if there were sweating, flushing, movement or swallowing. Cessation of anesthesia was timed to facilitate emergence as early as possible. Choice of airway management, perioperative analgesia and administration of intravenous fluid were left to the discretion of the attending anesthesiologists.

A manual or target-controlled infusion device was used to deliver propofol. The effect site propofol concentration was estimated from the infusion record using the Computer Control Infusion Pump (CCIP) program.* Sevoflurane dose requirement was expressed as age-adjusted MAC value.193

At each participating center, patients were interviewed by research personnel at regular intervals until hospital discharge. Patients were phoned again at 30 days after surgery to enquire about their current health status using the short-form health survey (SF-36).194-198

3.2.4 Trial Outcome The attending anesthesiologists were not blinded to the treatment allocation for safe patient management during surgery. However, the patients, surgical teams and the research personnel responsible for all outcome assessments were blinded to

*The

Computer Control Infusion Pump (CCIP) program is available for non-

commercial use at http://www.cuhk.edu.hk/med/ans/softwares.htm

37

the treatment allocation. The primary outcome was the change in quality of recovery (QoR) measured by the Chinese QoR score199 over the first 30 days after surgery. The psychometric properties of the Chinese QoR score are reported in Appendix 1. Secondary outcomes included recovery times from the cessation of anesthetic administration to eye opening, respond to verbal command, tracheal extubation, recovery room discharge (defined as when Aldrete score ≥ 9),200 ambulation, resume diet, hospital discharge and back to routine (defined as any of the followings: take care of self, take care of others, prepare meals, readings, do household duties and run errands).201 Incidence of postoperative nausea and vomiting (PONV) was recorded at 0, 6, 24 and 48 hours after surgery. Postoperative nausea was rated using a 10-cm visual analog scale (0 = no nausea; 10 = worst imaginable nausea). Significant nausea was defined as a score > 4. Retching or vomiting separated by ≥ one minute was considered as separate episodes.

Postoperative complications were also recorded. These included myocardial infarction (as defined by the universal definition),202 arrhythmias that required treatment (either pharmacological or electrical), stroke (defined as a new focal neurological deficit due to vascular causes that lasted ≥ 24 hours or leading to death), acute renal failure (defined as a rise of postoperative plasma creatinine concentration > 200 μmol/L) and infective complications involving the surgical wound (defined as the presence of purulent discharge, with or without a positive microbial culture), pneumonia (defined as the development of radiological infiltrate with fever > 38C or leukocytosis, or positive sputum or blood cultures), and infections from other sites.203

38

3.2.5 Statistical analysis and sample size calculation Statistical analyses were performed using Stata 13 (StataCorp LP, College Station, TX). Data were analyzed according to the principle of “intention to treat”. Differences in postoperative QoR scores were compared among groups using generalized linear models for repeated measures. Covariates included anesthetic regimens (sevoflurane versus propofol), perioperative characteristics and test for interactions. Multiple comparisons were adjusted by Dunn-Sidak procedure. Data that were not normally distributed were compared among groups using Kruskal-Wallis test. Recovery times were calculated by Kaplan-Meier analysis and compared among groups using log-rank test. Rates of perioperative adverse events were compared among groups using χ2 test. A p value of < 0.05 was considered statistically significant. The total sample size was determined based on a 15% change in the QoR score among groups. A total of 1,060 patients (i.e. 265 patients per group) would provide 92% power at two-sided α value of 0.05.

3.3

Results

3.3.1 Baseline characteristics The AEP trial enrolled a total of 1,063 (536 in the AEP-guided group and 532 in the routine care group) patients. Among these patients, 532 (50.4%) had sevoflurane and 531 (49.6%) had propofol infusion during anesthesia. All patients completed 30-day follow-up (Figure 3.2). Preoperative patient characteristics are shown in Table 3.1. On average, patients aged 53.1 years, 61.8% were females and 87.1% were classified as American Society of Anesthesiologists (ASA) physical status 1 or 2. Overall, 67.5% of patients had abdominal surgery for gynecologic or general surgical procedures.

39

Figure 3.2. Enrollments in the AEP trial

40

3.3.2 Anesthetic delivery with AEP monitoring During surgery, there was no difference in the amount of opioid given and administration of antiemetic agents between groups (Table 3.2). AEP monitoring however reduced anesthetic dosage significantly. In this respect, estimated effect site propofol concentration and end-tidal sevoflurane concentration were reduced by 16.1% and 29.3%, respectively. Consequently, the time-averaged AAI value in the AEP-guided group (20 ± 11) was significantly higher than that in the routine care group (9 ± 7), p < 0.001. The duration for which AAI < 15 was also longer in the routine care group compared with the AEP-guided group, p = 0.012. The lower anesthetic doses in the AEP-guided group resulted in faster emergence from anesthesia. The emergence times from cessation of anesthesia to eye opening, tracheal extubation and recovery room discharge were earlier in patients receiving AEPguided anesthesia than those in the routine care group (p < 0.001, log-rank test, for all three endpoints, Figure 3.3). However, emergence was not affected by the type of anesthetics (sevoflurane versus propofol) used (p > 0.495 for all interaction tests).

41

Table 3.1. Patient characteristics of the AEP trial

No. of patients Age, year Male sex, no. (%) Weight, kg Height, cm Current smokers, no. (%) ASA status, no. (%) 1 2 3 Pre-existing medical conditions, no. (%) Cardiovascular Respiratory Endocrine Hepatorenal Others Preoperative Chinese QoR score Median (Interquartile range) Surgical types, no. (%) Gynecology Orthopedics Urological General surgery Others Extent of surgery, no. (%) Minor/Intermediate Major/Ultra-major Duration of surgery, hours Median (Interquartile range)

AEP-guided anesthesia Sevoflurane Propofol Total 267 266 533 53 ± 15 54 ± 16 53 ± 15 110 (41.2%) 106 (39.8%) 216 (40.5%) 58 ± 11 59 ± 10 59 ± 11 160 ± 6 159 ± 7 159 ± 7 41 (15.4%) 30 (11.3%) 71 (13.3%)

Routine anesthetic care Sevoflurane Propofol Total 265 265 530 54 ± 17 53 ± 14 53 ± 16 113 (42.6%) 103 (38.9%) 216 (40.8%) 58 ± 11 58 ± 11 58 ± 11 160 ± 6 159 ± 6 159 ± 6 40 (15.1%) 37 (14.0%) 77 (14.5%)

112 (41.9%) 140 (52.4%) 15 (5.6%)

122 (45.9%) 128 (48.1%) 16 (6.0%)

234 (43.9%) 268 (50.3%) 31 (5.8%)

101 (38.1%) 147 (55.5%) 17 (6.4%)

106 (40%) 139 (52.5%) 20 (7.5%)

207 (39.1%) 286 (54.0%) 37 (7.0%)

84 (31.5%) 15 (5.6%) 38 (14.2%) 36 (13.5%) 58 (21.7%) 17.4 ± 0.9 18 (16-18)

72 (27.1%) 13 (4.9%) 45 (16.9%) 38 (14.3%) 46 (17.3%) 17.5 ± 0.8 18 (17-18)

156 (29.3%) 28 (5.3%) 83 (15.6%) 74 (13.9%) 104 (19.5%) 17.4 ± 0.9 18 (16-18)

91 (34.3%) 20 (7.5%) 45 (17.0%) 40 (15.1%) 59 (22.3%) 17.5 ± 0.9 18 (17-18)

86 (32.5%) 16 (6.0%) 41 (15.5%) 37 (14.0%) 51 (19.2%) 17.4 ± 0.9 18 (16-18)

177 (33.4%) 36 6.8%) 86 (16.2%) 77 (14.5%) 110 (20.8%) 17.5 ± 0.9 18 (16-18)

85 (31.8%) 37 (13.9%) 31 (11.6%) 95 (35.6%) 19 (7.1%)

88 (33.1%) 40 (15.0%) 28 (10.5%) 91 (34.2%) 19 (7.1%)

173 (32.5%) 77 (14.4%) 59 (11.1%) 186 (34.9%) 38 (7.1%)

86 (32.5%) 37 (14.0%) 30 (11.3%) 94 (35.5%) 18 (6.8%)

84 (31.7%) 38 (14.3%) 28 (10.6%) 95 (35.8%) 20 (7.8%)

170 (32.1%) 75 (14.2%) 58 (10.9%) 189 (35.7%) 38 (7.5%)

42 (15.6%) 227 (84.4%) 3.5 ± 2.0 3.2 (2.2-4.7)

43 (16.0%) 225 (84.0%) 3.6 ± 2.1 3.2 (2.0-4.8)

85 (15.8%) 452 (84.2%) 3.6 ± 2.1 3.2 (2.1-4.7)

37 (14.0%) 228 (86%) 3.7 ± 1.9 3.3 (2.3-4.8)

49 (18.5%) 216 (81.5%) 3.6 ± 2.1 3.1 (2.0-4.6)

86 (16.2%) 444 (83.8%) 3.6 ± 2.0 3.3 (2.1-4.6)

ASA = American Society of Anesthesiologists; QoR = Quality of Recovery

42

Table 3.2. Drug delivery and anesthetic depth in AEP trial AEP-guided anesthesia Sevoflurane Propofol Total 267 266 533 1.9 ± 0.9 1.8 ± 1.0 1.8 ± 0.9

Routine anesthetic care Sevoflurane Propofol Total 265 265 530 1.9 ± 1.1 1.9 ± 0.9 1.9 ± 1.0

107 ± 76

103 ± 106

105 ± 117

106 ± 75

107 ± 78

107 ± 74

25 (9.4%) 8 (3.0%) 2 (0.7%)

20 (7.5%) 9 (3.4%) 2 (0.8%)

45 (8.4%) 17 (3.2%) 4 (0.8%)

28 (10.6%) 10 (3.8%) 3 (1.1%)

21 (7.9%) 8 (3.0%) 2 (0.8%)

49 (9.2%) 18 (3.4%) 5 (0.9%)

Estimated effect site propofol concentration, µg/ml End-tidal sevoflurane concentration, MAC Nitrous oxide use, no. (%) End-tidal concentration

̶

2.40 ± 1.61

̶

̶

2.86 ± 1.32

̶

0.0003

0.65 ± 0.27

̶

̶

0.92 ± 0.38

̶

̶

15 min that required sympathomimetic agent or atropine; ‡Clinically significant hypertension was defined as systolic arterial pressure >160 mm Hg for > 15 min that required treatment; ‡‡Clinically significant tachycardia was defined as heart rate >100 beats/min for > 15 min that required treatment.

43

Figure 3.3. Emergence time from anesthesia

Sevoflurane

1.0

1.0

0.8

0.8

Proportion of patients opening eyes to command

Proportion of patients opening eyes to command

Propofol

0.6

0.4 AEP guided Routine Care

0.2

0.6

0.4 AEP guided Routine Care

0.2

0.0

0.0 0

5

10

15

20

25

30

0

1.0

1.0

0.8

0.8

0.6

0.4 AEP guided Routine Care

0.2

10

15

20

25

30

0.6

0.4 AEP guided Routine Care

0.2

0.0

0.0 0

5

10

15

20

25

30

0

Time after cessation of anesthetics (min)

5

10

15

20

25

30

Time after cessation of anesthetics (min)

1.0 Proportion of patients discharged from recovery room

1.0 Proportion of patients discharged from recovery room

5

Time after cessation of anesthetics (min)

Proportion of patients with tracheal extubation

Proportion of patients with tracheal extubation

Time after cessation of anesthetics (min)

0.8

0.6

0.4

0.2

AEP guided Routine Care

0.0

0.8

0.6

0.4

0.2

AEP guided Routine Care

0.0 0

1

2

Time after cessation of anesthetics (hour)

3

0

1

2

3

Time after cessation of anesthetics (hour)

Time to eye opening (upper panels), tracheal extubation (middle panels) and recovery room discharge (lower panels) after cessation of anesthesia. Data for sevoflurane and propofol infusion are shown in the left and right panels, respectively.

44

3.3.3 Outcomes after AEP monitoring Figure 3.4 shows the changes of QoR scores in the postoperative period. Patients receiving AEP-guided anesthesia rated their quality of recovery higher than that in the routine care group (p < 0.001). Interestingly, there was no impact of anesthetic agent used on the changes of postoperative QoR scores (p = 0.539).

The median (interquartile range) duration of hospital stay in patients receiving AEP-guided anesthesia was 6.4 (4.0-8.2) days and was shorter than that in the routine care group, 7.0 (4.5-11) days [hazard ratio (95% CI): 0.35 (0.29-0.42), p < 0.001, logrank test]. Patients in the AEP-guided group also resumed diet earlier than the routine care group [hazard ratio (95% CI): 0.57 (0.48-0.69), log-rank test, p < 0.001, Figure 3.5]. There was no impact of sevoflurane or propofol on hospital discharge or resumption of diet (p > 0.617 for both interactions). However, time to resumption of daily activities was not different among groups (p = 0.589).

45

Figure 3.4. Quality of recovery after AEP-guided or routine care anesthesia

Propofol

16

16

14

14

12 10 8 6 4

12 10 8 6 4

Routine care AEP guided

2

Sevoflurane

18

Chinese QoR score

Chinese QoR score

18

Routine care AEP guided

2

0

0 0

5

10

15

20

Day after surgery

25

30

0

5

10

15

20

25

30

Day after surgery

46

Figure 3.5. Late recovery times from surgery

Sevoflurane

1.0

1.0

0.8

0.8 Proportion of patients tolerating diet

Proportion of patients tolerating diet

Propofol

0.6

0.4

0.2

0.6

0.4

0.2

AEP guided Routine Care

0.0

AEP guided Routine Care

0.0 0

2

4

6

8

10

12

14

0

1.0

1.0

0.8

0.8

0.6

0.4 AEP guided Routine Care

0.2

4

6

8

10

12

14

0.6

0.4 AEP guided Routine Care

0.2

0.0

0.0 0

5

10

15

20

25

30

0

Time after cessation of anesthetics (days)

1.0

1.0

0.8

0.8

0.6

0.4 AEP guided Routine Care

0.2

5

10

15

20

25

30

Time after cessation of anesthetics (days)

Proportion of patients resumed daily activities

Proportion of patients resumed daily activities

2

Time after cessation of anesthetics (day)

Proportion of patients discharged from hospital

Proportion of patients discharged from hospital

Time after cessation of anesthetics (day)

0.6

0.4 AEP guided Routine Care

0.2

0.0

0.0 0

5

10

15

20

25

Time after cessation of anesthetics (days)

30

0

5

10

15

20

25

30

Time after cessation of anesthetics (days)

Time to first diet resumption (upper panels), hospital discharge (middle panels) and resumption of daily activities (lower panels) after surgery. Data for sevoflurane and propofol infusion are shown in the left and right panels, respectively.

47

There was a lower rate of PONV in the AEP-guided group compared with routine care (47.9% versus 19.7%, p < 0.001). It should be noted that propofol reduced the incidence of nausea and vomiting [odds ratio (95% CI): 0.28 (0.21-0.37), p < 0.001] compared with sevoflurane, regardless whether AEP monitoring was used. Other secondary outcomes are listed in Table 3.3. The rate of infection was also lower in patients receiving AEP-guided anesthesia compared with controls. These complications were not trivial. The median (95% CI) hospital stay was 6.0 (5.6-6.4) days for patients with no major complication, and 12.5 (10.6-13.4) days for patients with at least one major complication [hazard ratio (95% CI): 2.87 (2.38-3.45), p < 0.001].

At 30 days after surgery, patients in the AEP-guided group reported higher SF-36 scores compared with those in the routine care group (Table 3.3). This was primarily attributed to the higher scores in the general health perception and mental state index. Sevoflurane or propofol anesthesia did not impact on general health status (p = 0.845).

48

Table 3.3. Secondary outcomes after AEP-guided and routine care anesthesia AEP-guided anesthesia Sevoflurane Propofol Total 267 266 533

Routine anesthetic care Sevoflurane Propofol Total 265 265 530

58 (21.7%) 65 (24.3%)

35 (13.2%) 40 (15.0%)

83 (15.6%) 105 (19.7%)

137 (51.7%) 176 (66.4%)

66 (24.9%) 78 (29.4%)

143 (27.0%) 254 (47.9%)

99.9% of patients (see Section 1.5: Monitoring Anesthetic Drug Delivery).85 The subgroup analysis confirmed this hypothesis showing no effect of BIS monitoring on the risk of awareness when compared with an active control group using ETAG as an alternative target. However, the use of ETAG has some limitations. Firstly, it cannot be used in total intravenous anesthesia. In these patients, our meta-analysis showed that they were at higher risk of awareness. Secondly, significant hemodynamic change may preclude the delivery of anesthetic, even at the lower limit of ETAG target (0.5 MAC).

It was hypothesized that BIS monitoring may be more effective in patients having cardiac surgery. Cardiac surgical patients have limited physiological reserves. Anesthetic delivery in these patients are often restrained because of hypotension associated with many anesthetic drugs. The pooled analysis did not confirm this hypothesis. In fact, BIS monitoring appeared to be more effective in noncardiac surgery. It should be noted that the majority of the noncardiac surgical patients were

69

derived from trials that included routine anesthetic care, i.e. adjusting anesthetic dosage according clinical signs to indicate autonomic and somatic responses. The higher incidence of awareness in these patients may maximize the difference from BIS-guided anesthesia.

4.4.1 BIS monitoring to prevent awareness The result of this meta-analysis extends the findings in previous studies showing the capability of BIS in tracking level of hypnosis.100,205,244 By targeting a range of BIS values (usually between 40 and 60), anesthesiologists are able to detect inadequate anesthesia during intense surgical stimulation. Similarly, anesthesiologists may avoid excessive anesthetic doses by limiting unnecessary deep level of anesthesia.

Awareness may occur during BIS monitoring (total of 23 events in our metaanalysis). In these patients, awareness generally occurred when BIS was at the upper limit of the recommended range (i.e. ≥ 55-60). This represents the underlying variation of BIS. Furthermore, given that it is a probability function of hypnosis, it would not be surprising that a small proportion of patients may become aware of intraoperative events at the upper BIS limit (i.e. outliers).

The number needed to treat for BIS monitoring to prevent awareness was 843 (95%CI: (472-3,874)). This may appear to be clinically irrelevant. However, considering the incidence of awareness ranges from 1 to 2 per 1,000,41,230 the use of BIS monitoring in a 1,000 patients to prevent one episode of awareness may be considered as effective. It should be clear that awareness is associated with

70

devastating long-term psychological consequences.234-237 A monitor to prevent awareness may therefore be useful, especially to those who have experienced awareness in previous anesthesia.

4.4.2 Strengths and limitations This systematic review has strength in that a comprehensive search has been performed. The included trials, except one,243 were of high quality with low risk of reporting bias. The data were also confirmed by original authors. There is however a major limitation. A substantial proportion of patients could not be interviewed in the follow-up period. Many of these patients had significant complications resulting in impaired postoperative consciousness or requiring prolonged sedation in the intensive care unit. It remains unclear how the missing patients may influence the results of the included trials.

There are also issues related to the low incidence of awareness and the robustness of statistics. In this respect, a few additional events in the BIS treatment group will render the p value insignificant. The problem is known as fragility of trials. The smallest number of events that changed the p value from < 0.05 to one that is ≥ 0.05 is known as the fragility index.245 In this systematic review, the fragility index for Puri’s trial was one,243 and that for the B-Aware trial174 and Zhang’s trial231 were 3 and 8, respectively. One should be cautious when interpreting these trials.

4.4.3 Conclusions In conclusion, the results of this systematic review and meta-analysis suggest that BIS monitoring is effective in preventing awareness after general anesthesia,

71

especially in noncardiac surgical patients, and those receiving total intravenous anesthesia. Further study is required to define the optimal range of BIS value to prevent awareness. More studies will also be needed to demonstrate the effectiveness of other monitors to prevent awareness.

72

CHAPTER 5 5.1

Cognitive Outcome after BIS-Guided Anesthesia It is commonly believed among clinicians that anesthetic effects such as

unconsciousness, amnesia, and areflexia are transient phenomena. Therefore, patients who receive a general anesthetic should recover completely as the drugs are eliminated. However, there are now strong evidence from animal experiments to suggest that routine doses of common anesthetics induce profound neurochemical changes that persist for weeks or months after anesthetic exposure.175,176,246-251 The “anesthetic imprints” in the brain include deposition of amyloid β-42 protein,256,259-261 τ-phosphorylation,252-254 and activation of cell apoptotic markers (esp. caspase-3 pathway).255-258 Each of these events have been shown to produce long-lasting learning and memory impairments in rodents. Interestingly, these processes are also linked to the development of Alzheimer’s disease in humans. Although results of clinical studies are still controversial,259 the current data suggest that the anesthetic per se may be an important contributor to adverse cognitive outcome after surgery.177,260,261

5.1.1 Brain State Monitoring and Anesthetic Administration The AEP trial (Chapter 3) and several meta-analyses have demonstrated that brain state monitoring, such as BIS, is useful in assisting anesthesiologists to optimize anesthetic drug administration. Specifically, anesthesia guided by brain state monitoring leads to lower drug consumption, fewer episodes of deep anesthesia and less risk for perioperative complications.118,172,173,204,208,262 However, it is unclear if lower doses of anesthetics with brain state monitoring will improve cognitive 73

outcome. The purpose of this RCT was to determine whether BIS-guided anesthesia would decrease the risk of postoperative cognitive dysfunction (POCD) and delirium in elderly patients having major noncardiac surgery.

5.1

Methods

5.2.1 Study Design The COgnitive Dysfunction after Anesthesia (CODA) trial was approved by the Clinical Research Ethics Committee (CRE 2003.298, NTWC/CREC/514/07), and all patients gave written informed consents. CODA was a prospective, parallel group, double-blinded, randomized controlled trial. Trial objectives, methods and design have been reported in the Chinese Clinical Trial Registry (ChiCTR-TRC-09000704). Three centers (Prince of Wales Hospital, North District Hospital and Tuen Mun Hospital) participated in the trial. The candidate was the principal investigator of the CODA trial, responsible for the design of the study, funding application, data collection, analysis and interpretation.

5.2.2 Study Population Patients were eligible for the study if they were ≥ 60 years, scheduled for elective major noncardiac surgery, lasting ≥ 2 hours, with an expected hospital stay of ≥ 4 days. Patients were excluded if they were not expected to be available for, or cooperate with, postoperative interviews and neuropsychology testing. Patients who were illiterate, patients with language difficulties and those with significant visual or hearing deficit were excluded from the trial. Other exclusion criteria included patients with major psychosis who were taking regular tranquillizers or antidepressants within the last 3 months. Patients with known or suspected dementia or other forms of

74

memory impairment with a score on mini-mental state examination (MMSE) ≤ 23 were also excluded.263

Since patient performance in neuropsychology testing improves with repeated test administration, another 221 non-surgical patients were recruited from the medical specialist clinics in the Prince of Wales Hospital to quantify this learning effect. These patients fulfilled the above inclusion and exclusion criteria but they were not planned to undergo surgery with general anesthesia within 3 months of enrollment. All patients had identical neuropsychological assessments according to the protocols in the treatment cohorts.

5.2.3 Randomization and Blinding After obtaining informed consents and immediately before induction of anesthesia, patients were assigned to receive either BIS-guided or routine care anesthesia, stratified by center, using a computerized randomization service, accessed through an intranet system. Patients, surgical staff, other healthcare providers and all research personnel, including those who conducted neuropsychology testing and outcome assessment were blinded to the treatment allocation.

5.2.4 Study Procedure Patients were assessed within one week prior to the scheduled surgery. Details on concomitant medical illnesses, surgical disease and the level of education received were recorded. All aspects regarding surgery, perioperative care and monitoring were provided according to local hospital practice.

75

In the operating room, all patients received BIS monitoring. A BIS Quatro sensor (Covidien, Mansfield, MA) was applied to the forehead, before the induction of anesthesia according to the manufacturer’s recommendations. This was connected to an A-2000 System XP monitor. In the BIS group, anesthetic dosage was adjusted so as to achieve a BIS value between 40 and 60 during the maintenance of anesthesia. An audible alarm was set to alert attending anesthesiologists when the BIS value fell out of the prescribed range. In patients allocated to receive routine care, anesthetic drug administration was titrated according to clinical signs. In general, anesthesiologists aimed to maintain systolic arterial pressure within 15 mmHg of the baseline and the heart rate within a range of 40-90 beats/min. When there were signs of inadequate anesthesia, such as sweating, flushing, movement, or swallowing, anesthetic dose was increased as per routine practice. BIS monitoring was continued in the routine care group but the EEG waveform, the BIS number and its trend record were concealed from the monitor display using a data acquisition program, specifically designed for this trial.* BIS values, hemodynamic, and expired gas concentrations were recorded every 5 seconds using the same software.* The timeaveraged BIS value was calculated during the maintenance of anesthesia, roughly defined as the period from 10 min after induction of anesthesia to 5 minute before the last suture was placed). The amount of time with BIS < 40 was also calculated to indicate duration of deep anesthesia.

In all patients, general anesthesia was provided so as to facilitate early emergence after wound closure. Recovery times were recorded from the end of

*The Monitor for PC program is available for non-commercial use at http://www.cuhk.edu.hk/med/ans/softwares.htm 76

anesthesia to eye opening and discharge from the recovery room, when the modified Aldrete score was ≥ 9 (maximum of 10).200 For patients who were transferred to the intensive care unit (ICU) for postoperative mechanical ventilation, the time to tracheal extubation was recorded as a surrogate marker for early recovery.

Following surgery, patients were regularly reviewed by the research staff until hospital discharge. Delirium was assessed daily in the mornings after surgery using the confusion assessment method criteria.264 In this regard, delirium was defined as acute fluctuating course of inattention, and either disorganized thinking or an altered level of consciousness according to the Diagnostic and Statistical Manual of Mental Disorders, 5th edition.265 Patients who were alert, were asked to rate their quality of recovery (QoR) using the Chinese QoR score (Appendix 1).199 Before the surgical procedure and at 3 months after surgery, all patients were asked to complete a short form health survey (SF-36) to indicate their functional health status.

5.2.5 Cognitive Measurements In CODA, cognitive function was measured within a week prior to surgery, and then again at 1 week and 3 months after surgery. All assessments were conducted in a quiet room by certified research staff. All patients were first asked to complete a Chinese version of the cognitive failure questionnaire to indicate potential subjective problems with perception, memory and motor function.266 A battery of neuropyschological tests (verbal fluency test,267 Chinese auditory verbal learning test267 and color trail tests268) was then administered in a standardized fashion. Details of the neuropsychological tests and cognitive failure questionnaire are described in Appendix 3. The tests were chosen because they were sensitive to deficits in verbal

77

memory, language, attention, psychomotor skills and executive functions, and have been validated for local population.267,268 Parallel forms of each test and the sequence of test administration was randomly assigned, so as to minimize possible learning effects.

Change in objective cognitive function was measured by comparing the baseline performance of neuropsychology tests with results obtained at 1 week and 3 months after surgery. A Z score was then calculated to indicate the standardized change in each of the neuropsychology tests by subtracting the average learning effect measured in the non-surgical controls, divided by the standard deviation in the nonsurgical cohort.

𝑍=

∆𝑋 𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡 − ∆𝑋𝑁𝑜𝑛−𝑠𝑢𝑟𝑔𝑖𝑐𝑎𝑙 𝐶𝑜𝑛𝑡𝑟𝑜𝑙 𝑆𝐷(∆𝑋)𝑁𝑜𝑛−𝑠𝑢𝑟𝑔𝑖𝑐𝑎𝑙 𝐶𝑜𝑛𝑡𝑟𝑜𝑙

where, ∆𝑋 𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡 = mean change in patient test score ∆𝑋𝑁𝑜𝑛−𝑠𝑢𝑟𝑔𝑖𝑐𝑎𝑙 𝐶𝑜𝑛𝑡𝑟𝑜𝑙 = mean change in nonsurgical control test score 𝑆𝐷(∆𝑋)𝑁𝑜𝑛−𝑠𝑢𝑟𝑔𝑖𝑐𝑎𝑙 𝐶𝑜𝑛𝑡𝑟𝑜𝑙 = standard deviation of mean in non-surgical controls

A large Z score indicates deterioration in a particular cognitive measure from baseline compared to the non-surgical controls. POCD was defined as two or more Z scores ≥ 1.96.269,270 Similarly, subjective cognitive dysfunction was defined by calculating a Z score for cognitive failure questionnaire using the same method.

78

5.2.6 Outcome The primary outcome was POCD at 3 months after surgery. Secondary outcomes included POCD at 1 week, delirium during hospital stay, and quality of postoperative recovery. Major postoperative complications were also recorded up to 3 months after surgery. The following definitions were used: (1) Myocardial infarction was defined as typical rise and fall in plasma troponin concentrations, associated with either ischemic symptoms, positive findings in electrocardiography, echocardiography, coronary angiography or pathological examination);271,272 (2) Heart failure was diagnosed if there were clinical signs (the presence of any of elevated jugular venous pressure, crackles, crepitations or third heart sound) associated with typical radiographic features (viz. vascular redistribution, interstitial or alveolar pulmonary edema); (3) Thromboembolism was defined by unequivocal findings in venography, duplex ultrasonography, ventilation:perfusion scan or spiral computerized tomography; (4) Pneumonia was defined as pulmonary infiltrates in radiological studies, associated with fever, leukocytosis, desaturation (oxygen saturation < 90% for > 5 min) or positive culture in sputum and/or blood sample.272,273 (5) Other infective complications included wound infection (defined as purulent discharge with or without positive microbial culture), or isolation of microbial pathogens in normally sterile tissue and requiring antibiotic therapy.274,275

5.2.7 Statistical Analyses Statistical analyses were performed using Stata 13 (StataCorp LP, College Station, TX). Sample size was calculated based on an incidence of POCD of 30% in

79

the routine care group.276 In order to detect an absolute risk reduction of 15%, a sample size of 450 patients per group would provide ≥ 90% power at  value of 0.05 (twosided, binomial analysis).

All primary and secondary analyses were conducted according to the principle of intention-to-treat (ITT). The ITT population comprised patients having surgery with general anesthesia and randomized to either BIS monitoring or routine care. Differences in the rate of POCD were compared between groups using the Fisher’s exact test. Recovery times were calculated using Kaplan-Meier analysis and compared between groups using the log-rank test. The Cox proportional hazards model was applied for covariate adjustment and assessment for proportionality assumptions. Secondary endpoints were analyzed using Fisher’s exact test or chi-squared tests and unpaired t test for categorical and continuous data between groups, respectively. Multiple logistic regressions were used to investigate the associations between potential risk factors, POCD and postoperative delirium. All reported p values are two-sided.

5.3

Results We approached a total of 1,657 elderly patients having major noncardiac

surgery. Following screening, 62 (3.7%) patients were excluded because the preoperative MMSE was ≤ 23 points. Another 674 patients were excluded for other reasons. The CODA trial included 921 patients, of whom 462 patients (50.2%) received BIS-guided anesthesia, and 459 patients (49.8%) were randomized to the routine care group. About 85.0% and 90.7% of patients completed the 1 week and 3

80

months assessments, respectively (Figure 5.1). Baseline characteristics and surgical details of patients at entry of the trial were similar between study groups (Table 5.1).

Figure 5.1. Flow chart of trial enrollment

81

Table 5.1. Patient characteristics at entry of CODA trial Nonsurgical controls

BIS Group

Routine Care Group

No. of patients Age, years† Male sex, no. (%)

221 67.3 ± 8.1 135 (61.1%)

450 68.1 ± 8.2 280 (62.2%)

Weight, kg†

60.5 ± 10.8

62.0 ± 11.5

452 67.6 ± 8.3 273 (60.4%) 61.4 ± 10.7

ASA status, no. (%) 1-2 3-4

183 (82.8%) 36 (16.3%)

373 (82.8) 76 (16.9)

382 (84.5) 70 (15.5)

Pre-existing comorbidity, no. (%)# Cardiovascular Respiratory Endocrine Others

166 (75.1%) 36 (16.3%) 51 (23.1%) 40 (18.1%)

374 (83.1) 75 (16.7) 109 (24.2) 80 (17.8)

330 (73.0) 67 (14.8) 101 (22.3) 87 (19.2)

0.541 0.701 0.532 0.633

Surgery for cancer, no. (%)

̶

338 (75.1)

352 (77.9)

0.392

Duration of anesthesia (hours)† ̶

2.1 ± 1.0

2.0 ± 1.1

0.671

Years of education received‡

6 (0-19)

6 (0-22)

6 (0-18)

0.841

Chinese Geriatric Depression Scale‡

2 (0-13)

2 (0-13)

2 (0-14)

0.181

Mini-Mental State Examination score‡

28 (24-30)

28 (24-30)

28 (24-30)

0.640

p value*

0.421

0.471 0.582

Values are number (%) or †mean ± standard deviation or ‡median (range). *Comparing between BIS versus routine care groups # Patient may have more than one pre-existing medical conditions.

82

Details on anesthetic administration are shown in Table 5.2. BIS monitoring decreased the mean (± standard deviations) end-tidal volatile concentration from 0.93 ± 0.34 MAC to 0.57 ± 0.29 MAC, relative reduction was 29.7% (95% CI: 25.9-32.8, p < 0.001). Similarly, the mean (± standard deviations) estimated propofol effect site concentration was reduced from 3.3 ± 1.1 µg/L to 2.7 ± 0.9 µg/L, relative reduction was 20.7% (95%CI: 12.1-31.9, p < 0.001). Consequently, the median BIS value (interquartile range) during surgery in the BIS-guided anesthesia, 53 (48-57) was higher than that in the routine care group, 36 (31-49), p < 0.001. The amount of time (mean ± standard deviations) when BIS < 40 was also lower in the BIS group compared with controls (7.2 ± 7.8 min versus 22.8 ± 7.3 min, p < 0.001).

5.3.1 Primary and Secondary Outcomes Table 5.3 reports the test scores of cognitive failure questionnaire and their performance in neuropsychology testing at baseline, 1 week and 3 months after surgery. There were fewer patients with delirium in the BIS-guided group compared with routine care during hospital stay, absolute risk reduction was 8.6% (95% CI: 3.413.7), but the rates of POCD at 1 week after surgery were not different between groups. In contrast, BIS-guided anesthesia significantly reduced the rate of POCD up to 3 months after surgery (Table 5.4). The absolute risk reduction was 4.5% (95% CI: 0.25-8.9). The number needed to treat was 23 (95% CI: 6-391). The benefit of BIS monitoring was unaffected with a multivariate analysis that adjusted for age, gender, preoperative MMSE score, education status, average BIS value during anesthesia and postoperative delirium while in hospital [adjusted odds ratio (95% CI): 0.67 (0.320.98), p = 0.025].

83

Table 5.2. Comparison of anesthetic techniques

BIS group

Routine Care group

No. of patients

450

452

Propofol dose, mg Opioid dose Fentanyl, μg/kg no. (%) Morphine, mg no. (%) Midazolam dose, mg no. (%) Estimated effect site propofol concentration (µg/ml) no. (%) End-tidal volatile concentration, MAC equivalents,‡ no. (%) Nitrous oxide use no. (%) End-tidal concentration (%)

136 ± 30

148 ± 33

0.647

1.6 ± 0.4 400 (88.8) 0.12 ± 0.07 373 (83.0) 2.5 ± 1.1 33 (7.3) 2.7 ± 0.9

1.6 ± 0.3 411 (90.9) 0.12 ± 0.06 381 (84.4) 2.9 ± 1.4 27 (5.9%) 3.3 ± 1.1

0.752