Beruflich Dokumente
Kultur Dokumente
Anaesthesia
Ana Castro
Departamento de Engenharia Electrotcnica e de Computadores
Faculdade de Engenharia da Universidade do Porto
Advisor:
Co-Advisor:
Clinical Advisor:
Acknowledgements
The work developed during the course of this doctoral program, and here presented,
was possible with the contribution of a wide group of people that somehow touched me
during this period, and to only some of whom it is possible to give particular mention.
Let me begin by thanking the support and scientic rigor forwarded by my advisor
Prof. Catarina S. Nunes, as well as for the refreshing look and tireless meetings with
my co-advisor Prof. Fernando Gomes de Almeida.
My special thanks to the person who introduced me into the hospital environment,
and gave me a privileged knowledge on the medical specialty of Anesthesiology, Dr.
Pedro Amorim, for his endless enthusiasm for discovery and readiness to face new challenges and projects.
I would like to acknowledge the Investigao Clnica do Servio de Anestesiologia,
of Santo Antnio Hospital, who welcomed me during this period, to the Director of
the Servio de Anestesiologia Dra. Isabel Arago Fechs, the Director of the Servio de
Bloco Operatrio Dr. Simo Barros Esteves, and to the Director of the Departamento
de Anestesiologia Cuidados Intensivos e Emergncia Dr. Antnio Marques da Silva, for
their support in the approved research studies.
I would like to thank the prompt and tireless collaboration of a large group of
professionals, who always made available to support data collection performed at the
Urology operating room: nurses, surgeons, auxiliaries, anesthesiologists, interns, there
were many who contributed for a successful outcome, and to which I leave a warm thank
you. I leave my special thanks to the anesthesiologists who accompanied me throughout
these years Dra. Eduarda Amadeu, Dra. Ftima Martins and Dra. Paula S Couto,
it was thanks to their support, patience, teaching, and personal involvement that this
vii
work has gained strength, it was with them that I have learned and gained inspiration
for Medicine. A warm thank you to the intern Diana Afonso, to which was a pleasure
to work and share experiences with.
I would like to thank all patients and volunteers who agreed to participate in the
studies, from whom I always received the maximum collaboration. I also would like to
acknowledge Neurinbloc, Helena and Filipe, for their expertise and collaboration.
I was fortunate to nd several funny and interesting friends during the course of
this work, I would like to leave a particular warm mention to my lab mates, Susana
Brs and Nadja Bressan, who attended over the years always with a friendly word and
stories of happiness. To Snia Gouveia for the wise counsels, and Ana Maria and Aura
Maia for the interesting discussions, my warm thank you.
To my friends Fernanda and Tiago, for the friendship and good years in the university, and to my friend Clia Cruz, who has accompanied me since the rst steps in
school, companion and witness of life, thank you for the certainty of mutual support
that has always united us.
A special and caring thank you to Silvio for the love, and companionship that carried
and kept me going in the nal moments.
Of course, I would not forget my doggy roommates, scar and Maggie, for the love,
friendship, and no doubt endless joy with which they wake me up, and receive me at
home, every single day.
I thank my family, in particular my parents in my heart for all the dedication and
unconditional love. My eternal gratitude and love.
Last but not least, I would like to acknowledge the nancial support provided by
the Fundao para a Cincia e a Tecnologia, Portugal (SFRH/BD/35879/2007), for the
scholarship which allowed my dedication in full time to the development of this thesis.
Abstract
This thesis aims to study and develop objective measures of the nociception/antinociception balance during anaesthesia. In spite of the increasing interest in recent years
on both pain regarding to surgery and objective assessment of the analgesia component
adequacy, these are still open issues due mainly to the fact that general anaesthesia presupposes an unconscious subject, which implies lack of collaboration from the subject.
Due to this fact, the term nociception is used instead of pain whenever regarding
to unconscious subjects.
ix
Later, this
information was used to construct a measure of noxious stimulus intensity, using Rasch
analysis, which was presented in the form of scale. Regarding to a surgical procedures
under general anaesthesia, this scale allows an estimative of the stimuli intensity during
the procedures, and also allows to model the input drugs' doses and stimuli intensities
impact on the physiological variables linked to noxious activation.
The rst clinical study was focused on the response to precise noxious stimuli, at
dierent analgesic drug doses under a TIVA TCI general anaesthesia of propofol (hypnotic) and remifentanil (analgesic).
with emphasis on the collected waves in order to extract relevant information on noxious activation.
amplitude, systolic blood pressure, respiration rate and EEG derived indexes.
Data
The rst
was an homeostasis index, combination of steady-state individual indexes of each variable shown to contain information on noxious activation, applying wavelet analysis
adjusted to each physiological signal characteristics.
to painful electrical stimuli were measured on the somatosensorial brain area, and related to the stimulation intensity and anaesthetic drugs' attenuation.
This approach
allowed the detection of individual sensing thresholds and expected response to pattern
stimulation, which presented large interindividual variability. In this study cortical somatosensory evoked potentials were found to be related to the stimulus intensity, to
pain reported by the volunteer in a numerical rating scale (0 to 10), and to drug attenuation in a dose dependent manner, both with propofol and remifentanil. This may
congure a direct and objective method to translate the nociception/anti-nociception
balance in uncommunicative patients, allowing for the construction of normalized measures, comparable between subjects.
Finally, a simulation study on the closed-loop control of a general anaesthetic drug
was conducted.
years, with proposed automatic administration systems for the muscle relaxant and
hypnotic drug, which correspond to general anaesthesia components that already have
validated monitors to translate the patient's state. A new approach to the control of
anaesthetic drugs' administration is here presented for the hypnotic component, taking
into account inter-patient variability described in the modelling process. The technique
employed was a sliding-mode controller, that presented robust responses to patients
diering from the nominal. The common features to the analgesia component, and the
necessary studies to further evolve in the automatic administration of general anaesthesia were discussed.
Summarizing, two approaches to the nociception/anti-nociception balance assessment were proposed in this thesis, diering from the methods presented in the literature, and shown to adequately respond to noxious stimulus and anaesthetic drugs.
Future studies should be conducted to further develop and validate the proposed indexes. Following its validation, the indexes may be employed in the closed-loop control
of the nociception/anti-nociception balance during anaesthesia, taking into account the
observed inter-patient variability.
Resumo
xiii
wavelet
O segundo estudo clnico foi desenhado para investigar a aplicabilidade das respostas somatosensoriais evocadas na avaliao do balano nocicepo/anti-nocicepo,
para diferentes combinaes de propofol e remifentanil. No total, 10 voluntrios saudveis foram includos no estudo, aps consentimento informado e aprovao institucional.
Rsum
xvii
boucle de contrle d'un drogue anesthsique. Cette thse propose galement un examen complet sur le sujet de l'valuation de l'objectif de la nociception et l'quilibre
analgsique pendant l'anesthsie gnrale.
La premire tape d'une valuation de l'quilibre nociception/anti-nociception est
de rpondre la question suivante Quelle est l'intensit de la stimulation nociceptive
applique au patient qui dclenche les rponses physiologiques observes ?. Pour rpondre cette question, une questionnnaire a t mene dans les anesthsistes cliniques
an d'valuer la perception clinique de l'intensit des stimuli nocifs lis aux procdures
d'anesthsie et de chirurgie. Un total de 57 anesthsistes ont rpondu, classiant 35
stimuli dans une chelle ordinale allant de 0 10. Ultrieurement, cette information a
t utilise pour construire une mesure de l'intensit du stimulus nociceptif, en utilisant
l'analyse de Rasch, qui a t prsent sous la forme d'chelle. En ce qui concerne une
intervention chirurgicale sous anesthsie gnrale, cette chelle permet une estimative
de l'intensit des stimuli au cours des procdures, et permet galement de modliser les
doses des drogues d'entre et les stimuli intensits impact sur les variables physiologiques
lis l'activation nocives.
La premire tude clinique a porte sur la rponse des stimuli nocifs, notamment
direntes doses de drogues anesthsiques sous anesthsie gnrale TIVA TCI du propofol (hypnotique) et rmifentanil (analgsique). Les donnes physiologiques recueillies ont
t prtraites, en mettant l'accent sur les ondes recueillies visant l'extraction des informations pertinentes sur l'activation nocives. Ces informations comprennent le rythme
cardiaque, pouls plthysmographie amplitude de l'onde, la pression artrielle systolique,
taux de respiration et EEG driv indexes. Les donnes ont t inspectes an d'identier les variables physiologiques qui permettent une rponse adquate la stimulation
et l'attnuation de la dose analgsique. Un total de 34 patients ont t inclus dans
l'tude aprs consentement clair et l'approbation institutionnelle, et rpartis au hasard en trois groupes d'tude en fonction de la dose de rmifentanil (Ce=2,0, 3,0 ou
4,0 ng/ml). Bas sur l'tat actuel de la technique, deux mthodologies direntes ont
t utilises. Le premier tait un indice homostasie, la combinaison de l'tat stable des
indices individuels de chaque variable a montre contenir des informations sur l'activation nocive, en appliquant l'analyse en ondelettes ajusts chaque caractristique du
signal physiologique. Le second tait un modle physiologique dynamique intgrant les
signaux physiologiques lie l'activation nocives prcdemment identies, capable de
reproduire les relations entre l'estimation de relance peru par le patient comme une
aux drogues anesthsiques. tudes futures doivent tre menes an de dvelopper et
de valider les indices proposs. Aprs sa validation, les indices pe uvent tre utilises
dans le contrle en boucle ferme de l'quilibre nociception/anti-nociception pendant
l'anesthsie, en tenant compte de la variabilit inter-individuelle.
Contents
List of Figures
xxvii
List of Tables
xxxix
List of Abbreviations
xlv
1 Introduction
1.1
1.2
1.3
1.5
1.1.2
Hypnosis Monitors . . . . . . . . . . . . . . . . . . . . . . . . . .
Pain
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2.1
Sense Organs
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2.2
12
1.2.3
. . . . . . . . . . . . . . . . . . . . . . . .
14
1.2.4
16
Anaesthetic Drugs
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
20
1.3.1
Analgesics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
21
1.3.2
Hypnotics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
22
1.3.3
1.4
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
State of the Art on the Analysis and Control of the Nociception Level
24
32
1.4.1
. . . . . . . . .
32
1.4.2
Automatic Control . . . . . . . . . . . . . . . . . . . . . . . . . .
41
43
xxi
CONTENTS
2 Data Collection and Pre-Processing
47
2.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
47
2.2
48
2.2.1
Design Goals
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
48
2.2.2
Selection Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . .
49
2.2.3
Pre-Medication . . . . . . . . . . . . . . . . . . . . . . . . . . . .
50
2.2.4
Clinical Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
50
2.2.5
Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
51
2.3
Institutional Approval
2.4
Data Overview
2.4.1
2.5
2.6
. . . . . . . . . . . . . . . . . . . . . . . . . . . .
54
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
54
Files' Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . .
57
Waves Pre-Processing
. . . . . . . . . . . . . . . . . . . . . . . . . . . .
62
2.5.1
2.5.2
2.5.3
. . . . . . . . . . .
68
2.5.4
Respiration Rate . . . . . . . . . . . . . . . . . . . . . . . . . . .
68
Summary
. . . . . . . . . . .
62
. . . . . . . . . . . . . . . . . . . . . . .
67
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
70
71
3.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
71
3.2
Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
73
3.2.1
74
3.2.2
Scale Construction . . . . . . . . . . . . . . . . . . . . . . . . . .
74
3.2.3
77
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
79
3.3.1
81
3.3.2
82
3.3.3
88
3.4
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
89
3.5
Summary
92
3.3
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
xxii
CONTENTS
4 Data Analysis - Passive Nociception Measures
95
4.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
95
4.2
. . . . . . . . . . . . . . . . . . . . . .
96
4.2.1
Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
97
4.2.2
Results
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
100
4.2.3
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
110
112
4.3.1
. . . . . . . . . . .
113
4.3.2
Homeostasis Index
. . . . . . . . . . . . . . . . . . . . . . . . . .
124
4.3.3
. . . . . . . . . .
127
4.3.4
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
128
130
4.4.1
Individual Modelling . . . . . . . . . . . . . . . . . . . . . . . . .
136
4.4.2
Merged Modelling
. . . . . . . . . . . . . . . . . . . . . . . . . .
140
4.4.3
143
4.4.4
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
144
4.3
4.4
4.5
Summary
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
146
149
5.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
149
5.2
Evoked Potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
151
5.3
155
5.4
164
5.4.1
Design Goals
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
164
5.4.2
Selection Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . .
164
5.4.3
Clinical Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
165
5.4.4
Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
166
5.4.5
Institutional Approval
. . . . . . . . . . . . . . . . . . . . . . . .
169
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
169
5.5
5.5.1
5.5.2
5.5.3
. . . . . . . . . . . . . . . . . . . . . . . . . . . .
172
173
174
xxiii
. . . . . . .
CONTENTS
5.5.4
5.5.5
176
177
5.5.6
179
5.5.7
183
5.5.8
. . . . . . . . . . . . . . .
. . .
183
5.6
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
185
5.7
Future Developments . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
188
5.8
Summary
190
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6 On Nociception Control
193
6.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
193
6.2
198
6.3
. . . . . . . . . . . . . . . . . . . . .
. . .
202
6.3.1
203
6.3.2
Feedback Linearization . . . . . . . . . . . . . . . . . . . . . . . .
205
6.3.3
Sliding-Mode Control
. . . . . . . . . . . . . . . . . . . . . . . .
207
6.3.4
Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
210
6.3.5
Results
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
211
6.3.6
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
213
6.4
. . . . . . . . . . . . . . . . . . . . . .
216
6.5
Summary
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
217
219
. . . . . . . . . . . . . . . . . . . . . . . . .
List of Publications
222
225
229
A.1
232
A.2
235
. . . . . . . . . . . . . . . . .
239
243
xxiv
CONTENTS
D Informed Consent Form - Volunteers' Study
245
249
References
288
xxv
CONTENTS
xxvi
List of Figures
1.1
1.2
1.3
1.4
manufacturer recom-
. . . . . . . . .
1.5
1.6
Schematic representation of the dorsal horn of the spinal cord layers and
terminations of the three types of primary aerent neurons pathways. . .
13
1.7
14
1.8
. . . . . . . . . . . . . . . . .
From top to
. . . . . . . . . . . . . .
19
. . . . . . . . . . . . .
20
1.10 Remifentanil 2D molecular schematic representation (methyl 1 - (3 methoxy - 3 - oxopropyl) - 4 - (N-phenylpropanamido) piperidine - 4
- carboxylate).
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
21
23
1.12 Drugs' action path in the human body, from drug dose administration to
observed clinical eect. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
xxvii
24
LIST OF FIGURES
1.13 Three compartments pharmacokinetic (PK) model, followed by the pharmacodynamic (PD) model and Hill equation, translating the eect-site
concentration into the output measurable eect. Where
the drug transfer rate from compartment
is the elimination rate), and
mes; in the Hill equation
drug in the system (Ce
E0
V1 , V2
and
V3
kij ,
represents
to the compartment
j (k10
= 0), EC50
represents the steepness of the descent of the eect response to the drug.
26
1.14 Screenshot of the educational software specially developed for the simulation of target-controlled infusion anaesthesia using BIS or Entropy. It
shows the surface relational model between propofol and remifentanil's
synergistic eect on the hypnosis indexes.
. . . . . . . . . . . . . . . . .
28
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
33
1.16 Bispectral Index (BIS) trend during general anaesthesia, with the induction, maintenance and recovery phases. The dashed lines represent
the BIS manufacturer recommended target range for general anaesthesia
(40-60). Data from one of the patients in this study.
. . . . . . . . . . .
36
38
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
41
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
xxviii
44
LIST OF FIGURES
1.20 Representation of the relation between the stimulus and administered
drugs on a specic patient, producing an amplitude response (patient
sensitivity) in the nociception/anti-nociception balance index. . . . . . .
2.1
2.2
55
T M monitor; B)
Clinical setup used in data collection: A) BIS VISTA
Orchestra
R
Base Primea syringe pumps in TCI mode (propofol and re-
R
monitor and ventilator; D) Com-
49
2.3
45
c Waves installed. . . . . . .
56
2.5
57
2.6
. . . . . . . . . . . . . . . . . .
i = 1, ..., 34).
. . . . . . . . . . . . . .
61
Original ECG signal (black line), ltered ECG signal (QRS complex
enhancing, gray line), and correspondent QRS complex detection (black
dot). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.7
65
2.8
65
2.9
. . . . . . . .
66
. . . .
67
2.10 Photoplethysmography wave and correspondent wave amplitude denition (PPGA): A is the local minimum at the beginning of the pulse, and
B is the local maximum. PPGA is dened as the dierence between the
two.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
xxix
68
LIST OF FIGURES
2.11 Photoplethysmography wave and detected points for amplitude assessment (A, white dot, dened as the local minimum and B, black dot, as
the wave local maximum). Data from one of the patients in the study.
69
69
2.12 CO2 wave and detected point of maximum CO2 for each respiration cycle,
corresponding to the end of an expiratory cycle. Data from one of the
3.1
73
3.3
. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
78
3.4
80
Total score for each stimulus within the three groups of stimuli studied:
anaesthetic procedure stimuli (Ani, i=1,...,11), pre-incision and incision
surgical stimuli (PreIncj, j=1,...,10), and post-incision surgical stimuli
(PostInck, k=1,...,14).
. . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.5
3.6
3.7
81
81
i = 1, ..., 35,
stimulus location (i ,
. . . . . . . . . . . . . . . . . . . . . . . . . . .
84
3.9
80
85
. . . . . . . . .
86
Relation between score rates (0 to 10) and pain intensity measures (logit),
with overlapping 0.5 points of expected score (squares).
n = 1, ..., 57,
xxx
86
LIST OF FIGURES
3.11 Application of the developed stimulus intensity scale: schematic representation of an urological procedure under general anaesthesia with identied stimuli intensities of the anaesthetic and surgical actions.
. . . . .
87
3.12 Comparison between pre-intensities estimated in previous pharmacological studies, and the Rasch scale obtained using anaesthesiologists noxious
stimuli perception.
4.1
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
88
Schematic representation of proposed methods to assess amplitude responses prior and post-stimulation: a) and b) average versus maximum
response; c) and d) average values prior and post stimulation. . . . . . .
4.2
97
4.3
4.4
99
114
Diagram presenting the algorithm steps: the patient is the system, with
propofol and remifentanil drugs' eect-site concentrations (Ce) as inputs,
and output physiological variables entering the wavelet based algorithm
to determine the combined steady-state index as summarized.
Dashed
line rectangle presents the initially used input detection rule, replaced by
the wavelet based algorithm. . . . . . . . . . . . . . . . . . . . . . . . . .
4.5
4.6
118
121
4.7
122
Schematic representation of the signal processing steps to obtain the individual and combined steady-state indexes, from data collection, to signal
pre-processing unit, and homeostasis index presentation (individual and
combined), as implemented in the developed tool for on-line homeostasis
assessment.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
xxxi
125
LIST OF FIGURES
4.8
Designed software for on-line assessment of the developed steady-state indexes: individually for each input (propofol and remifentanil Ce), output
(bispectral index, electromyography, heart rate, blood pressure, photoplethysmography wave amplitude, respiration rate), and combinations of
input and output signals (In and
Out ).
is from one of the patients of the study, simulating an on-line assessment. 126
4.9
. . . . . . .
128
4.10 Representation of the attenuation and depression eects of the analgesic drug on a stimulus of known intensity, with the introduction of an
additional depression curve.
. . . . . . . . . . . . . . . . . . . . . . . . .
130
4.11 From top to bottom: stimulus intensity trend, using author's annotations and intensities estimated by the Rasch analysis;
preopioid intensity
estimated based on the Rasch scale extrapolation; estimated remifentanil eect-site concentration (Ce in ng/ml);
postopioidintensity estimates
preopioid intensity
based on extrapolated
total surgical stress. Data from one of the patients in the study. . . . . .
134
4.12 Structure representing the impact of anaesthetic drugs and noxious stimulus on the physiological measurable eects. . . . . . . . . . . . . . . .
135
. . . . . .
135
4.14 Modelled perceived stimulus, based on the stimulus intensity and attenuation provided by the analgesic remifentanil (output), and bispectral index, frontal electromyography, heart rate, systolic blood pressure,
pulse plethysmography wave amplitude, and propofol eect-site concentration as inputs.
postopioid intensity .
. . . . . . . . . . . . . . . . . . . . . . . . . . .
xxxii
138
LIST OF FIGURES
4.15 Modelled perceived stimulus, based on the stimulus intensity and attenuation provided by the analgesic remifentanil, using total surgical stimulus
(TSS) as output signal, and bispectral index, frontal electromyography,
heart rate, systolic blood pressure, pulse plethysmography wave amplitude, and propofol eect-site concentration as inputs.
(a) Individually
adjusted state-space model of order 1, in the data set with higher correlation to the output signal TSS (correlation of 0.97). (b) Individually
adjusted state-space model of order 1, to the data set with lowest correlation (correlation of 0.51). . . . . . . . . . . . . . . . . . . . . . . . . . .
139
4.16 Modelled perceived stimulus, based on the stimulus intensity and attenuation provided by the analgesic remifentanil, using total surgical stimulus
(TSS) as output signal, and bispectral index, frontal electromyography,
heart rate, systolic blood pressure, pulse plethysmography wave amplitude, and propofol eect-site concentration as inputs. (a) Data set with
higher correlation to the output signal TSS, considering the merged data
model. (b) Data set with lowest correlation, considering the merged data
model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
142
5.1
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2
144
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
151
5.3
5.4
153
154
xxxiii
158
LIST OF FIGURES
5.5
5.6
International 10-20 system for electroencephalogram collection. The letters used stand for:
161
. .
166
A) SEPs
T M CR (Nicolet-Vyasis); B)
169
5.9
. . .
. . . . . . . . . . . . . . . . .
171
BIS trends for all volunteers during the experiment (a), and one case in
detail (b), centered according to propofol start (time=0). The study was
interrupted as soon as the volunteer lost response to verbal command,
maintaining response to mechanical stimulation, or by clinical advice. . .
172
5.10 Sensitive (ST), motor (MT), painful (PT) and 1.3 painful (1.3PT) threshold currents of stimulation for each volunteer. . . . . . . . . . . . . .
173
5.11 Relation between the painful stimulus (a) and 1.3 painful stimulus (b)
intensity and the SEPs' latency and amplitude responses prior drug administration.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
174
5.12 Numerical rating scale pain evaluations (0-10) for each volunteer both
for the painful threshold (PT) and 1.3PT stimuli. . . . . . . . . . . . .
175
5.13 Numerical rating scale evaluations (NRS) versus original (a) and normalized (b) SEP amplitude. . . . . . . . . . . . . . . . . . . . . . . . . . . .
178
5.14 Numerical rating scale evaluations (NRS) versus: (a) SEP amplitude/SEP
latency; (b) normalized SEP amplitude/normalized SEP latency.
. . . .
179
180
5.16 Original (a) and normalized (b) ratio between cervical SEP amplitude
and latency, versus evaluations in the numerical rating scale. . . . . . . .
xxxiv
181
LIST OF FIGURES
5.17 Normalized cervical SEP ratio versus remifentanil eect-site concentrations for each volunteer.
. . . . . . . . . . . . . . . . . . . . . . . . . . .
182
5.18 Boxplot of the normalized SEP amplitude/normalized SEP latency ratio (RN orm ) for the maximum drug doses combination achieved in each
volunteer in each administration scheme: 1 - No drugs; 2 - Remifentanil
only; 3 - Propofol increasing steps and constant remifentanil (1.0 ng/ml);
4 - Remifentanil increasing steps and constant propofol (1.2 g/ml). . . .
182
5.19 Normalized SEP cortical ratio in the original linear scale (a) and logarithmic scale (b), versus evaluations in the numerical rating scale. . . . .
184
5.20 Collected data points from all volunteers considering propofol and remifentanil eect-site concentrations (Ce) as inputs, and the normalized
SEPs ratio (RN orm ) as output measurable eect.
. . . . . . . . . . . . .
185
6.1
Schematic representation of the analogy between the automatic pilot systems in the airplane, and the anaesthesiologists' action during general
anaesthesia: induction, maintenance, noxious stimulation regulation and
recovery. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.2
194
Schematic representation of the general anaesthesia triad, with corresponding anaesthetic drugs, and their direct or indirect interference on
each other: line arrows represent the synergy relation between the hypnotic and the analgesic drugs, dashed arrows represent the impact that
the paralysis state of the patient has on the monitoring devices used to
assess the hypnosis and analgesia components, and nally dash-dot arrow
represents the impact that the hypnotic drug may have on the variables
related to noxious activation used to monitor the analgesia component. .
6.3
199
Schematic representation of an automatic control system for the administration of general anaesthetics, using a multiple-input and multipleoutput (MIMO) approach for the general anaesthesia triad.
6.4
. . . . . . .
xxxv
201
210
LIST OF FIGURES
6.5
Control activity for the nominal patient using the sign function: on top
SE and SE reference, below the error sequence, and on bottom the infusion rate (ml/h).
6.6
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
211
Control activity for the nominal patient using the saturation function:
on top SE and SE reference, below the error, and on the bottom the
infusion rate (ml/h).
6.7
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
212
Control activity for the minimum patient using the saturation function:
on top SE and SE reference, below the error, and on the bottom the
infusion rate (ml/h).
6.8
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
212
Control activity for the maximum patient using the saturation function:
on top SE and SE reference, below the error, and on the bottom the
infusion rate (ml/h).
6.9
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
213
Representation of the patient modelled system, with possible error inputs, controller, and state observer, for an on-line implementation of the
automatic control of the hypnotic.
. . . . . . . . . . . . . . . . . . . . .
214
E.1
massi
kij
= 1, ..., 3).
ce the eect-site
. . . . . . . . . . . . . . . . . . .
251
Step response simulations for the Minto model, considering dierent combinations of age, weight, height and gender.
E.4
250
E.3
215
. . . . . . . . . . . . . . . .
252
R
, with the system re-
xxxvi
. . . . .
253
LIST OF FIGURES
E.5
Representation of the diagram block for a possible on-line implementation, with the input incoming from the monitored values of State Entropy
(SE) in the patient, and with the output infusion rate (ml/h) directed to
the syringe pump actuator.
. . . . . . . . . . . . . . . . . . . . . . . . .
xxxvii
254
LIST OF FIGURES
xxxviii
List of Tables
1.1
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2
11
1.3
Types of nerve bers that carry aerent impulses to the central nervous
system. A and B bers are myelinated, C bers unmyelinated. . . . . . .
1.4
12
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.5
1.6
1.7
17
. . . . .
18
. . . . . .
18
pharmacokinetic and dynamic model, and corresponding percent coecient of variation (% CV). The parameters incorporate patient's age and
lean body mass (LBM).
1.8
. . . . . . . . . . . . . . . . . . . . . . . . . . .
30
Estimated relations for the Schnider three compartments pharmacokinetic model parameters. It incorporates patient's age, weight, height and
lean body mass (LBM).
1.9
. . . . . . . . . . . . . . . . . . . . . . . . . . .
(1=1,...,11) parame-
2.1
31
. . . . . . . .
31
Study interruption rules: for each case the anaesthesiologist should redene drugs' theoretical eect-site concentration targets, according to
patient's needs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2
51
deviation). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
xxxix
56
LIST OF TABLES
2.3
58
2.4
59
2.5
3.1
. .
scale.
60
3.2
. . . . . . . . . . . . . . . . . . . .
75
Total and average scores obtained for each stimulus, divided according
to the considered groups, stimuli measures (i ,
i = 1, ..., 35,
in the Rasch
model, with greater logit values indicating higher pain perception), standard errors (S.E.), expected raw score, Rasch measure parameterized in
a scale from 0 to 10, and stimulus identication, in descending order of
stimulus intensity perception (reliability of 0.98).
3.3
. . . . . . . . . . . . .
4.1
83
88
. . . . . . . . . . . . . .
100
4.2
Patients' distribution on the three possible incision sites, for each group
of study, in the nal sample (no statistical dierence between groups).
4.3
101
P < 0.05).
102
Median values for each cardiovascular variable analyzed in the study, for
each stimulus (average pre and post stimulation), divided according to
the drug dose group: remifentanil eect-site concentration (RemiCe) of
2.0, 3.0, or 4.0 ng/ml (* stands for signicant dierences with
4.4
Median values of BIS related variables analyzed in the study, for each
stimulus (average pre and post stimulation), divided according to the
drug dose group: remifentanil eect-site concentration (RemiCe) of 2.0,
3.0, or 4.0 ng/ml (* stands for signicant dierences with
4.5
P < 0.05).
. .
103
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
xl
105
LIST OF TABLES
4.6
Median amplitude response of BIS related variables for each stimulus and
remifentanil eect-site concentration (RemiCe) target considered in the
study.
4.7
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
106
4.8
. . . . . . . . . . . . . . . . . . . . . . . . . . .
107
4.9
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
108
Median values of available CVI trends, for each stimulus (average pre and
post stimulation), divided according to the drug dose group: remifentanil
eect-site concentration (RemiCe) of 2.0, 3.0, or 4.0 ng/ml. . . . . . . . .
109
j = 1, 0.
119
j > 5,
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
119
4.12 Estimated thresholds based on historic data for each signal analyzed, in
a training set of ve patients.
. . . . . . . . . . . . . . . . . . . . . . . .
120
4.13 Percentage of total time (total of 31 patients), that each signal analyzed
in this study phase was found to be in steady-state (SS) conditions (data
as meanstandard-deviation). . . . . . . . . . . . . . . . . . . . . . . . .
121
4.14 Median values of the proposed homeostasis index (HI): average value in
the periods baseline, prior and post-stimulation, for each remifentanil
eect-site concentration (RemiCe) group.
. . . . . . . . . . . . . . . . .
123
scale
. . . . . . . . . . . . . . . . .
124
Stimulus
intensities estimated and translated in the Rasch scale, as well as extrapolated values for the pre-intensities, considering the linear relation
shown in Figure 3.12. Table entries presented in bold correspond to the
stimuli that have pre-intensities estimations; original values presented
between parenthesis.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
xli
132
LIST OF TABLES
4.17 Order 1 state-space models adjusted to the individual data, t statistics,
and correlation between output and modelled output. Model 1 corresponds to the output TSS, and Model 2 to the output
postopioidintensity .
Three input sets were considered using bispectral index (BIS), frontal electromyography (EMG), heart rate (HR), systolic blood pressure
(SBP), pulse photoplethysmography wave amplitude (PPGA), and propofol eect-site concentration (PropCe). Data presented as meanstandarddeviation.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
137
4.18 Order 1 state-space models adjusted to the merged data sets, t statistics, and correlation between output and modelled output, considering
baseline removed data, and baseline normalized data. Three input sets
were considered using bispectral index (BIS), frontal electromyography
(EMG), heart rate (HR), systolic blood pressure (SBP), pulse photoplethysmography wave amplitude (PPGA), and propofol eect-site concentration (PropCe). Data presented as meanstandard-deviation.
. . .
141
4.19 Average, minimum and maximum values of the merged model output perceived stimulus during the maintenance phase, considering all data sets
and input set of BIS, EMG, HR, SBP, PPGA and PropCe: population
trends. Data presented as meanstandard-deviation. . . . . . . . . . . .
143
4.20 Parameters of the state-space models using the merged data, considering
the three input sets.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.1
5.2
. . .
143
170
tion rules applied in the study: the study began with remifentanil administration only, in increasing steps as described in the text, until one of
the interruption rules occurred; after interruption the drug doses were reduced to the rst step considered in the next phase, and drug doses were
increased again according to the scheme. The study had three sequent
schemes, rst remifentanil only, followed by constant remifentanil and
propofol increasing steps, and nally constant propofol and remifentanil
increasing steps. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
xlii
170
LIST OF TABLES
5.3
Numerical rating scale (NRS) evaluations in each volunteer for the baseline and maximum drug doses combination achieved in each administration scheme (NE - no evaluation). . . . . . . . . . . . . . . . . . . . . . .
5.4
175
5.5
176
. . . . . . . . . . . . . . . . . .
177
6.1
Estimated parameters for the Hill equation using State Entropy monitoring.205
A.1
Recommended doses of remifentanil in co-administration with other anaesthetic agents for adult patients. . . . . . . . . . . . . . . . . . . . . . .
230
A.2
231
A.3
E.1
. . . . . . . . . . . . . . . . . . .
232
xliii
255
LIST OF TABLES
xliv
List of Abbreviations
AAI
ACC
AEP
ANSSI
ASA
BIS
Bispectral Index
BP
Blood Pressure
Ce
CHEPS
CO2
Carbon Dioxide
Cp
CSSA
Clinical Signs-Stimulus-Antinociception
CVI
DLPFC
DNIC
ECG
Electrocardiogram
ECoG
Electrocorticogram
EEG
Electroencephalogram
xlv
List of Abbreviations
EKF
EMG
Electromyography
fMRI
HBI
HR
Heart Rate
HRbaseline
HRV
IBP
IIR
LOC
Loss of Consciousness
MAC
MIMO
MIR
Noc/ANoc
Nociception/Anti-Nociception
NRS
NSRI
PET
PKPD
PPC
PPG
Pulse Plethysmography
PPGA
RE
Response Entropy
RespR
Respiration Rate
xlvi
List of Abbreviations
RN
ROC
Recovery of Consciousness
RR
SBP
SBPbaseline
SD
Standard-Deviation
SE
State Entropy
SEP
SI
SII
SISO
SPI
SS
Steady-State
SSI
TCI
TSS
TTPE
VAS
VEP
WT
Wavelet Transform
xlvii
List of Abbreviations
xlviii
Chapter 1
Introduction
Anaesthesiology is a medical specialty that has greatly evolved in the last 50 to 60 years,
from the simple idea of inducing pain relief and unconsciousness for a clinical procedure
to occur, to a complex set of concepts and interactions, with action in a wide range of
clinical procedures [1].
Nowadays, anaesthesia involves sophisticated technology, administration of multiple
drugs, trained professionals and demanding regulations. From being a very unsafe and
simple procedure, requiring almost no specialized skills, equipment or monitoring (just
a few drops of ether poured over the face were enough to provide anaesthetic state), to
a complex and high technology procedure.
1. INTRODUCTION
and sensible to painful stimulation. Besides being a traumatic and terrifying experience,
these interventions had a high risk of complications, during and following the procedure.
Aware of these facts, the concept of producing insensibility was quite appealing for
both patients and clinicians, improving patients' wellbeing, and also allowing surgical
technical developments with superior quality and success.
Figure 1.1: First public demonstration of diethyl ether general anaesthesia at the Massachusetts General Hospital, on October 16, 1846. Recreation of the event by Robert
Cutler Hinckley completed in Paris in 1892 [2].
16
1846) was wrapped in controversy, due to the use of a secret compound which was in fact
pure unadulterated sulfuric ether. Although ether use was registered two years before,
for short dentistry procedures, however only in the referred public demonstration, a
longer surgical intervention took place at the Massachusetts General Hospital (see Figure
1.1) [2, 3].
Later, in 1847, the use of chloroform was introduced and the concept of general
anaesthesia began to gain form. Although this technique showed great advantages compared to previous approaches, there were (and still are) associated risks that soon proved
to be the worst possible. In January 1848, soon after the introduction of chloroform for
anaesthetic use, the rst of many fatalities occurred: a 15-year-old girl died while receiving a chloroform anaesthesia [4], showing that it required careful titration, beginning
the long road of drug development, patient monitoring and eects' balance, which the
anaesthesiologist still needs to go through.
Figure 1.2: General anaesthesia components and corresponding informative physiological signals.
Electroencephalogram
tors give information on the patient's state regarding the hypnosis component. Usually
1. INTRODUCTION
these monitors translate the information contained in the EEG patterns into simple and
easy to use indexes, varying from 0 (isoelectric EEG) to 100 (fully awake patient) [6].
For the paralysis component, neuromuscular monitors allow to assess the state of the
patient regarding muscle relaxation, through the evaluation of the response to dierent
electrical stimulation patterns and currents, using movement sensors (accerelerometry)
or
Electromyography
Nociception/Anti-Nociception
(Noc/ANoc) balance in
Figure 1.3: Guedel's stages and planes of ether anaesthesia [9], withdrawn from [6].
Guedel dened the anaesthesia stages and planes presented in Figure 1.3 for ether
anaesthesia [9]; the idea was to dene clinical signs characteristic of dierent anaesthesia
levels, that would place the patient in consecutive clinical states.
However, Guedel's
signs cannot be fully applied due to the dierent action mechanisms in the brain of the
currently used anaesthetic drugs. This is worsened by the use of other agents that may
mask the physiological responses. Jameson and Sloan [6] reviewed the use of the EEG
to monitor patients hypnosis state during general anaesthesia from the rst denition of
the anaesthesia planes by Guedel, to the proliferation of EEG monitors in the operating
room.
In addition to the use of clinical signs to assess the patients' state, several studies
aimed at dening average population drug doses to achieve a certain clinical eect. The
term
1. INTRODUCTION
reaction to surgical incision [6]. The anaesthetic eect that causes immobility occurs in
the spinal cord, not guaranteeing adequate cortical eect, and with the use of muscle
relaxants the use of movement as a sign of adequate anaesthesia has decreased, however
the MAC that causes 50% of patients to be unconsciousness is lower than the MAC
for immobility. Other anaesthetic agents (intravenous agents) may not have the same
action, and the use of MAC initially developed in volatile anaesthesia, was replaced by
the
Minimum Infusion Rate (MIR). Althoug the MAC and MIR are usefull guides, they
Expected Response
Score
(Alert) 5
(Unconscious) 0
Observer
This score scheme was developed and used to relate the EEG patterns with dierent
sedation levels. However, its use is limited to light sedation levels, since it depends on
the subject's collaboration to verbal and mechanical stimulation. The core idea of this
approach is to determine levels of sedation, to identify and classify EEG patterns on
each level.
Although clinical assessment of patient's state, and denition of optimum drug doses, are essential to deliver anaesthesia, the existence of objective and reproducible
methods to assess the hypnosis state of a patient is quite appealing. Thus several tools
have been developed based both on EEG and derived signals, such as the
Evoked Potentials
Auditory
each sedation/anesthesia state, to construct valid indexes, that are correlated to the
clinical state of the patient (clinical endpoints). These indexes improved the way how
anaesthesia is delivered allowing more secure and tailored drug administration, aiding
Bispectral Index
TM,
(BIS
processing techniques over the frontal EEG, and transform the information contained in
the original chaotic signal, into easy to understand indexes. Although both indexes have
the same manufacturer recommended target range for general anaesthesia, they dier in
the signal processing techniques employed in the index calculation. Moreover, although
both monitors try to prevent the impact of outliers and outer interference, they are subjected to factors that inuence the extracted indexes, such as EMG, electromechanical
artifacts (e.g. electrocautery or forced-air warmers), abnormal EEG states, or certain
anesthetic agents' action [14].
measures of the EMG and BIS index, as described by Greenwald and Rosow [16], varying
1. INTRODUCTION
Figure 1.4: Bispectral index monitor during data collection: manufacturer recommended target range for general anaesthesia of 40 to 60.
from 0 to 10. Figure 1.4 shows the BIS bilateral monitor during a general anaesthesia,
with both BIS and CVI trends for the left EEG channel. The CVI was also used in this
thesis, thanks to the collaboration of Aspect Medical Systems.
A bilateral BIS was provided for acquisition of the new index. The Bilateral BIS
monitor allows the EEG collection from both sides of the frontal cortex, providing two
EEG channels (left and right), allowing asymmetry assessment which is an indicator of
central ischemia.
the
State Entropy
Response Entropy
1.2 Pain
The information contained in the RE and SE, similar to what was described for the
BIS monitor, has been linked to noxious activation [20, 21].
1.2 Pain
The physical adjunct of an imperative protective reex.
Sherrington
According to the
It is therefore a
external and internal environments through a variety of sensory receptors which activate
the central nervous system.
energy from the environment into action potentials in the neurons [23, 24].
Figure 1.5: Schematic representation of the sensory process in the human body, redrawn
from [24].
1. INTRODUCTION
A sensation essentially involves sampling selected small amounts of energy from the
environment, and using it to control the generation of trains of action potentials as
represented in Figure 1.5.
The sensory receptors are activated in response to dierent forms of energy, which
requires dierent specialized types of receptors. Table 1.2, reproduced from [23], shows
dierent sensory modalities present in the human body, from which the rst eleven are
conscious; there are however other sensory receptors involved in information processing
that do not imply consciousness.
In the human body, the way a stimulus is perceived is conditioned by the sensing grid,
but also by the intrinsic mechanisms transmitting the information. As an example of the
rst, the cutaneous sense organs translate four dierent sensations, touch/pressure, cold,
warmth and pain. The sensors present in the skin for all these sensations are mapped
according to a grid that is not uniform, and one can only sense the stimulus if an area
of the mapping sensor is elicited. As an example of the second, if a stimulus of constant
strength is applied to a receptor, an adaptation phenomenon occurs, making the action
potentials of the sensory nerve decline over time.
occurs in dierent ways depending on the sensor nerve involved in the process. Another
principle implicated in sensing is the law of projection. By this law, independently of
where the sensory pathway is stimulated, the conscious sensation produced is referred
to the location of the receptor [23].
An important issue in the sensory process is the stimulus intensity discrimination.
The way stimulus intensity is perceived is not linear and the way the information is
sent to the brain depends on two factors: the frequency of the action potentials, and
the number of receptors activated. The magnitude of the perceived sensation used to
be described by the Weber-Fechner law [23], which states that the magnitude of the
sensation is proportional to the logarithm of the stimulus intensity. The Weber-Fechner
law was replaced by the power function, as presented in Equation 1.1.
R = KS A
(1.1)
where R is the sensation felt, S is the intensity of the stimulus, K and A are constants
[23].
A power function is also used to describe the relation between frequency of the action
potentials generated in a sensory nerve ber and the intensity of the initiating stimulus.
10
1.2 Pain
Table 1.2: Sensory modalities and corresponding receptors and sense organs, reproduced
from [23].
Sensory Modality
Receptor
Sense Organ
Vision
Eye
Hearing
Hair cells
Smell
Olfactory neurons
Olfactory
mucous
mem-
brane
Taste
Taste bud
Rotational
Hair cells
Linear acceleration
Hair cells
Touch-pressure
Nerve endings
Warmth
Nerve endings
Cold
Nerve endings
Pain
Nerve endings
Nerve endings
Various
Muscle length
Nerve endings
Muscle spindle
Muscle tension
Nerve endings
Arterial blood
Nerve endings
acceleration
movement
pressure
Central venous
pressure
Ination of lung
Nerve endings
Temperature of
Neurons in hypothalamus
blood in head
Arterial PO2
Glomus cells
pH of CSF
Osmotic pressure of
Cells
plasma
bly
in
other
OVLT
and
possi-
circumventricular
glucose dierence
costats)
11
1. INTRODUCTION
As the strength of a stimulus is increased not only the sense organs in immediate contact
with the stimulus are activated, but also sense organs in the surrounding area become
activated, resulting in a spread eect [24].
Fiber Type
Function
Fiber
Conduction
Spike
Absolute
Diameter
Velocity
Duration
Refractory
(m)
(m/s)
(ms)
Period
12-20
70-120
0.4-0.5
0.4-1
5-12
30-70
3-6
15-30
2-5
<3
12-30
3-15
1.2
1.2
0.4-1.2
0.5-2
0.3-1.3
0.7-2.3
Proprioception;
somatic motor
Touch, pressure,
motor
Motor to muscle
spindles
Pain, cold, touch
Preganglionic
autonomic
Dorsal root
Sympathetic
Pain, temperature,
some
mechanoreception,
reex responses
Postganglionic
sympathetic
The human central nervous system (brain and spinal cord) is a complex system that
contains about
1011
network [24]. The types of nerve bers that carry impulses generated in them to the
central nervous system are presented in Table 1.3. Some of these bers are covered with
myelin and convey information at faster rates.
The spinal cord also has a complex structure.
allows the identication of its most important structures: it is divided in the so called
horns, due to its shape. Dorsal horns are divided into laminas I-VII, with I being the
most supercial and VII the deepest (see Figure 1.6).
12
1.2 Pain
Figure 1.6: Schematic representation of the dorsal horn of the spinal cord layers and
terminations of the three types of primary aerent neurons pathways, redrawn from
[23].
2. Small myelinated A bers, some of which transmit impulses from either cold
receptors, nociceptors or mechanoreceptors;
3. Unmyelinated C bers that are concerned primarily with pain and temperature;
some C bers also transmit impulses from mechanoreceptors.
Cerebral cortex areas related to the cutaneous senses are shown in Figure 1.7; the
areas involved in sensation in the cortex have been studied clinically using
Positron
Emission Tomography (PET) and functional Magnetic Resonance Imaging (fMRI). The
use of these technologies has increased the knowledge about the ways information is
processed in the human brain.
13
1. INTRODUCTION
Figure 1.7: Brain areas involved in somatic sensation (withdrawn from [24]).
The neuronal network is not innate or immutable, and can rapidly recongure itself
by experience to reect the use of the represented area.
explained by the cortical connections of sensory units to the cortex, which have extensive
convergence and divergence, with connections that can become weak with disuse and
strong with use [23].
Some of the axons of the dorsal horn neurons end in the spinal cord and
brainstem, others enter the ventrolateral system, including the lateral spinothalamic
tract, a few ascend in the dorsal portion of the cord.
project to the ventral posterior nuclei, which are the specic sensory relay nuclei of the
14
1.2 Pain
thalamus, and from there to the cerebral cortex.
somatosensory cortices ), and the cingulate gyrus of the side opposite the stimulus.
In
addition, the mediofrontal cortex, the insular cortex and the cerebellum are activated
[23, 24].
As a previous step to design a tool capable of translating the balance Noc/ANoc
during general anaesthesia, it is important to understand the inherent mechanisms
and pathways from the sensor activation to the central processing of pain potentials.
This means to understand and dene the physiological signals which could be used in
translating nociception objectively.
Pain may be described considering some studied features, such as conduction velocity, site of occurrence, and pathways [2325]. Impulses resulting from noxious activation are transmitted through two nerve ber systems, resulting in the physiological
observation of two dierent pain sensations: one sharp and localized, the other dull,
intense and diuse, usually referred as fast and slow pain, or rst and second pain.
Observations indicate that the further away from the brain stimulation is applied, the
greater the time lag between fast and slow pain.
that the fast pain is due to the activation of the A -bers and the second due to the
activation of the C pain bers.
Enervation in body tissues is not equally distributed, conducting to dierences in
the nature of the evoked pain for supercial and deep structures.
are poor in A nerve bers, inducing little rapid sharp pain.
Deep structures
localized, frequently associated with nausea and autonomic symptoms like sweating
and changes in blood pressure.
those in the skin, however there are no proprioceptors, and few temperature and touch
receptors. In the central nervous system visceral and somatic sensations travel along
the same pathway, and cortical receiving areas of visceral sensation are intermixed with
the somatic receiving areas.
A dierent specic pain modality is muscle pain, resultant from contraction and
blood supply occlusion. It is thought that the muscle pain is the result of the accumulation of a chemical agent known as the Lewis P factor, and that as soon as the blood
supply is reestablished, it is washed out and the pain disappears.
Visceral and deep somatic pain initiate a reex contraction of nearby skeletal muscle,
more frequent in the abdominal wall, originated to protect the structures beneath from
15
1. INTRODUCTION
trauma.
Irritation of a viscus producing pain may not always be felt in the viscus,
but in some referred somatic structure; this pain is said to be referred to the somatic
structure. Deep somatic pain may also be referred, but supercial pain is not. When
the pain is referred it usually is associated to a structure that has evolved from the same
embryonic segment or dermatome (dermatomal rule). This is the reason why pain from
cardiac ischemia is usually felt in the arms (besides the chest).
Pain may be centrally inhibited; examples of this are stress analgesia in battle
wounded soldiers, or the occasions when touching and shaking an injured area leads
to pain relief. The last, called central inhibition is explained by the inhibition of pain
pathways in the dorsal horn gate, by stimulating large diameter touch-pressure aerents.
Inammatory processes also perform important roles in pain sensing, leading to
hyperalgesia (minor stimuli produce an exaggerated response) and allodynia (normally
innocuous stimuli as light touch cause pain); this is the result of the liberation of
cytokines and growth factors, referred to as the inammatory soup in the inamed
area, which facilitate perception and transmission in cutaneous areas and in the dorsal
horn [23] .
Finally, another important phenomenon in pain sensing is the pain resulting from
injury in the nerve bers themselves, called neuropathic pain, which is usually accompanied by hyperalgesia and allodynia. It is a dicult condition to treat and occurs in
various forms [23, 24].
16
1.2 Pain
Table 1.4: Stimulus modalities for induction of experimental cutaneous pain in humans,
adapted from [26].
Modality
Stimulus
Electrical
Transcutaneous
Intracutaneous
Intraneural
Thermal
Heat
Cold
Ice water
Contact thermode
Mechanical
Pinprick
Pressure
Impact Stimuli
Pinch
Chemical
Capsaicin
Mustard oil
Hydrogen ions
Inammatory-related substances (serotonin, bradykinin, histamine,
potassium ions, substance P)
trying to translate the way humans perceive pain, and the way the sensation may be
objectively described.
(VAS),
In the rst method the subject is asked to rate pain by positioning a cursor along a ruler;
this ruler has the left side extremity marked as No Pain, and the right side extremity
identied as Very Severe Pain. The back of the ruler, not seen by the patient, usually
has a scale from 0 to 10 or 100, allowing the measurement of relative positions for pain
assessment. In the second method a numerical rating scale usually varying from 0 (no
pain) to 10 or 100 (worst possible pain) is explained to the subject who auto-evaluates
his/her pain in the ordinal scale. In the last method, the subject is asked to classify
his/her pain in one of the verbal descriptors, usually translated into an ordinal scale. A
17
1. INTRODUCTION
Table 1.5: Stimulus modalities for induction of muscular pain in humans, reproduced
from [26].
Type
Modality
Stimulus
Endogenous
Ischemia
Tourniquet + contractions
Exercise
Concentric contractions
Eccentric contractions
Exogenous
Electrical
Intramuscular
Intraneural
Mechanical
Pressure
Chemical
Hypertonic saline,
potassium,
levoascorbic acid,
capsaicin bradykinin, serotonin, calcitonin generelated peptide, neurokinin A, substance P, hydrogen ions
Table 1.6: Stimulus modalities for induction of visceral pain in humans, reproduced
from [26].
Modality
Stimulus
Electrical
Transmucosal
Thermal
Heat
Water
Cold
Water
Mechanical
Distension
Chemical
Capsaicin
schematic representation of these pain assessment techniques are shown in Figure 1.8.
Other methods of pain evaluation using the subject's participation are the multimodal questionnaires, in which pain is evaluated as a multifaceted process.
As an
example, the McGill Pain Questionnaire, also known as McGill pain index, is a scale
for pain rating developed at the McGill University, which is used to evaluate pain as a
multifaceted phenomenon using Psychology techniques (see Figure 1.9) [29]. This is one
of the developed methods to evaluate pain in patients able to cooperate, describe pain
sensation and activity limitation; several were developed for specic recurrent chronic
18
1.2 Pain
Figure 1.8: Schematic representation of pain assessment techniques. From top to bottom: visual analogue scale, numerical rating scale (0 meaning no pain and 10 worst
possible pain), and a verbal scale.
nevertheless to use this method it is important to have the document translated, adapted and validated to the subjects' mother language. In [33], the original McGill pain
questionnaire was translated and adapted to Portuguese (Brazil), trying to maintain
the original objective of the questionnaire, and adapting some of the pain descriptors
to better t the subjects. Figure 1.9 shows the original (on left) and adapted (on right)
versions of the McGill Pain Questionnaire.
The use of the described rating scales is limited by patients' collaboration and selfevaluation of pain sensation. In the case of the multifaceted questionnaires they have
practical applicability only in the assessment of chronic pain, with no practical use
in acute pain patients (time consuming).
19
1. INTRODUCTION
Figure 1.9: McGill pain questionnaire: on left the original version in English, and on
right the adapted version in Portuguese (Brazil), withdrawn from [33].
in many demanding practical clinical situations, such as in patients unable to communicate. This reveals the need in the clinical setting for a tool to objectively translate
the patient's state concerning pain or Noc/ANoc balance. This is the case for patients
under general anaesthesia, subjected to severe surgical noxious stimulation, and unable
to express the Noc/ANoc balance state. Objective methods for nociception assessment
are discussed in section 1.4.
General anaesthetics are potent drugs that depress the central nervous system; they
are used to allow patients to be subjected to otherwise painful or very unpleasant
procedures, namely surgery [34].
care of a specialized physician, severe noxious stimuli may be rendered virtually harmless. Anaesthesiologists titrate a set of drugs according to specic requirements of each
20
1.3.1 Analgesics
Analgesics are potent anaesthetic drugs used to abolish responses to noxious stimulation, meaning nociception [34].
according to patients' needs, and manually titrated to obtain homeostasis in conjunction with other anaesthetic agents.
prevent adverse eects, and only by experienced personnel properly assisted by all the
required equipment to quickly respond to a crisis situation [35].
Opioid analgesics are currently used in clinical practice to produce analgesia. The
term
opioid
the poppy ower. Well known examples of drugs derived from opium include morphine,
codeine, pethidine, and many semisynthetic derivatives. Opioids mimic the endogenous
opioid peptides endorphins, the naturally occurring ligands for opioid receptors, that
are part of the endogenous opioid system [34].
1.3.1.1 Remifentanil
Figure 1.10:
21
1. INTRODUCTION
Remifentanil, also known by its molecular description methyl 1- (3-methoxy -3oxopropyl) -4- (N-phenylpropanamido) piperidine -4- carboxylate, (see Figure 1.10) is a
fast acting opioid drug, used to provide analgesia to patients under general anaesthesia
or in the intensive care unit under mechanical ventilation [35, 37].
The endogenous opioid system plays an important role in the sensory processes, by
inhibiting responses to noxious stimuli. It has also impact in gastrointestinal, endocrine
and autonomic functions, and emotional and cognitive roles in the modulation of learning and memory. There are four major opioid receptor types:
, k ,
and N/OFQ
[34].
Remifentanil is a selective
tion. The
-opioid
-opioid
in adult healthy patients is of about 2 hours. In patients with normal renal function
the time for 95% of renal elimination of the primary metabolite of remifentanil is of 7
to 10 hours. Due to its very short action duration, the side eects of remifentanil are
limited to the immediate period following administration. Interruption or cessation of
drug administration leads to the return to the initial state in 10 minutes [34, 35, 37].
Remifentanil use is approved by the Portugal National Authority of Medicines and
Health Products (INFARMED), and its main characteristics are presented in Appendix
A, including side-eects and dosage recommendations. Remifentanil may be administered using either manual infusion or model based infusion, with the
and Dynamic
Pharmacokinetic
1.3.2 Hypnotics
Hypnotic drugs are used either for induction of an altered state of consciousness or
for its maintenance, and are titrated according to patient's needs and procedures' requirements. As the analgesic drugs, hypnotics are also potent anaesthetic drugs with
associated risks of overdosing (e.g. higher mortality or cognitive dysfunction incidence)
[4042] and underdosing (awareness) [12, 13].
22
1.3.2.1 Propofol
Figure 1.11: Propofol 2D molecular schematic representation (2,6 - bis(propan - 2 yl)phenol) [43].
23
1. INTRODUCTION
hypercapnia [34, 44]. Propofol decreases brain metabolism, cerebral blood ow, intracranial pressure, and presents a neuroprotective eect [44].
Propofol is the most widely used hypnotic anaesthetic.
whether propofol, besides being an hypnotic drug, could also have some analgesia eect
[45, 46]. It is known to act on a specic site on the
Figure 1.12: Drugs' action path in the human body, from drug dose administration to
observed clinical eect (adapted from [51]).
When the clinician administers a certain drug amount to a patient, a specic clinical
eect is expected; this is based on the understanding of what happens following drug
administration, meaning the drug path in the human body, as represented in Figure
1.12.
This drug action path may be resumed in the relations between administered
24
eect-site concentration
(Ce).
plasma concentration
kij ,
j (k10
V1 , V2
and
V3
to the compart-
The central compartment may be considered as the plasma compartment (although the
volume
V1
may be higher than the real plasma (blood) volume), the compartment 2
corresponding to the fast distribution compartment or vessel rich, and the compartment
3 represents the slow distribution compartment or vessel poor.
V1 , V2
and
V3
are howe-
ke0 .
The
ke0
e,
of the drug in the system (bolus). The TTPE represents the moment in which the eectsite and plasma concentrations equalize. In the studies leading to the development of
PKPD models, EEG derived parameters are used as surrogates of the Ce to assess the
clinical/dynamic eect, and determine the point of maximum eect (TTPE).
The PKPD presented in Figure 1.13 is a dynamic model, with state-space representation as presented in Equation 1.2.
25
1. INTRODUCTION
Figure 1.13: Three compartments pharmacokinetic (PK) model, followed by the pharmacodynamic (PD) model and Hill equation, translating the eect-site concentration
into the output measurable eect.
compartment
Where
kij ,
V1 , V2
and
V3
are
E0
= 0), EC50
represents the
k12 k13 k10 k21
k31
0
m 1 (t)
m 2 (t)
k
k
0
0
12
21
m 3 (t) =
k13
0
k31
0
ke0
0
0
k
Ce(t)
e0
V1
m1 (t)
1
m2 (t) 0
m3 (t) + 0
Ce(t)
0
r(t)
(1.2)
where
r(t)
mi
i (i =
1, 2, 3).
The relation between eect-site concentration and the measurable clinical eect (E )
26
E0
= 0), EC50
is the
drug eect-site concentration necessary to achieve half of the maximum eect (drug
potency), and
E(t) = E0 1
Ce(t)
Ce(t) + EC50
(1.3)
UA (t) =
CeA (t)
EC50A
(1.4)
UH (t) =
CeH (t)
EC50H
(1.5)
As the normalized concentration of analgesic (UA ) and hypnotic (UH ) are computed
by Equations 1.4 and 1.5, a family of drugs may be dened as a ratio of
UA
and
UH
(t) =
By denition,
UH (t)
UH (t) + UA (t)
(1.6)
Thus, the
concentration response relation for any ratio of the type of interaction may be described
by Equation 1.7.
E(t) = E0 1
where
E0
27
U50
(1.7)
is the
1. INTRODUCTION
The relation in Equation 1.7, and the expected response to noxious stimulation, were
employed in research studies [5860] addressing the combination of the hypnotic and
analgesic drugs' concentrations to obtain an index of probability of response to noxious
stimulation,
(NSRI).
Figure 1.14: Screenshot of the educational software specially developed for the simulation of target-controlled infusion anaesthesia using BIS or Entropy. It shows the surface
relational model between propofol and remifentanil's synergistic eect on the hypnosis
indexes [56, 57].
The bivariate Hill methodology was also employed in some studies developed by
us, nevertheless outside the context of this thesis, in the assessment of the relations
between hypnotic propofol, analgesic remifentanil and the measurable eect translated
by BIS and Entropy indexes [56, 57]. In [61, 62], data during propofol and remifentanil
anaesthesia, with BIS and Entropy monitoring were used to adjust dierent surface
model structures, and construct the synergy curves of eect, later implemented in an
education simulation tool (see Figure 1.14).
Although these models try to predict the average patient response to drug administration, there is no such thing as an average patient. The problem resides in the
known inter and intra-individual variability, which should be taken into account [54, 63]
when administering these or any drugs.
The aforementioned concepts and all the PKPD modelling, allow to implement these
28
Infusion
Target Controlled
(TCI) systems, continuously simulate the drug course in the human body,
commercialized in several countries, the United States of America Food and Drug Administration (FDA) has not yet approved it.
Two clinical studies have been conducted for the purpose of this thesis, both used
propofol as hypnotic, and remifentanil as analgesic. For both studies, TCI was applied
as method of drug administration, with the Schnider [50] and Minto [39] PKPD models.
The clinical protocols for these studies are detailed in Chapters 2 and 5. In the next
sections, the main features of these models are described.
model, and the model parameters found to be related to the demographic data of the
subjects were age, weight, height and gender.
Table 1.7 shows the relations between patient's demographic data and the estimated
model parameters, with corresponding estimated percent coecient of variation (CV).
Weight (in kg), height (in cm) and gender were used in the assessment of lean body
mass (LBM), one of the model variables, given by Equation 1.8 for men, and Equation
1.9 for women.
Men:
LBM = 1.1W eight 128
29
W eight
Height
2
(1.8)
1. INTRODUCTION
Women:
LBM = 1.07W eight 148
W eight
Height
2
(1.9)
The relation used to obtain the LBM is a quadratic expression, that produces paradoxical values for overweighted patients [64]. Pharmacokinetic models were not adjusted
or validated on obese patients, and should be carefully used in this population. Some
methods to overcome the LBM paradoxical values issue have been proposed in the commercially available TCI systems, suchlike limiting the body weight for a certain height,
guaranteeing that the LBM calculus is not on the descending portion of the quadratic
curve [65].
comprise this thesis was the three compartments Schnider PKPD model [50].
The Schnider PKPD model parameters were obtained using a population of 24
healthy adult volunteers, between 26 and 81 years of age. Tables 1.8 and 1.9 present the
estimated population parameters, adjusted to subjects' demographic data age, weight,
Table 1.7: Estimated relations for the parameters of the three compartments Minto
pharmacokinetic and dynamic model, and corresponding coecient of variation (CV).
The parameters incorporate patient's age and lean body mass (LBM) [38].
CV (%)
5.1-0.0201*(Age-40)+0.072*(LBM-55)
26
9.82-0.0811*(Age-40)+0.108*(LBM-55) (fast)
29
5.42 (slow)
66
(2.6-0.0162*(Age-40)+0.0191*(LBM-55))/V1
14
(2.05-0.0301*(Age-40))/V1
36
(0.076-0.00113*(Age-40))/V1
41
(k12*V1)/V2
(k13*V1)/V3
0.595-0.007*(Age-40)
60
30
(LBM) [50].
2 +7 *(Age-53)
3
(1=1,...,11) parameters
4.27
0.278
18.9
2.33
238
34.9
1.89
0.059
1.29
0.112
0.836
0.044
-0.391
0.07
0.0456
0.009
-0.0681
0.017
0.0264
0.009
-0.024
0.005
height and gender. LBM was employed as described for the Minto model in the previous
section (Equations 1.8 and 1.9).
The parameters obtained by the models are average population parameters, and
31
1. INTRODUCTION
exhibit variability. Although these models try to predict the drug path in the human
body, they are only estimations of the expected drug's concentrations for an average
patient.
Nociception may be dened as the aerent activity produced in the peripheral and
central nervous system by stimuli that have the potential to damage tissue [23]. Nociception triggers responses that include tachycardia, hypertension, sweating and electromyographic activity.
Every time a noxious/painful stimulus is applied, the human body reacts, alerting
the organism to avoid or stop the noxious stimulus source.
a phenomenon known as
32
1.4 State of the Art on the Analysis and Control of the Nociception Level
order to give a comprehensive analysis of this topic, the following sections are organized
according to the physiological signals employed in the search of a Noc/ANoc balance
index.
Most of the proposed approaches to the assessment of the Noc/ANoc balance use
pre-processed indexes, and are unitless.
Figure 1.15: Results obtained by Seitsonen and colleagues, withdrawn from [20]: Relative (to prestimulus interval) group average ( standard error of mean) responses to
skin incision in movers (dashed line) and non-movers (solid line) of (A) EEG response
entropy (n=10 and n=12 for movers and non-movers, respectively), (B) RRI (n=12 and
n=14), (C) PPG amplitude (n=12 and n=14), and (D) PPG notch amplitude (n=12
and n=14). Skin incision marked with t=0.
33
1. INTRODUCTION
Seitsonen et al. [20] found a relation between several variables and the nociceptive
response to skin incision during general anaesthesia.
Electrocardiogram
(ECG),
Pulse
distinguished between patients who moved following incision and those who did not,
was composed by the Response Entropy of the EEG (RE, Entropy hypnosis monitor),
the
time interval between ECG QRS complexes (RR interval) and PPG notch amplitude
Clinical Signs-Stimulus-Antinociception
(CSSA)
score. This score was a combination of patients' reactions and clinical signs of nociception, remifentanil levels and an estimative of noxious intensity of incision. The proposed
scale was based on few experts opinion. Dierences between post and pre-incision values of
Heart Rate Variability (HRV), PPG and pulse transition time related parameters
were analyzed o-line to evidence the best predictors of CSSA. Those best predictors of
CSSA were used to develop a so called
0 to 100. This index was later tested in a dierent data set composed by 10 patients'
data. The authors concluded that HRV, RE, RE-SE and PPG variability were the best
predictors of CSSA [77].
In a dierent study by Huiku et al [78], a similar approach was used in the assessment of surgical stress, where nger photoplethysmography and electrocardiography
waveforms were recorded and various waveform parameters extracted o-line. The
To-
remifentanil concentration.
A so called
ped to correlate with the TSS estimate and the performance of SSI validated within
a validation data set [78].
Beat Interval
(HBI) and
Heart
(PPGA), expressed
(1.10)
According to the authors, in this study the developed index (SSI) responded to
changes in the remifentanil level and to dierent nociceptive stimuli.
34
The proposed
1.4 State of the Art on the Analysis and Control of the Nociception Level
index was then used in dierent clinical studies [8085] to verify its applicability, seeming
to reect changes in surgical stress and analgesic dose.
In a study by Struys et al. [83], the SSI was monitored under general anaesthesia for
dierent combinations of propofol and remifentanil target eect-site concentrations [78].
At each steady-state of the remifentanil eect-site concentration, the maximum change
of the SSI, SE, RE and PPGA was compared with its baseline value. It was found that
the static and dynamic values of SSI correlated better to the remifentanil eect-site
concentration than the other recorded variables, and that this was independent of the
hypnotic eect-site concentration.
Recently a new method was proposed in [86], also using PPG amplitude (P P GA)
and pulse-to-pulse interval (P P I ), the so called
(AN SSI ). According to the authors the proposed index adequately reected stimulation and was able to dierentiate between patients who received an inltration with a
local anaesthetic in the tonsils, from patients who did not, demonstrating a greater amplitude response when compared to the
to as SSI).
(1.11)
AN SS = P P I P P GA
(1.12)
where
and
AN SSmax
is the maximum
AN SS ,
ent.
ANSSI is an open source index, and may be computed by Equations 1.11 and 1.12.
Some authors [15, 16, 8789] have related hypnosis indexes variability to the nociceptive response of anaesthetized patients, using BIS, Entropy indexes (SE, RE) and
(AAI).
Figure 1.16 shows a typical trend of the Bispectral Index (hypnosis index) during general anaesthesia, with the descending phase (anaesthesia induction), the steady phase
(maintenance), and the nal ascending phase (recovery). Examining the index during
the maintenance phase, while the patient is subjected to surgical noxious stimulation, it
may be observed that the index has an intrinsic variability. This variability may be affected by several external factors. The possibility that BIS variability could be aected
35
1. INTRODUCTION
Figure 1.16: Bispectral Index (BIS) trend during general anaesthesia, with the induction, maintenance and recovery phases. The dashed lines represent the BIS manufacturer recommended target range for general anaesthesia (40-60). Data from one of the
patients in this study.
by the Noc/ANoc balance was explored [15, 17]. The core idea behind this approach
is to use the apparently random variability of the signal, to translate the nociceptive
responses of an anaesthetized patient, based on the arousal responses induced in the
EEG by noxious stimulation.
Noxious stimulation activates peripheral nociceptors which send aerent signals via
the spinal cord, midbrain and thalamus to a broad distribution of regions in the cerebral
cortex.
stimulate the cerebral cortex and elicit changes in the EEG [15].
Following this line of thought, one of the developed indexes is the so called Composite
Variability Index (CVI) [16]. It was developed as a logistic regression based on the BIS
index and frontal EMG activity, recorded by the same BIS monitor.
The authors
in [15, 16] related the increase of the standard-deviation of each signal and of EMG
power to somatic responses, and demonstrated that early detection of arousal could
be achieved this way. In another study [17], the authors also observed that CVI, BIS
standard-deviation, but not EMG standard-deviation increased in response to surgical
incision, with slightly increased amplitude in the contralateral site to stimulation. A
bilateral sensor was used in these studies.
Entropy indexes were also proposed as surrogate measures of nociception [90].
study from Takamatsu et al. [89] employed Entropy monitoring to assess the depth of
anaesthesia, and used the dierence between the two indexes provided in this monitor
36
1.4 State of the Art on the Analysis and Control of the Nociception Level
- the State Entropy (SE) and the Response Entropy (RE) - to estimate nociception
during general anaesthesia.
the dierence between RE and SE, however, it did not always indicate inadequate
analgesia.
In another study [91], authors observed that RE increased in response to tracheal
intubation, while SE and BIS did not, and also that landiolol (an antiarrhythmic) and
remifentanil suppressed this increase.
by Kawaguchi et al. [92], where the authors demonstrated that the dierence between
SE and RE was suppressed by increasing the muscle relaxant, meaning that it diers
according to the paralysis state of the patient. In a dierent study conducted by Valjus
et al. [93], the authors concluded that RE did not seem to be more sensitive than SE to
guide opioid administration, and Weil et al. [94] observed that an increase in Entropy
predicted motor response to intubation and incision, but not a haemodynamic response.
Although the methodology presents several limitations, specially for paralyzed patients,
a study by Mathews et al. [21] proposed an automatic algorithm for the administration
of remifentanil based on the RE-SE dierence, demonstrating its feasibility.
It should be highlighted that Entropy based, BIS based and PPG wave based indexes, were in the majority studied through the direct initiative or direct support of the
monitors' developers. Nevertheless, from all the indexes described to this point, only
SPI became commercially available.
Some studies [95, 96] using AAI monitoring also proposed a method that used this
index to translate the nociceptive response of an anaesthetized patient. The dierences
between this index and the previously referred are the site of signal collection, postprocessing and involved processes.
cortical activity, while the AAI index is obtained as a response to an auditory stimulus, corresponding to the transfer of auditory information from the medial geniculate
body to the primary auditory cortex (closer to the somatosensory area). This method
was evaluated in [95], demonstrating that during a BIS guided anaesthesia, the AAI
response diered according to the analgesic regimen, and it could be used to translate
the anti-nociception of a patient under balanced anaesthesia.
A dierent approach was proposed by Cividjian et al. [97], where beat-by-beat blood
pressure and heart rate (Finapress
R
measurements), were used to assess the baroreex.
Nociception is known to depress cardiac baroreex, and the authors proposed a new
index, the
37
T M ), varying from 0
index (CARDEAN
1. INTRODUCTION
to 100, that reects baroreex inhibition [97]. It was also demonstrated that CARDEAN
allows to predict intraoperative movement better than BIS [97, 98], and its use reduced
the occurrence of unexpected movement during endoscopy [99, 100].
In a dierent
study [101], CARDEAN was directly associated to stimulus intensity and inversely to
analgesic concentration.
Figure 1.17: Facial muscles involved in the innate response to noxious stimulation [102].
A study conducted by Bennet et al. [103], explored the human facial expression in
response to painful stimuli.
muscles as shown in Figure 1.17. This study the autonomic facial expression response
to noxious stimulation as a measure of Noc/ANoc balance.
The innate face expression to noxious stimuli has well dened patterns according to
the sensation, with dierent muscles contraction and intensities. Basically this study
[103] suggested that to measure electromyography activity based on two specic muscles - the
orbicularis oculi
and the
corrugator
corrugator muscle
these two and the frontalis
over the
orbicularis oculi
R=
(C F ) + (O F )
C + O + 2F
38
(1.13)
1.4 State of the Art on the Analysis and Control of the Nociception Level
The collection of physiological signals such as facial EMG must be driven with
caution due to the possible internal and external interferences, suchlike cable motion
artifacts [5, 104, 105].
The developed
monitor displays the number of peaks per second of skin conductance bursts, and the
area under the curve.
Skin conductance has been applied to asses postoperative pain intensity and in response to standardized stimuli, responding adequately to stimulation [110112], however
its use during general anaesthesia has some limitations. In a study conducted in children, the monitor was not able to distinguish evaluations in the NRS, showing low
correlation between the NRS and the number of skin uctuations [113].
The Stress Detector monitor application was also explored by us. Initial usage in
anaesthetized patients did not show promising results, and actually presented technical
problems (unpublished data).
39
1. INTRODUCTION
the nociceptive response to noxious stimulation and the impact of analgesic drugs to
abolish these reexes.
Barvais et al. [115] explored the changes in pupil diameter and reactivity to noxious stimulation, and observed that increasing remifentanil doses decreased pupillary
responses. In a dierent study [116], the authors explored the pupillary response and
correspondent delay under general anaesthesia, inspecting if the reex was conducted by
slow C bers (increased delay for distant sites of stimulation). The authors concluded
that there was no dierence in the time delays and therefore the reex is not conducted
by slow C bers, observing that higher doses of fentanyl blocked the pupillary reex.
Although pupillary reactions are related to nociceptive responses, this approach has
limited applicability due to its potential for eye damage.
Evoked Po-
tentials (EP), is broadly used in clinical neurophysiology to determine the state of nerve
conduction for dierent stimulation sources such as visual, auditory and somatosensory.
The somatosensory evoked potentials are of interest for the problem of nociception measurement, and represent a more active approach in pain response measurement, dierent
from the passive methods proposed so far.
As part of this thesis addresses a dierent perspective on nociception measurement,
involving active interference of the system, it is treated separately in Chapter 5
Evoked
Potentials - Active Nociception Measures, were its background, specic state of the art
and clinical protocol designed for data collection in a volunteers' study, are described
in detail.
Most of the developed research in this topic is concerned with the impact
of anaesthetics on the evoked potentials, not for their potential information on the
Noc/ANoc balance, but rather as an interference to the wave used to monitor nerve
40
1.4 State of the Art on the Analysis and Control of the Nociception Level
conduction.
Also, most of the studies suggesting the use of evoked potentials to the
pain assessment are antique, and only few address the topic during general anaesthesia.
Figure 1.18: Control structure for a complete automatic administration of general anaesthesia, considering its three components: hypnosis, analgesia and muscle paralysis.
The future of
41
1. INTRODUCTION
Automatic drug administration may improve drug titration and combination, maintaining homeostasis during the intervention and improving patient outcome.
Some
studies [55, 121127] have addressed this issue for the paralysis and hypnosis control,
however there is still not a way of closing the loop for the analgesia component. Knowing
this, the development of a nociception index would bring a way of closing the loop for
the remaining component of general anaesthesia.
A nociception index would most likely lead to a totally independent drug delivery
or advisory system. Combined to a robust control algorithm it would adjust to each
patient the drug doses, according to his/her needs throughout the procedure.
Most or all of the observed physical and chemical phenomena are nonlinear, and
although linear control techniques are well dened, its required applicability conditions
usually can not be met. A good example of nonlinearity is the human body, the drug
interactions and respective nal eect. Due to the adverse consequences that human
error can carry, it is very important to know and estimate the drug processing stages in
the human system. Nonlinear control may be a good approach for the three anaesthesia
components multivariate control, due to their nonlinear course in the human body, and
for being a regulatory control for each component, taking in consideration the known
deviations from the average patient [55, 128].
To maintain the patient in an optimum state, drugs are used in the correct amount
at the right moment, to minimize risks of sub or overdosing. Each case has negative
consequences that need to be minimized for maximum satisfaction and rapid recovery
[13, 40]. The importance of guaranteeing the use of the strictly necessary drug amount
would be translated in a lower incidence of side eects and lesser costs in anaesthesia.
Although for the analgesia level, monitors to assess the nociceptive responses of a
patient are still being developed, attempts have been made using the dierence between
Response and State Entropy, to automatically guide opioid administration [21]; recently
a controller for co-administration of propofol and remifentanil (hypnotic and analgesic)
based on BIS has been proposed and tested in a multi-center study [127].
Although
feasibility of these methods was demonstrated, because the Noc/ANoc assessment presents limited reliability, the control methods present practical limitations and limited
performances in the analgesic administration.
Until a monitor of the Noc/ANoc balance is validated, automatic analgesic administration may be feasible, but with limitations. In fact, considering the model of the
42
ke0
this study, using an EEG derived measure to assess the opioid maximum eect, this may
not adequately translate the maximum drug eect directly on the Noc/ANoc balance,
and even more if one considers that depending on the eect measure used, dierent
delays are associated and should be regarded in the dynamic model adjustment for the
controller design. Further, the relation between the Ce and the measured eect is still
unknown (maybe a sigmoidal relation), both the structure and inter-patient variability
in the adjusted parameters need to be studied in order to produce a full patient model
from the drug dose to the nal measurable eect.
Current research [129] suggests that soon it will be possible to control the three
main drugs for general anaesthesia (see Figure 1.18), allowing a more secure and stable
state for each component. These systems must be robust in the detection of monitoring
errors and unexpected events like hemorrhage or hemodilution [123], and the relation
between drugs (e.g. synergy between analgesic and hypnotics), must also be taken into
account and modelled.
Anaesthesia is a medical eld that has rapidly evolved in recent years accompanying the
technological developments and the tools they carried. General anaesthesia is a complex
combination of drugs used to induce a state of hypnosis, analgesia and paralysis, and
depends on adequate monitoring to provide to the patient the best care. Hypnosis and
muscle paralysis components may be monitored, since several monitors for these anaesthesia components are available, informing the clinician on patients' state throughout
the procedure, aiding to prevent inadequate dosing and possible awareness, which may
be worsened by conscious pain perception during the clinical procedure. Nevertheless
there is not still an available monitor fully accepted to translate the Noc/ANoc balance
during general anaesthesia.
43
1. INTRODUCTION
Pain is a complex and multifaceted process, necessary for the organism preservation
and survival. For years, the physiological process behind pain sensing has been a theme
of great research interest due to the increasing need of managing pain therapies, and
improving patient welfare. The understanding of pain pathways, central processes and
inhibitory drugs' actions, is fundamental in the search for a tool capable of translating
the Noc/ANoc balance.
Although several studies have already addressed the nociceptive response during
general anaesthesia, there is still no broadly accepted index to translate this anaesthesia
component, nor clinically broadly used. The proposed indexes so far present limitations
in the denition of the adequate state of Noc/ANoc balance. Studies suggested that
a multivariate approach is necessary for a robust nociception index. To this moment
SPI, and Stress Detector are the only developed indexes commercially available. ANSSI
is an open source index, available for individual implementation, and CVI is available
only for clinical studies.
Similar to the process of obtaining a valid hypnosis index, a Noc/ANoc indicator
should be able to dierentiate levels of responsiveness to painful stimulation.
This
process necessarily involves not only the painful stimulus intensity assessment, but also
the impact on the sensory responses of dierent levels of analgesic drug.
44
Figure 1.20:
A possible structure for the conception of a Noc/ANoc balance index is proposed and
shown in Figure 1.19. An index that follows this structure is able to translate the patient
sensitivity to the stimulation, meaning the sensitivity to a stimulus while receiving a
determined anaesthetic drug dose (see Figure 1.20).
to explore passive and active measures linked to the Noc/ANoc balance during
general anaesthesia;
45
1. INTRODUCTION
to implement the Noc/ANoc developed index in a system to aid the anaesthesiologist control the analgesic administration and, in conjunction with the information
provided by the monitors present in the operating room, maintain patient's homeostasis.
This thesis is organized as follows: Chapter 2 to 4 present the study of passive nociception measures in anaesthetized patients during general anaesthesia, Chapter 5 is
focused on active measures of nociception on anaesthetized volunteers using somatosensory evoked potentials, Chapter 6 discusses the control in anaesthesia, and Chapter 7
presents the thesis global conclusions.
46
Chapter 2
Data Collection and Pre-Processing
This chapter is focused on a clinical protocol design to collect data under general anaesthesia. It also includes the submission for institutional and ethics committee approval,
informed consent management, and data collection. The data were collected in the urology operating room of the Hospital de Santo Antnio (Centro Hospitalar do Porto),
located in Porto-Portugal, following approval and written informed consent. The structure of the data les collected is here presented, as well as the pre-processing techniques
applied to the wave data les collected, in order to extract the information of interest.
2.1 Introduction
Responses to noxious stimulation are widely described in the literature, expressed in
dierent physiological signals such as heart rate, blood pressure, skin conductance,
electromyography, pupil diameter, pulse wave and evoked responses. In order to fulll
the thesis objectives, a clinical protocol was designed to collect as much usable data as
possible, and exploit the relations between the physiological signals, in the search for
a tool to translate the clinical Noc/ANoc balance of the patient. The present chapter
47
48
evaluate the eect of changes in surgical stimulation and drug doses on the physiological signals during the intervention;
evaluate metrics proposed in the literature, and propose an adequate measure for
the Noc/ANoc balance during general anaesthesia;
Figure 2.1:
gists
(ASA), lled the requirements for enrollment. Consecutive patients scheduled for
surgery that fullled the inclusion criteria, and accepted to participate in the study by
signing the informed consent presented in Appendix B, were included. Informed consent
was obtained by one of the anaesthesiologists who participated in the study during the
pre-anaesthetic visit.
The set of enrolled patients was randomly divided in three groups (see Figure 2.1),
dened according to the theoretical eect-site concentration of the analgesic remifentanil
49
2.2.3 Pre-Medication
Patients were pre-medicated with 10 mg of diazepam, at 22h on the previous day, and
at 7 h on the day of surgery. In the operating room, following patient monitoring, 2 mg
of midazolam were administered intravenously.
All patients were monitored according to hospital's protocols, using the General Electric
monitor and ventilator equipping the operating room.
were continuously registered through the synchronization and data acquisition software
RugloopII
R
- monitoring and automatic ventilation;
Propofol 1% Lipuro
Remifentanil Ultiva
Fresenius-Kabi Orchestra
T M monitor;
R
R
R
/Fresenius
in B.Braun
Perfusor 50 ml syringe;
R
R
in B.Braun
Perfusor 50 ml syringe;
R
Base Primea pump (2 infusion pumps);
R
Infusomat;
50
Anaesthetic drugs' administration was conducted using the TCI system Orchestra
Base Primea. Remifentanil (C=20 g/ml) administration was conducted using Minto
[38, 39] PKPD model and propofol (C=10 mg/ml) was administered using Schnider
PKPD model [50]. All stimuli and events considered important for the study were annotated and registered in RugloopII
stimulus, intubation, incision, electric cauterization, drugs' administration, suture, extubation). After data collection, the les recorded were extracted to spreadsheets and
c (Demed).
2.2.5 Methods
Anaesthetic induction was conducted similarly in all patients using TCI of propofol
and remifentanil.
sed indexes changes, in the periods preceding and following precise noxious stimuli:
laringoscopy/intubation, tetanic stimulus and incision.
In the determination of pattern responses exclusively due to the noxious stimulus,
the patient was kept with no external interference, for one minute prior and two minutes
after stimulation, with the same doses of propofol and remifentanil (phase I).
Besides pattern responses to precise noxious stimuli, data referring to dierent combinations of drugs were obtained to analyze the relations between drugs' doses, noxious
stimulation and nociceptive activation. For that purpose, during the surgical procedure,
dierent drugs' combinations were dened and kept constant for 5 minutes intervals, to
obtain both dynamic and steady-state information (phase II).
Following, the designed clinical protocol for both phases of the study is described
step-by-step.
Table 2.1: Study interruption rules: for each case the anaesthesiologist should redene
drugs' theoretical eect-site concentration targets, according to patient's needs.
51
art Rate
(HR) and
He-
SBPbaseline );
2. Start saline at 400 ml/h and remifentanil infusions, dening the eect-site concentration of remifentanil according to the study group attributed in the randomization (2, 3 or 4 ng/ml, as shown in Figure 2.1);
6. When the eect-site of propofol reaches the target, patient should be maintained
with no external interference;
7. One minute after the eect-site concentration of propofol dened after induction
is stable, start laringoscopy; after intubation keep the patient with no stimulus
for two minutes;
8. During the period of preparation for the surgical procedure, change drugs' eectsite concentration, according to patient's needs;
9. Before surgery start, set eect-site concentration of remifentanil to the dose used
during intubation, and one minute after it reaches the theoretical target, apply a
tetanic stimulus (50 Hz for 5 s, and post-tetanic count); keep the patient with no
external interference for 2 minutes;
52
PHASE II:
Study responses to noxious stimuli during surgery, for dierent ascending and descending analgesic steps.
procedure, register baseline heart rate and systolic blood pressure (HRbaseline ,
SBPbaseline );
2. In the case of remifentanil increment after surgery start, the study should be
initiated with ascending remifentanil steps, otherwise start second phase of the
study with descending remifentanil steps;
53
approval and the results presented to the scientic community [114], and published [112].
Although the study was conducted in full collaboration with the author, the results are
not in the scope of this thesis (study focused on the analysis of the post-operative acute
pain), and therefore will not be here presented.
P < 0.05).
54
55
studies.
Figure 2.2: Flow diagram of the hospital's required procedures to obtain institutional approval for the realization of clinical
P < 0.05
(data as meanstandard-
deviation).
Figure 2.3:
Orchestra
18/13
5/5
8/3
5/5
Age (years)
54.913.4
58,215,4
50,813,8
56,110,6
Weight (kg)
69.912.3
66,010,9
74,914,4
68,510,3
Height (cm)
165.07.6
161,92,6
168,49,7
164,47,5
A) BIS VISTA
T M monitor; B)
R
Base Primea syringe pumps in TCI mode (propofol and remifentainl); C)
R
monitor and ventilator; D) Computer used in data collection
c Waves installed.
using Labgrab
Of the 34 collected
56
Figure 2.4: Monitored patient during a standard data collection with a BIS bilateral
sensor, on the left, and the BIS monitor with corresponding BIS and CVI trends for the
left channel, on the right.
In the nal data set, 16 patients had an epidural catheter placed prior to anaesthesia
induction for post-operative analgesia only, and the moment of post-operative analgesia
administration registered. Type of surgery varied, with incidence of 13 prostatectomies,
8 nephrectomies, 5 nephrolithotomies, 2 cystectomies and 3 smaller procedures.
Collected data from the monitors, using RugloopII, were sampled at a rate of 1 Hz.
Nevertheless, monitors export data to the synchronization software in dierent sampling rates: Datex monitor every 5 s, infusion pumps every 2s and BIS every 1s. For
lower sampling rates the synchronization software digitizes the signal at constant steps.
Waves data les were also sampled at dierent sampling rates, depending on signal
characteristics. Due to the presented dierences, les' structure was carefully dened
for subsequent analysis.
Figure 2.3 shows the clinical setup assembled in every data collection, with the
R
R
T M monitor, Orchestra
, the BIS
syringe pumps in TCI mode, and the computer used for data collection and synchronization, with the Rugloop
2.4 shows a patient during data collection, with the BIS bilateral sensor placed on the
forehead (left), and corresponding BIS monitor and collected trends (right).
R
for o-line analysis.
variables analyzed in the study. Data collected at 1 Hz containing information from the
57
Datex
TCI Data
1. Time
38. Remifentanil Cp
1 )
(beats/min
Frequency L (Hz)
(ng/ml)
3. BP Systolic (mmHg)
18. BIS L
39. Remifentanil Ce
(ng/ml)
4. BP Diastolic (mmHg)
40. Remifentanil Ct
(ng/ml)
5. BP Mean (mmHg)
6. BP Heart Rate
43. Propofol Cp
1 )
(beats/min
7. Non-Invasive BP
(g/ml)
Systolic (mmHg)
23. SD BIS L
9. Non-Invasive BP Mean
24. SD EMG L
o
10. Temperature ( C)
25. CVI L
8. Non-Invasive BP
Diastolic (mmHg)
(ml/h)
11. SPO2 (%)
26. Impedance L ()
1 )
(beats/min
13. CO2 Et (%)
29. BIS R
1 )
Rate (cycles/min
31. EMG R (dB)
32. SQI R (%)
33. SD BIS R
34. SD EMG R
35. CVI R
36. Impedance R ()
58
infusion pumps, Datex and BIS monitors, exported to Excel les using Labgrab
later imported to Matlab
R
using ve structure variables as follows:
Table 2.4: Collected waves, number of channels, sampling rate, and data units.
Signal
EEG
ECG
1 (ecg)
PPG
1 (pleth)
IBP
1 (invp)
CO2
1 (co2)
V
V
128
100
mmHg
100
300
Carbon Dioxide
25
(IBP),
c , and later im
R
. The collected waves included four EEG channels (ch0, 1, 2 and 3),
ported to Matlab
ECG, CO2 , PPG and IBP, all with dierent sampling rates as shown in Table 2.4. Each
wave le is composed by two columns: the rst with the sampling time, synchronized
with the le sampled at 1 Hz; the second with the collected physiological signal wave
values.
For each patient
i in the study data were saved to a *.mat le pati , where i = 1, ..., 34,
containing the ve dened structures and the eight wave les.
To facilitate data inspection, a
R
using Matlab
to visualize and assess collected data. This tool allows the synchronized
navigation through the les, with a zoom feature for more detailed analysis of each
signal and event (see Figure 2.5).
59
scale.
Noxious Stimulus
Code
No stimulus
Ureteroscopy/Cystoscopy
Retractors placement and Introduction of laparoscopic surgery instruments
Retractors placement in abdominal surgery
Tetanic stimulus
Lesion extraction in laparoscopic surgery
Extubation
Skin incision for nephrectomy
Abdominal incision
Percutaneous nephrostomy incision
Skin incision for prostatectomy
Inate abdominal cavity for laparoscopy
Intubation
Laringoscopy
Tissue manipulation open surgery
Tissue manipulation cystectomy
Tissue manipulation laparoscopy
Tissue manipulation nephrectomy
Tissue manipulation prostatectomy
Skin incision for drain or Small laparoscopic incision
Periods in the surgery without surgical stimulus (any break in surgery)
Wall suture/Skin staples
Skin suture
Abdominal cavity tissues traction
Gastric drainage tube/Temperature probe nasal/oral
Vein puncture/Arterial puncture/Jugular or sub-clavian puncture
Folley catheter placement
Laparoscopy nalization
Intraoperative lithotripsy with ultrasound
Laringoscopy for oro-pharyngeal throat-packing
Visceral tissue manipulation
Nasal broscopy for intubation
Skin puncture and epidural catheter placement under general anaesthesia without
local anaesthesia
Ganglia removal in the inguinal region
Internal sutures (muscle or viscera)
60
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
Figure 2.5: Software developed to analyze and navigate o-line through the collected
data for each patient (pati , where
i = 1, ..., 34).
As
referred before, the study was divided in two phases: the rst phase focused on the
analysis of punctual events (laringoscopy, tetanic stimulus and incision), and a second
phase during the maintenance in which changes in the drugs' doses and stimulus intensity were to be related to changes in the physiological signals. To analyze changes
in stimulus two variables were included in each patient le, the rst being an array
containing the time points of precise events in the procedure, such as loss of response
to verbal and mechanical stimulus (LOC),
Recovery Of Consciousness
tubation (times variable), and a second variable (scale), which translated by a code
number the dierent stimuli annotated in the events' text le throughout the whole
case. Table 2.5 shows the code numbers attributed to each stimulus annotated by the
author during data collection, and later introduced in the
array was composed by the following events in time:
61
scale
variable. The
times
noise disturbances, and to proceed with the analysis waves must be pre-ltered, and
external interferences such as EMG, or electrical power line, removed.
One of the
problems in ECG signals processing is the existence of baseline drift, which should
be removed minimizing interference in heart beat morphology. Baseline drifts present
in ECG signals are usually below 0.5 Hz, nevertheless movement artifacts may lead to
increased frequency bounds. The two techniques usually applied in these cases are linear
lters and polynomial tting. In the collected data a linear lter was applied [131].
The high-pass lter designed must allow removal of all low-frequency components,
keeping the most of the signal clinical information. For the ECG, during a bradycardia
episode heart rate may drop to 40 beats/min, making the lowest frequency of the signal
to drop to 0.67 Hz. Nevertheless, heart rate is irregular, and its best to choose a cut-o
62
allow optimum magnitude response lters, however with non-linear phase response.
Nevertheless if the lter is applied in the two directions, a zero-phase transfer function
lter is obtained. Since in this case data were processed o-line, this was the method
applied. Let us consider
h(n)
x(n)
s(n)
the IIR lter [131], the output signal is given by the sequence of Equations 2.1,
(2.1)
(2.2)
s(n) = z2 (n)
(2.3)
Taking the discrete Fourier transform of the signal, Equation 2.4 is obtained.
X(ejw ) =
x(n)ejwn
(2.4)
n
with the property in Equation 2.5.
x(n) X (eejw )
(2.5)
(2.6)
(2.7)
(2.8)
The lter used was an IIR Butterworth of order 3, applied in both directions: cuto
frequency of 0.5 Hz (30 beats/min), and order 6.
After removing baseline wander, the next step is to detect the QRS complex and
extract the RR intervals. A linear lter given by Equation 2.9 was applied to the signal
to highlight the signal features of interest.
63
(2.9)
where
(2.10)
(2.11)
is a detected complex,
complexes and
nI =
ze,i , n = i , i + 1, ...
z(i )
ze,i
and
0.5 0.7
D1
D0,
during which
n = i + 1, ..., i + D0
nmax
g(n i D0 1) n = i + D0 + 1, ..., i + D1
n(n) =
ze,i
n = i + D1 + 1, ...
(2.12)
This method was applied to the RR intervals detection in the collected data. Figure
2.6 shows the results of the RR detection algorithm in one of the collected data les.
Upon detecting the QRS complexes, RR intervals sequence must be extracted. The
RR interval was dened at the moment of QRS detection (tRi ) as the time dierence
between two consecutive QRS complexes (tRi
tRi1 ).
64
Figure 2.6: Original ECG signal (black line), ltered ECG signal (QRS complex enhancing, gray line), and correspondent QRS complex detection (black dot).
Figure 2.7: Original RR time intervals (black line), and interpolated signal re-sampled
at 1 Hz: using hermitian cubic spline interpolation (gray line), and cubic spline interpolation (dashed line).
HR(tRi ) =
tRi
60
tRi1
(beats/min)
(2.13)
To obtain
equally spaced in time sequences, data was interpolated using a cubic Hermite spline,
maintaining the original signal trend (see Figure 2.7), and re-sampled at 1 Hz.
65
C2
the second order derivative is continuous in time, the cubic Hermite spline has no
overshoot and less oscillations, which is more adequate for the current application [133,
134].
Figure 2.8:
correspondent ltered signal after outlier removal (gray line), and RR sequence provided
by the Aisys monitor through the heart rate sequence (dashed line). Data from one of
the patients in the study.
Following QRS peaks detection, RR intervals extraction and interpolation, the obtained RR sequence was still contaminated with outliers due to poor quality in the
ECG data acquisition. To proceed with the analysis, the extracted sequence was postprocessed for outlier detection and removal. To obtain clean RR sequences the originally extracted RR sequence was smoothed by a smoothing lter using the robust Loess
method [135], which uses a local regression with linear least squares, and a 2nd degree polynomial model, assigning lower weights to outlier values in the regression. This
method allows the construction of a smoother version of the original signal, diminishing
the impact of outliers. To detect an outlier a comparison between the original and smoothed signal was performed: if the dierence between the two sequences was above two
standard-deviations of the median error observed in the population data, an outlier was
identied and substituted by the corresponding point in the smoothed signal version.
66
Figure 2.9: Invasive blood pressure signal with overlapping systolic pressure peaks detected (black dot). Data from one of the patients in the study.
For the IBP wave, information regarding systolic, diastolic and mean arterial pressure may be extracted. Due to signal physiological characteristics, systolic blood pressure is the parameter more closely related to noxious activation, and compensation
mechanisms. The methods previously described were applied to obtain blood pressure
peaks, and systolic blood pressure measured in the original signal. In this case, baseline
wander is important information contained in the physiological signal. Figure 2.9 shows
the invasive blood pressure wave, and the detected peaks later used to interpolate the
systolic blood pressure sequence.
67
For the PPG wave, the objective is to obtain the wave amplitude, which has been
related to changes in vasoconstriction and dilation, as a cardiovascular response to
noxious stimulation [78, 86, 130]. Wave amplitude is the height of the pulse, usually
obtained as the amplitude dierence between the local minimum at the beginning of
the pulse (A), and the pulse wave maximum amplitude (B) as depicted in Figure 2.10
[130].
Beat-to-beat PPG wave amplitude (PPGA) was extracted considering the consecutively registered wave values at points A (tAi ) and B (tBi ), given by Equation 2.14, and
illustrated in Figure 2.11.
(2.14)
The obtained PPGA sequence was then interpolated and re-sampled at 1 Hz.
Respiration Rate
68
Figure 2.11:
ment (A, white dot, dened as the local minimum and B, black dot, as the wave local
maximum). Data from one of the patients in the study.
Figure 2.12: CO2 wave and detected point of maximum CO2 for each respiration cycle,
corresponding to the end of an expiratory cycle. Data from one of the patients in the
study.
method adapted from the previously described, in which the moment of maximum CO2
value was detected and considered to be the end of an expiration cycle (tExpi ).
respiration rate sequence was given by Equation 2.15.
69
The
RespR(tExpi ) =
tExpi
60
tExpi1
(cycles/min)
(2.15)
Figure 2.12 shows the CO2 collected wave with the expiration detected points used
to compute RespR in cycles/min, and re-sampled at 1 Hz using cubic Hermite splines.
2.6 Summary
A clinical protocol to evaluate nociceptive responses to changes in stimulation intensity,
and drug doses was designed and submitted for approval to the Hospital de Santo
Antnio Ethics Committee and Administration Council.
Following institutional approval, and patients' written informed consent, 34 patients
scheduled for urological surgical procedures under general anaesthesia were enrolled in
the study. Of the collected data, 3 cases were excluded due to protocol deviations or
technical diculties.
Following data collection, the data were pre-processed and exported to an unique
R
le. This le contained synchronized data from all monitors used in the study.
Matlab
R
in order to ex-
tract the information contained in them, synchronized and re-sampled for the following
analysis.
70
Chapter 3
Stimulus Intensity Analysis
survey designed and disseminated in the Portuguese anaesthesiologists community (Sociedade Portuguesa de Anestesia and Servio de Anestesiologia do Hospital de Santo
Antnio), to evaluate clinical perception of the noxious intensity of a list of stimuli that
the patient may be subjected to during general anaesthesia, is reported.
Moreover,
a scale based on clinical perception constructed using Rasch analysis is proposed for
later application in the search for relations between drugs' doses, stimuli intensity and
amplitude response of the physiological variables related to noxious activation [136].
3.1 Introduction
During general anaesthesia patients are subjected to various sources of noxious stimuli originating from anaesthetic and surgical procedures. Anaesthesiologists rely on
their experience to recognize stimulation timing and intensity, anticipating stimulus
occurrence, and administering analgesics to control physiological responses to noxious
simulation.
71
72
3.2 Methods
intensity for a wide list of stimuli, that can be later used in the search of pattern
responses in the objectively measured physiological variables during general anaesthesia.
The most commonly used scales in the assessment of acute pain are visual analogue
scales, numerical rating scales, and verbal scales. Although these scales are easy to understand and use, they reect individuals' assessment and cannot be compared between
subjects, troubling statistical analysis and operations between and within individuals.
This problem may be surpassed using the Rasch model [140, 141], a mathematical tool
that uses the population tendency rather than individual's assessment to obtain scaling
measurements [31, 142, 143].
Figure 3.1: Schematic representation of the patient's system and involved variables in
the triggered noxious responses. The amplitude of response to noxious stimulation, in
this simplied version, is dependent of the drugs' doses (known) and of the stimulus
intensity (estimated).
This carries important information in the search for relations between dierent stimuli, and the triggered responses in the physiological signals associated to noxious
activation.
The idea is to have an estimate of the stimulus intensity and assess the
impact of both stimulus intensity and drugs' attenuation, as shown in Figure 3.1.
3.2 Methods
This section intends to describe the methods employed in the development of the noxious
stimulus intensity scale. The survey design, the scale construction using Rasch analysis,
and comparison of the results with previous pharmacological studies, are reported in
detail.
73
A total of 35 stimuli were analyzed in this study and shown in Table 3.1.
For each stimulus in the survey, anaesthesiologists were asked to give an evaluation
in an ordinal rating score varying between 0 (non-noxious stimulus) to 10 (severe noxious
stimulus).
Besides stimulus assessment, a few questions considered relevant to the analysis
were added to the survey: clinician's perception of the existence of a tool to assess the
analgesic state of the patient, similar to the existing tools for the hypnosis state (likely
or unlikely); what scale would be chosen for such an index (0-1, 0-5, 0-10, 0-100, other);
what is the most intense stimulus, laringoscopy/intubation, or abdominal incision for
classic apendicectomy; number of years of experience since the beginning of the training
period; and nally gender of the participant (male or female).
The survey was implemented using Google Docs (https://spreadsheets.google.
74
3.2 Methods
Table 3.1:
stimuli derived from the anaesthesia procedure; stimuli derived from the surgical procedure, pre-incision and incision; and stimuli derived from the surgical procedure postincision.
Anaesthesia Procedure
Stimulation (11, An)
Surgical Procedure
Stimulation Pre-Incision
and Incision (10, PreInc)
Surgical Procedure
Stimulation Post-Incision
(14, PostInc)
1. Intraoperative lithotripsy
with ultrasound
2. Ureteroscopy/Cystoscopy
3. Intubation (moment of
passage of the tube through
vocal cords)
2. Retractors placement in
abdominal surgery
3. Abdominal cavity tissues
traction
6. Tissue manipulation
nephrectomy
7. Tissue manipulation
prostatectomy
9. Tissue manipulation
laparoscopy
75
The Rasch
model was rst developed for dichotomous data [140, 145, 146] (painful/not painful for
example) later expanded for the analysis of polytomous data, meaning for more than
two ordered response categories [147151].
This study aimed at obtaining measures of perceived stimulus intensity, based on
rates attributed by clinicians considering their clinical perception of the stimulus intensity; this problem corresponds to a two-facet Rasch model, the two facets being the
painful stimulus and the raters who evaluated each stimulus, in a rating scale from 0
to 10. The rating scores were analyzed using a two-faceted rating scale model [141], as
presented in Equation 3.1.
Px
j=0 [n (i
Pk
j=0 [n
k=0 exp
exp
nix = Pm
where
0 0
nix
so that
Pk
j=0 [n
perception of rater
(stimulus location,
i = 1, ..., 35),
n,
(rater location,
= 1, ..., 10).
x = 0, ..., 10
(i + k )]
(3.1)
(i + k )] = 1.
categories (k
+ k )]
and
giving a rating of
n = 1, ..., 57), i
k
x,
on stimulus
i, n
is the pain
i , k )
for each stimulus (35 stimulus57 raters), providing insight not only on the perceived
stimulus intensity, but also on the raters' locations regarding pain scores attributed.
The subsequent analysis will include both facets analysis, since they are intertwined in
the measurement process (rater location and item location).
Model parameters are all in the same continuum of pain intensity/sensibility, in
a logit scale.
were estimated. Extreme scores (perfect scores) imply extreme measures, but indenite
since they congure innite values of the latent trait. Nevertheless the location of the
boundaries may be estimated, considering a 0.5 score from the extreme perfect scores,
in order to obtain a more user friendly scale (linear transformation) [152, 153].
Item response theory methodology and Rasch analysis have been used in several
elds, such as educational applications in which the latent trait is usually referred to
76
3.2 Methods
as ability, but also dierent applications in the medical eld have explored this analysis
technique, in the search of measurement and tuned questionnaires to better evaluate
pain [31, 32, 142, 143, 154158], fatigue [159], physical functioning ability and outcome
[160165], and even analysis of the written anaesthesiology examination [166], among
other variables of interest.
Rasch analysis was performed using Minifac (Facets student version 3.66.3, Linacre
2010).
Plaryngoscopy =
1+
laryngoscopy
laryngoscopy
where:
Plaringoscopy
77
(3.2)
C50,remif entanil,laryngoscopy
UP laryngoscopy
C50,propof ol,laryngoscopy
no response to laryngoscopy);
laryngoscopy
no response to laryngoscopy;
and
laryngoscopy
In the same study, the opioid-hypnotic interaction was also modelled as an hierarchical model (see Figure 3.2), where pain is blocked by the use of opioids (attenuation),
and then the attenuated pain is projected to the cortex where hypnotics act. This is
78
3.3 Results
the concept behind the use of information contained in EEG based indexes to translate nociceptive activation, since the noxious stimulation is projected interfering with
cortical EEG patterns [129].
This approach interprets the potency of the opioids in attenuating the stimulus as
a function of the stimulus intensity given by Equation 3.3.
postopioid intensity=
preopioid intensity 1
opioid
opioid +(opioid50 preopioid intensity)
(3.3)
where:
preopioid intensity
postopioid intensity
tion;
opioid
opioid50
and
preopioid intensity
The intensities estimated using the hierarchical model described are unitless and
are related to the probability of response. After attenuation the
postopioid intensity
is projected to the cortex and suppressed by the hypnotic drug, described by a similar
model structure (refer to [167] for more detailed analysis).
The drug needs were evaluated and stimulus intensities were estimated based on the
drug needs to abolish 50% of the response (pre-intensity estimation).
This approach
was further explored in a subsequent study by Huiku and colleagues [78], where the
authors estimated other noxious stimuli intensities based on drug needs reported in the
literature [167, 170, 171]. The pre-intensities estimated for the hierarchical model were
later used in the search for noxious responses in the physiological signals collected.
3.3 Results
There were 57 responders to the survey, 19 of which were men. Average years of experience in clinical anaesthesiology since the beginning of the internship of respondents
79
Frequency of each
entry in the rating scale was (rating scale entry [0-10] and correspondent frequency of
occurrence): 0-25, 1-87, 2-154, 3-189, 4-281, 5-315, 6-290, 7-281, 8-218, 9-125 and 10-30.
Figure 3.3:
or laringos-
copy/intubation (74%).
Figure 3.4: Total score for each stimulus within the three groups of stimuli studied:
anaesthetic procedure stimuli (Ani, i=1,...,11), pre-incision and incision surgical stimuli
(PreIncj, j=1,...,10), and post-incision surgical stimuli (PostInck, k=1,...,14).
Regarding the question, which is the most painful stimulus, incision for apendicectomy or laringoscopy/intubation, 74% of anaesthesiologists who answered the survey
considered laringoscopy/intubation as the most painful stimulus (see Figure 3.3).
Total scores obtained for each stimulus analyzed (anaesthetic, pre-incision and incision, post-incision; 0 to maximum of 570) are shown in Figure 3.4; total scores rated by
each respondent (0 to 350) are shown in Figure 3.5. There is a wide variation in total
80
3.3 Results
Figure 3.5: Total scores for each respondent, considering the 35 evaluated stimuli.
scores obtained for each respondent, consequence of dierent rater severities and scale
use/interpretation.
Figure 3.6:
Assuming the existence of such a tool, anaesthesiologists were also asked to give
81
score when compared to the remaining responders, demonstrating lower severity. This
may have been due to a misunderstanding of the rating criteria, leading to unexpected
responses within the measurement process, or to loss of collaboration in the survey
response, since the responses by this rater were of 0 in many of the stimuli analyzed.
Raters who had recurrent incidence of outlying entries, meaning unexpected responses within the adjusted Rasch model to the population trend (response outside modelled
boundaries resulting in high residuals), were removed from the data set, and the analysis
conducted with the remaining evaluators. Each removed rater was individually analyzed to identify the possible exclusion cause. The removed raters either had unrealistic
ratings (loss of collaboration in the survey response), or dierent opinions regarding
some of the stimuli within the population trend, leading to occurrence of high residuals.
Stimuli measures (i ,
i = 1, ..., 35,
each group of stimuli considered. The model obtained presents a reliability index of 0.98
(98% of the observed variance resulted from true measure variance). After obtaining
the Rasch measurements for each stimulus in logit, measures were linearly rescaled to
present values in a linear scale from 0 to 10, for ease of understanding and application
(Measure [0-10] in Table 3.2).
using values of 0.5 from the perfect score; values of -3 for 0, and 3.3 for 10 were used.
Stimulus intensity measures (logits) provided by the adjusted Rasch model are also
82
3.3 Results
Table 3.2: Total and average scores obtained for each stimulus, divided according to the
considered groups, stimuli measures (i ,
i = 1, ..., 35,
logit values indicating higher pain perception), standard errors (S.E.), expected raw
score, Rasch measure parameterized in a scale from 0 to 10, and stimulus identication,
in descending order of stimulus intensity perception (reliability of 0.98).
Anaesthesia Stimuli
7,67
7,00
6,38
5,81
5,17
4,33
4,21
3,92
3,90
3,81
3,73
1,28
0,88
0,56
0,28
0,00
-0,37
-0,42
-0,55
-0,56
-0,60
-0,64
S.E.
(logit)
0,11
0,10
0,10
0,09
0,09
0,09
0,09
0,09
0,09
0,09
0,09
Expected Measure
Stimulus
Score
[0-10]
7,74
7,08
6,46
5,85
5,20
4,32
4,20
3,89
3,87
3,78
3,68
6,81
6,16
5,65
5,21
4,76
4,17
4,09
3,88
3,87
3,80
3,74
An2
An3
An6
An4
An10
An11
An8
An1
An7
An5
An9
Brief Descriptor
Laringoscopy
Intubation
Throat-packing
Fibroscopy
Tetanic
Extubation
Jugular or sub-clavian puncture
G18 catheter
G20 artery
Gastric tube/Temperature probe
Epidural catheter
8,02
7,94
7,29
7,08
6,79
5,88
5,50
5,25
4,92
3,23
1,53
1,47
1,04
0,92
0,76
0,32
0,15
0,03
-0,11
-0,87
S.E.
(logit)
0,12
0,12
0,11
0,10
0,10
0,09
0,09
0,09
0,09
0,10
Expected Measure
Stimulus
Score
[0-10]
8,10
8,01
7,36
7,16
6,86
5,94
5,55
5,27
4,94
3,17
7,20
7,11
6,43
6,23
5,98
5,27
4,99
4,81
4,58
3,37
PreInc6
PreInc8
PreInc7
PreInc9
PreInc10
PreInc5
PreInc4
PreInc3
PreInc2
PreInc1
Brief Descriptor
Incision supra-umbilical
Incision nephrectomy
Incision prostatectomy
Incision lumbar
Incision craniotomy
Inate abdominal cavity
Retractors in laparoscopy
Incision drain/lapar./nephrostomy
Ureteroscopy/Cystoscopy
Folley catheter
Post-Incision Stimuli
6,98
6,25
6,00
5,73
5,62
5,12
5,00
5,00
4,88
4,79
4,75
4,67
4,54
2,04
0,87
0,49
0,37
0,25
0,20
-0,03
-0,08
-0,08
-0,13
-0,17
-0,19
-0,22
-0,28
-1,51
S.E.
(logit)
0,10
0,10
0,10
0,09
0,09
0,09
0,09
0,09
0,09
0,09
0,09
0,09
0,09
0,11
Expected Measure
Stimulus
Score
[0-10]
7,07
6,31
6,05
5,78
5,67
5,13
5,01
5,01
4,89
4,79
4,75
4,67
4,53
1,95
6,15
5,55
5,36
5,16
5,07
4,72
4,64
4,64
4,56
4,49
4,46
4,41
4,32
2,35
83
PostInc3
PostInc2
PostInc6
PostInc7
PostInc5
PostInc11
PostInc13
PostInc8
PostInc14
PostInc10
PostInc1
PostInc9
PostInc4
PostInc12
Brief Descriptor
Abdominal tissue traction
Retractors abdominal
Nephrectomy manipulation
Prostatectomy manipulation
Cystectomy manipulation
Lumbar manipulation
Wall suture
Ganglia removal
Skin staples
Lesion extraction laparoscopy
Lithotripsy/Ultrasound
Laparoscopy manipulation
Internal sutures
No surgical stimulus
1, ..., 35,
i=
in the Rasch model), and correspondent expected score in the original rating
scale.
84
3.3 Results
shown in Figure 3.7, demonstrating the dispersion of the stimuli along the stimulus
intensity continuum.
Figure 3.8: Representation of the probability of response of each rating score (0 to 10),
depending on the location in the pain intensity continuum (logit).
The probability of response in a determined rating score, regarding the pain intensity
location, is shown in Figure 3.8, and the relation between scores, and pain perception
measures given by the model, shown in Figure 3.9. Approximately between -1 and 1
logit the relation is close to linear; the relation is given by an ogival shape function,
which transforms non-linear scores into linear pain perception measures.
Distribution of raters' location (n ,
n = 1, ..., 57,
Figure 3.10, measures are not presented for brevity. Raters' model presents a reliability
of 0.94 (6% of the observed variance results from measurement error).
The most intense stimuli considered were: abdominal incision 15cm supra-umbilical,
nephrectomy incision, laringoscopy, prostatectomy incision, lumbar incision, intubation,
abdominal cavity tissues traction, and incision for craniotomy. Stimuli with lower perceived intensity were:
insertion or temperature probe, epidural puncture, Folley catheter, and periods in surgery with no surgical stimulus.
As aforementioned, for the comparison of the relative pain intensity considering laringoscopy/intubation and abdominal incision, 74% of anaesthesiologists agreed that
85
Figure 3.9: Relation between score rates (0 to 10) and pain intensity measures (logit),
with overlapping 0.5 points of expected score (squares).
n = 1, ..., 57,
86
3.3 Results
presented a measure of 1.28 logits, followed by intubation with 0.88 logits, with laringoscopy presenting a higher measure than incision for prostatectomy, only surpassed
by incision for nephrectomy (1.47 logits) and transversal abdominal incision of 15cm
supra-umbilical (1.53 logits), of all stimuli considered.
Figure 3.11: Application of the developed stimulus intensity scale: schematic representation of an urological procedure under general anaesthesia with identied stimuli
intensities of the anaesthetic and surgical actions.
For demonstration purposes, the developed scale was applied to describe changes in
noxious stimuli for one of the urological procedures collected under general anaesthesia,
and shown in Figure 3.11 with the illustration of dierent moments of the anaesthetic
and surgical procedures, characterized by dierent intensity levels. The stimulus map
corresponds to a prostatectomy procedure, with the dierent stimuli moments both from
anaesthetic and surgical procedures; observation of the estimated intensity variations
allows to infer the changes in the analgesic drug to compensate and anticipate the
stimulus occurrence.
Observing Figure 3.11 a limitation in the study is encountered, since moments
without stimulus post-intubation were not dierentiated from moments with no stimulation prior-intubation. The stimulation of an endotracheal tube should not be disregarded, and should be dierentiated when evaluating the stimulus intensity. Although
87
Stimulus
Laringoscopy
0,83
6,81
An2
Laringoscopy + Intubation
1,57
6,16
An3
0,8
4,81
PreInc3
1,25
4,99
PreInc4
4,41
PostInc9
1,5
6,15
PostInc3
0,8
4,76
An10
Incision (laparoscopy)
Trocar
Laparoscopy surgery
NoStim
88
3.4 Discussion
were compared to the Rasch scale obtained in this study, using the clinical perception.
The referred studies obtained an estimation of the stimulus intensity based on analgesic
drug needs to abolish the response to stimulation. Results of the comparison are shown
in Table 3.3 and Figure 3.12, with the stimuli considered in the pharmacological studies,
and correspondent stimulus in this study, the pre-intensities estimated, and corresponding value in the Rasch scale. Notice that laringoscopy+intubation was compared to
intubation in this study, leading to lower values in the Rasch scale, most probably
due to the way the question was posed to the respondents who were asked to evaluate
intubation (moment of passage of the tube through vocal cords) independently from
laringoscopy. Nevertheless, the relation between the two assessments is close to linear,
supporting anaesthesiologists' stimulus assessment and their evaluations, which lead to
the Rasch scale proposed in this thesis.
3.4 Discussion
Rasch models are probabilistic models which lead to the construction of linear continuous measures, based on ordinal observations of the population. A measure of noxious
stimuli intensity was successfully constructed, based on anaesthesiologists' perception,
which may be used in the development of a tool to assess Noc/ANoc balance during
general anaesthesia. To our knowledge this is the rst time such information was obtained. It is a simple study to conduct and its importance may be disregarded, however
we think that the information gathered through such a simple method is relevant and
may be useful.
The number of respondents to the survey (57) may seem small, however this number
corresponds to approximately 5% of the total number of Portuguese anaesthesiologists.
One may argue that the number of respondents is low, nonetheless the model of the
population trend presented a reliability of 98%, and results should not vary signicantly
with the introduction of more raters in the process. Also, although it was applied to
the Portuguese anaesthesia community only, there is no reason for bias.
Portuguese anaesthesiologists who answered the survey in their majority believe
that a tool to assess the Noc/ANoc balance, similar to the tools used in the assessment
of hypnosis and neuromuscular blockade, will be available in the future. Nevertheless
there is still a high incidence of skepticism for this matter, and 30% of the enquired
professionals are not sure that this tool may be in the future developed.
89
There are
90
3.4 Discussion
observations were removed. The occurrence of outlying observations was not surprising,
since the survey was long and loss of collaboration was a real concern, also some stimuli
may be dierently interpreted by raters, leading to unexpected responses considering
the population modelled trend.
Regarding the stimuli perceived as less intense, the one assessed to be the less
intense of all considered was periods in surgery with no surgical stimulus, even when
compared to a venous catheter puncture. This may be due to the way the problem
was posed to the respondents, who were asked to evaluate stimulus noxious intensity
considering a venous puncture prior to drug administration, and for the remaining
stimuli noxious intensity under general anaesthesia.
surgical stimulus during a surgical procedure still imply the presence of noxious stimuli
namely from the presence of a tracheal tube, positioning, or of retraction, nevertheless it
also implies the use of anaesthetics to maintain the patient in a state of unconsciousness,
as for the venous catheter puncture stimulation was considered to occur in an awake
patient, which is in many cases suering from anxiety in the period prior to anaesthesia
induction.
This was the probable reason why a venous catheter puncture in the
awake patient was considered to be a more intense stimulus than periods in surgery
with no surgical stimulation. When trying to evaluate personal opinions it is of great
importance to adequately state questions to obtain usable information.
One concern which may rise with this study is the use of subjective individual assessment of noxious stimuli intensities, which may be prone to bias and misinterpretations.
Although with the Rasch analysis the dierence in raters' severity should be compensated to produce independent measures, to understand the implications of this matter,
relation between pre-intensities obtained in pharmacological studies, which are based
in objective data of drug needs, and the proposed Rasch scale were investigated.
linear relation between the two approaches was observed; the result strengthens the
method proposed in this study, showing that clinical perception may be used as a tool
to estimate noxious stimulus intensity, as it would be expected since it is based on clinical experience and perception, that anaesthesiologists guide anaesthesia (drug doses)
in their daily activities. Since anaesthesiologists' perception of stimulation intensity is
based on the stimulus anticipation and drugs' adjustment, it was expectable that this
relation was strong, nevertheless the conrmation of this relation, leaves room for the
use of this simple method in a wider number of noxious stimuli.
91
An anti-
nociceptive drug which may be adjusted very quickly is already available - remifentanil what is lacking is an objective measure of nociception. In previous studies, a search for a
nociception index has lead researchers to compare dierent precise stimulus, for dierent
drug doses [77, 78, 86, 87, 138], and attempts have been made to obtain scaling methods
connecting stimulus and analgesic level, nevertheless taking into account few experts
opinions [77]. In this study a survey was designed to evaluate the clinical perception
of several anaesthesiologists on the stimuli intensity, leading to the construction of a
measure of stimuli intensity that can be later used to further investigate changes in
physiological parameters in response to stimulus variations.
Besides the usability of this technique in the search for a tool to evaluate and quantify
the Noc/ANoc balance, results in this study allow clinicians an overview of stimuli
intensities they encounter in their daily activity, and the inference of the relations
between them. This scale provides a term of comparison between dierent stimuli, and
a chronological prediction of intensity in the anaesthetic and surgical procedures.
The present survey must be simply regarded as a tool to obtain data and develop
a scale to be applied to clinical studies oriented at relating noxious stimulus, antinociception drugs, and correspondent measurable physiological responses in the patient.
The novelty brought by this study was the use of Rasch analysis to construct the noxious
intensity measure, with the input of a group of anaesthesiologists, for a more objective
assessment of these relations.
3.5 Summary
A measure of stimulus intensity was constructed, allowing scaling and comparability
between the considered noxious stimuli. This information was used to better understand
92
3.5 Summary
the relationship between stimulus intensity, drug doses and physiological responses in
the collected data.
The original
scale
during the stay in the operating room, was updated to contain the corresponding stimulus intensity values of the developed noxious stimulus intensity scale. A new variable
containing this information was created (sca) and employed in the following analysis,
to translate the stimulus intensity the patient was under throughout the entire case.
93
94
Chapter 4
Data Analysis - Passive Nociception
Measures
This chapter is focused on the analysis of passive Noc/ANoc balance measures. After the
pre-processing of the collected data, as described in Chapters 2 and 3, it was analyzed
in the search for the best descriptors of noxious activation, and of the relations between
input and output signals.
analyzed signals, and variations in response to noxious stimulation, are merely observed
without outer interference.
4.1 Introduction
The objective during general anaesthesia is to maintain the patient in homeostasis, and
provide conditions for the clinical procedures to be performed. Considering the system
represented in Figure 1.20, if the input drugs are constant, the output should only vary
with a stimulus variation, and therefore measures of variability of the system outputs
may translate the Noc/ANoc balance of the patient.
95
This
analysis allows the denition of the more informative variables responding to noxious
stimulation in standardized conditions.
96
response to precise stimuli, and analgesic dose impact on the amplitude response.
4.2.1 Methods
Several methods are proposed in the literature to assess amplitude response to precise stimulation, from comparison of baseline values prior stimulation with maximum
response after stimulation, average values before and after, among others. The methods
should be applied carefully since comparisons between dierent extracted measures may
lead to erroneous assumptions, for example comparing average values in a time window
to the maximum value reached post-stimulation, is doomed to bias, and tends to show
dierences (see Figure 4.1).
average value in a time window prior stimulation, versus the average post-stimulation
were used, followed by amplitude response assessment given by Equation 4.1.
97
tP re =60 s
and
(4.1)
tP ost =180 s.
Besides obtaining average values for each time span, dierences observed in individual baseline values of variables analyzed in this study should be taken into consideration. To increase comparability between individual responses all
Amplitude Response's
extracted were normalized by the baseline values observed in each patient prior drugs'
administration, in this way, more precise results may be guaranteed when comparing
groups of patients, for example to compare the impact of the drug dose on amplitude
response for the dierent groups of study.
Amplitude ResponseN
Equation 4.2.
Amplitude ResponseN =
where
tP reDrugs =180 s
Amplitude Response
Average(tP reDrugs )
(4.2)
The variables analyzed in this study were HR, SBP, PPGA, BIS, EMG, BIS SD,
EMG SD, CVI, SEF, TOTPOW, and ASYM. Besides the cardiovascular and EEG
derived indexes collected during general anaesthesia, an additional index proposed in
the literature for the Noc/ANoc balance assessment (ANSSI) was implemented, since it
is an open source index. ANSSI was obtained as described in Chapter 1, in Equations
1.11 and 1.12.
During data collection, a BIS bilateral sensor was used, and laterality eects reported
in the literature were taken into account [17]. Since BIS data were available for the two
EEG channels, contralateral data according to the site of surgical incision was used:
if the incision was on the left side, EEG right channel was used and vice-versa; if the
incision was central, the average of the indexes from both channels was used instead.
98
Figure 4.2: Schematic representation of the laringoscopy and intubation moments, registered in each data set, and the corresponding time lag - stimulus duration.
Although the stimuli chosen to assess the analgesic impact between groups of patients are considered to be similar between individuals, the fact is that stimulation
duration from laringoscopy to intubation may dier according to the diculty of intubation. Although the stimulus is the same, in some cases the duration of the stimulus
diers (consequence of a dicult intubation), leading to more pronounced responses
(e.g.
anatomical dierence).
from rst laringoscopy to intubation were analyzed, to verify if there was any bias
factor (see Figure 4.2).
In addition to the dierence in stimulus duration for laringoscopy/intubation, the
incision stimulus may also introduce bias. Incision stimuli intensities dier according to
the surgical procedure, as described and concluded in the previous chapter (Rasch scale).
Since the study groups are composed of dierent surgical procedures, and consequently
dierent incision types, dierences between groups considering incision types, meaning
intensities, were investigated.
99
P < 0.05
was considered
R
software was used for statistical analysis.
4.2.2 Results
Table 4.1: Patients' demographic data (original and phase I nal sample): global and
for each considered study group of remifentanil target eect-site concentration (data as
meanstandard-deviation).
5/5
8/3
5/5
Age (years)
54.913.4
58,215,4
50,813,8
56,110,6
Weight (kg)
69.912.3
66,010,9
74,914,4
68,510,3
Height (cm)
165.07.6
161,92,6
168,49,7
164,47,5
4/5
7/2
4/4
Age (years)
55.012.4
57,015,8
54,011,1
53,910,6
Weight (kg)
69.412.2
63,49,1
77,212,9
66,810,9
Height (cm)
164.57.5
161,62,5
168,79,6
163,17,6
Data were collected from 34 patients under general anesthesia (3 excluded) as described in Chapter 2. From the 31 remaining data sets, ve were eliminated from this phase
of the study due to protocol deviations: one case from Group 1, two cases from Group
2, and two cases from Group 3. The nal data set was composed by 26 patients, nine in
Group 1, nine in Group 2 and eight in Group 3. Table 4.1 shows the demographic data
of the nal sample. No statistical dierences between groups considering demographic
100
Incision Site
Left Central Right
Group 1
Group 2
Group 3
data were found (Lilliefors, Kolmogorov-Smirnov and Fisher's tests). Data presented
as meanstandard-deviation.
Table 4.2 shows the frequency of occurrence in each study group of each incision
site: left, central of right. No statistical dierence was found between groups considering
incision larerality distribution.
Time from rst laringoscopy to intubation was registered and analyzed in the search
for possible bias factors. In general time from rst laringoscopy to intubation was below
60 seconds, only two cases exceeded the one minute interval, and in only one case
intubation was guided using GlideScope
R1
(median of intubation times, excluding the
guided intubation was of 16 s, ranging from 4 s to 144 s), with no statistical dierence
between groups considering time for laringoscopy (Kruskal-Wallis test).
Incision intensity was compared between study groups to verify if there was any
dierence that could bias the observations. No statistical dierence was found between
groups considering surgical incision intensity (Kruskal-Wallis test). Average observed
intensity in the study groups were 5.951.11, 5.810.93 and 6.300.92 in the Rasch
scale.
101
Baseline
P < 0.05).
Laringoscopy
Pre
Post
Tetanic
Pre Post
Incision
Pre
Post
HR
RemiCe=2.0
67,78*
63,98*
71,66*
60,58
59,68
58,69
59,00
RemiCe=3.0
68,98*
58,85*
65,14*
64,86
61,00
59,60
58,59
RemiCe=4.0
82,61*
59,38*
65,71*
63,81
63,48
62,72
62,35
RemiCe=2.0
126,34*
91,66*
108,16*
103,37
100,97
104,90*
108,07*
RemiCe=3.0
125,05*
97,43*
102,04*
100,95
99,23
98,51*
115,37*
RemiCe=4.0
120,54*
101,26*
96,76*
101,28
95,02
96,83*
104,75*
RemiCe=2.0
1,12*
6,14*
4,10*
1,11
1,07
1,62*
1,02*
RemiCe=3.0
1,53*
5,75*
5,18*
6,38
7,32
6,52*
2,33*
RemiCe=4.0
1,17*
2,46*
2,68
1,82
2,49
2,56
2,29
RemiCe=2.0
82,78*
38,34*
54,39*
85,70
85,52
78,49*
81,92*
RemiCe=3.0
90,15*
48,85*
55,49*
49,93
46,93
55,59*
67,09*
RemiCe=4.0
82,73*
59,48*
57,36
60,64
52,66
60,95*
67,12*
SBP
PPGA
ANSSI
102
Table 4.4: Median values of BIS related variables analyzed in the study, for each stimulus (average pre and post stimulation), divided according to the drug dose group:
remifentanil eect-site concentration (RemiCe) of 2.0, 3.0, or 4.0 ng/ml (* stands for
signicant dierences with
P < 0.05).
Baseline Laringoscopy
Pre Post
Tetanic
Pre Post
Incision
Pre Post
BIS
RemiCe=2.0
81,43*
43,63*
46,60*
46,79
42,50
41,79
43,44
RemiCe=3.0
95,81*
50,51*
47,00
42,42
43,17
42,63
43,78
RemiCe=4.0
87,91*
47,14*
45,08
43,22
45,95
45,39
45,21
RemiCe=2.0
3,60
3,21*
7,13*
3,26
3,25
2,44*
3,11*
RemiCe=3.0
1,93
2,52*
4,88*
2,79
3,28
2,94*
3,33*
RemiCe=4.0
3,17
2,83*
5,33*
3,84*
4,51*
3,19*
5,25*
RemiCe=2.0
44,67*
26,52*
28,19*
26,51
26,90
25,68
26,46
RemiCe=3.0
48,62*
27,23*
27,67*
26,70
26,93
25,90
25,94
RemiCe=4.0
49,85*
28,93*
29,29*
26,51
26,65
26,67
26,63
RemiCe=2.0
2,98*
0,88*
3,41*
1,21
0,43
0,44*
1,04*
RemiCe=3.0
3,76*
1,63*
2,85*
0,40*
1,56*
0,33
0,34
RemiCe=4.0
3,18*
1,12*
3,10*
0,41
1,22
1,25
1,02
RemiCe=2.0
22,50*
16,19*
16,31
16,17
15,93
15,67
15,68
RemiCe=3.0
20,11*
16,47*
16,42
15,42
15,02
15,38
15,40
RemiCe=4.0
24,55*
16,10*
15,92
16,31
16,26
16,36
16,60
RemiCe=2.0
56,76
62,67
62,42
60,99
61,74
61,01
62,06
RemiCe=3.0
61,59
61,55
61,71
61,17
60,88
61,11
60,79
RemiCe=4.0
59,31
64,41
63,20
61,19
61,16
61,00
60,52
RemiCe=2.0
51,02
52,47
51,36
52,65
53,05
54,78
53,81
RemiCe=3.0
51,68
48,79
49,75
50,29
49,48
50,02
48,59
RemiCe=4.0
54,13
48,89
48,60
48,97
49,39
49,76
49,31
BIS SD
EMG
EMG SD
SEF
TOTPOW
ASYM
103
Also
EMG activity was observed in all study groups for the laringoscopy/intubation stimulus.
EMG SD was assessed demonstrating that there was a signicant depression of EMG
SD in response to induction. Also a signicant increase in EMG SD was observed in all
study groups considering the laringoscopy/intubation stimuli. An EMG SD signicant
increase was also observed for the tetanic stimulus, in study group 2, and in response to
incision in study group 1, although not statistically signicant in the remaining study
groups.
SEF was globally depressed at induction, with a tendency of depression in response
to increasing remifentanil doses for the tetanic stimulus.
in response to induction.
104
Original Data
Baseline/Pre-Laringo Laringoscopy Tetanic Incision
HR
RemiCe=2.0
-6,0
7,91
-0,95
0,31
RemiCe=3.0
-2,8
2,01
-1,45
1,08
RemiCe=4.0
-14,3
2,58
-1,42
-0,93
RemiCe=2.0
-30,6
11,63
0,09
6,49
RemiCe=3.0
-20,6
2,91
-0,20
4,22
RemiCe=4.0
-17,7
0,71
-3,01
6,43
RemiCe=2.0
3,9
-1,07
0,03
0,02
RemiCe=3.0
3,4
-0,39
-0,04
-0,53
RemiCe=4.0
1,1
0,36
0,06
-0,17
RemiCe=2.0
-45,1
19,77
-0,56
0,60
RemiCe=3.0
-30,9
6,64
0,64
5,15
RemiCe=4.0
-18,3
-5,64
-2,25
2,87
SBP
PPGA
ANSSI
Following the analysis of individual stimulus amplitude response, amplitude responses for the three stimuli considered in the study were analyzed within each study group.
Tables 4.5 and 4.6 present the median amplitude responses (baseline to pre-laringoscopy,
pre-laringoscopy to post-intubation, pre-tetanic to post-tetanic, and pre-incision to postincision) for each variable within each study group.
Statistical signicant dierences in HR amplitude response were observed considering the stimuli, being laringoscopy/intubation the stimulus with most pronounced
response, followed by incision. SBP amplitude responses were also dierent when comparing tetanic and incision stimuli.
105
Original Data
Baseline/Pre-Laringo Laringoscopy Tetanic Incision
BIS
RemiCe=2.0
-39,8
3,78
-0,09
0,55
RemiCe=3.0
-44,5
-4,75
-0,96
1,03
RemiCe=4.0
-42,7
-0,88
1,44
-1,40
RemiCe=2.0
-0,8
3,93
0,53
0,93
RemiCe=3.0
1,4
1,62
0,18
0,61
RemiCe=4.0
-0,5
1,59
1,12
1,67
RemiCe=2.0
-18,2
2,02
-0,19
0,33
RemiCe=3.0
-22,6
0,20
0,08
0,02
RemiCe=4.0
-18,3
1,04
0,03
0,22
RemiCe=2.0
-2,4
3,03
-0,30
0,42
RemiCe=3.0
-1,3
1,54
0,42
0,02
RemiCe=4.0
-1,7
2,27
0,62
-0,01
RemiCe=2.0
-7,3
0,18
-0,23
-0,05
RemiCe=3.0
-3,6
-0,35
-0,17
-0,01
RemiCe=4.0
-8,2
-0,10
-0,03
0,05
RemiCe=2.0
5,9
-0,25
-0,04
0,05
RemiCe=3.0
-0,7
0,16
-0,22
-0,32
RemiCe=4.0
4,9
-0,56
-0,05
-0,24
RemiCe=2.0
-1,0
-1,00
-0,51
-0,41
RemiCe=3.0
-2,0
0,76
-0,59
-0,67
RemiCe=4.0
-5,5
-0,02
-0,49
-0,87
BIS SD
EMG
EMG SD
SEF
TOTPOW
ASYM
Amplitude response for each stimulus was compared between the three study groups
considered, meaning according to the analgesic remifentanil dose. To perform this com-
106
-11,8
10,8
-1,5
0,5
RemiCe=3.0
-3,6
3,0
-1,6
1,5
RemiCe=4.0
-15,9
3,6
-1,7
-1,4
RemiCe=2.0
-26,9
10,2
0,1
5,9
RemiCe=3.0
-18,7
2,7
-0,2
5,3
RemiCe=4.0
-14,8
0,6
-2,3
5,1
RemiCe=2.0
235,2
-66,3
1,7
2,2
RemiCe=3.0
224,4
-20,9
-4,3
-67,1
RemiCe=4.0
77,9
29,3
5,8
-14,0
RemiCe=2.0
-54,1
23,9
-0,8
0,7
RemiCe=3.0
-34,3
6,7
0,7
6,8
RemiCe=4.0
-21,7
-6,5
-3,1
3,3
SBP
PPGA
ANSSI
parison original and normalized amplitudes were used. Each stimulus was considered
separately and compared between groups using a non-parametric sign rank test for independent samples. Tables 4.5 and 4.6 present the median original values for each variable
analyzed, and Tables 4.7 and 4.8 present the median normalized (%) amplitude responses, by the baseline value prior drug administration (Equation 4.2), for each variable
analyzed.
Statistical dierences were found between amplitude responses observed in each
study group at induction (HR, PPGA and ANSSI) and in response to the stimulus
laringoscopy/intubation for the HR, SBP, PPGA, ANSSI, BIS, EMG and EMG SD,
with a tendency of decreasing amplitude responses for increasing remifentanil doses.
107
-48,8
4,7
-0,1
0,6
RemiCe=3.0
-46,1
-5,1
-1,1
1,1
RemiCe=4.0
-46,5
-0,9
1,5
-1,6
RemiCe=2.0
-22,1
80,6
2,7
12,9
RemiCe=3.0
68,1
82,1
6,3
22,8
RemiCe=4.0
-13,8
53,4
32,9
74,5
RemiCe=2.0
-40,7
4,8
-0,3
0,5
RemiCe=3.0
-45,3
0,4
0,2
0,0
RemiCe=4.0
-40,6
1,9
0,0
0,4
RemiCe=2.0
-68,8
89,8
-9,8
14,2
RemiCe=3.0
-57,2
52,3
10,7
0,5
RemiCe=4.0
-59,6
63,3
17,7
-0,2
RemiCe=2.0
-31,5
0,8
-1,3
-0,2
RemiCe=3.0
-18,1
-1,5
-0,9
-0,1
RemiCe=4.0
-34,8
-0,4
-0,1
0,2
RemiCe=2.0
10,6
-0,4
-0,1
0,1
RemiCe=3.0
-1,1
0,2
-0,4
-0,5
RemiCe=4.0
8,5
-1,0
-0,1
-0,4
RemiCe=2.0
-1,8
-2,1
-1,1
-0,9
RemiCe=3.0
-3,9
1,6
-1,1
-1,3
RemiCe=4.0
-9,6
0,0
-0,9
-1,6
BIS SD
EMG
EMG SD
SEF
TOTPOW
ASYM
108
remifentanil eect-site
2,47
4,80
1,75
1,54
1,23
1,81
RemiCe=3.0
5,94
3,41
1,09
1,16
1,01
1,24
RemiCe=4.0
6,57
6,61
1,52
1,53
1,39
1,79
109
pose as a problem, since the BIS index is already alerting to the arousal episode.
4.2.3 Discussion
Analysis of amplitude response to precise noxious stimulation of several physiological
variables was conducted. The stimuli employed in this analysis were stimuli considered
to have similar noxious intensity between patients, and therefore comparable between
individuals.
Bias factors that could introduce variations to amplitude responses, and inuence
the observed results were analyzed both for laringoscopy time and incision intensity.
No statistical dierence was found between groups, nevertheless an increase of incision
intensity was observed in study group 3. Although it was not signicant, regarding the
reduced number of patients in each study group this may account for the observation
of increased responses to incision of some of the analyzed physiological variables, even
with a higher remifentanil dose, when compared to the remaining study groups.
Of the physiological signals analyzed in this study, some responded both to stimulus,
and remifentanil attenuation. These variables were HR, SBP, PPGA, BIS SD, EMG
and EMG SD.
Two proposed indexes in the literature were also assessed to verify if they responded
adequately to stimulus and drug attenuation. Being these indexes post-processed measures which include the previously referred physiological signals responding to stimulation and drug attenuation, it was expected that they too presented adequate responses
to both factors.
This was observed for ANSSI, nevertheless the same tendency was
not possible to observe for the CVI due to limitations in the index presentation for the
110
comparability between patients [86]. Although this was not taken into account, indication of the information contained in this variable regarding the Noc/ANoc balance was
found, and in future studies these issues should be addressed and corrected.
In general the variables that have been linked in previous studies to noxious activation [16, 17, 20, 77, 86] presented signicant results in response to a noxious stimulus,
and for the amplitude response attenuation of the analgesic drug dose. Therefore, these
variables were concluded to contain information on noxious activation, and to provide
indicators of the patient's Noc/ANoc balance. This is in accordance to previous results
presented in the literature, and described in Chapter 1.
Statistical signicant dierences in amplitude responses considering the three noxious stimuli applied in this phase of the study were observed. Laringoscopy/intubation
was the stimulus with more pronounced response (increased amplitude response to stimulation), followed by incision, and tetanic stimulus.
results presented in the literature [168170], and by the observed tendency described
in Chapter 3, where laringoscopy and intubation were considered to be high intensity
stimuli (6.81 and 6.16 in the Rasch scale) even when compared to some incision types
(see Figure 3.7), and tetanic stimulation (4.76).
The main objective of this phase of the study was to identify variables linked to
noxious activation, that adequately translate the impact of the analgesic drug and
stimulus intensity, this was achieved in accordance to the tendencies described in the
literature.
111
Steady-State
since it allows the identication of models, to translate the patient's response to changes
in drugs' doses and stimuli intensities.
Throughout daily activities body natural balance suers constant interferences, automatically and unconsciously adjusting physiological variables to maintain homeostasis.
During general anaesthesia, the patient is subjected to the action of drugs and
noxious stimulation, which interfere with the natural balance and also trigger natural
compensatory mechanisms (see Figure 1.20). These relations were investigated to explain the involved mechanisms in response to drugs and stimuli inputs, and obtain a
measure of the patient Noc/ANoc balance.
In a rst approach a SS measure was extracted from the input signals (drugs' doses
and stimulus intensity), and output measurable physiological signals related to noxious
activation (BIS, EMG, HR, SBP, PPGA, as presented in the previous section), in order
to evaluate patient's homeostasis [174177].
were used in the search of dynamic models that may explain the individual responses
to noxious stimulation and drugs' doses, observed in the output signals.
Most of the indexes proposed in the literature to assess the patients' Noc/ANoc
balance are based on measures of physiological signals variability associated to noxious
activation and their normalized values: BIS, EMG, HR, PPGA, SBP [17, 20, 77, 78].
Some of the methods proposed so far present limitations in their application and solely provide information on the variation of the physiological signal, and not on the
direction of the occurring change. To respond to this problem a dierent signal processing technique was applied, adequate to non-stationary signals, capable of assessing
and translating the change and shift direction of the physiological signals associated to
112
hypnotic dose may be adjusted using information derived from EEG based indexes and
clinical observation (see Table 1.1) to maintain an adequate unconscious state, avoiding
awareness or overdosing, and possible adverse consequences [12, 13, 4042]. Considering
anaesthesia as a combination of the three components, if a certain unconscious state
is maintained, variations in the patient's state may be considered to derive only from
changes in the stimulation intensity. Bouillon et al. [167] present anaesthesia opioidhypnotic interaction as an hierarchical model (see Figure 3.2), where pain is blocked by
the use of opioids (attenuation), and then the attenuated pain is projected to the cortex
where hypnotics act. If one considers this hierarchical model it is possible to infer that
changes in the patient's physiological signals are a consequence of noxious stimulation
variation, and therefore closely linked to the Noc/ANoc balance. This notion is applied
by anaesthesiologists in their daily activities, detecting changes in the physiological
signals' trends, and adjusting drugs' doses to compensate the uctuations, acting as
controllers to maintain patients in optimal SS conditions. A measure of combined SS
for dierent physiological signals was obtained.
inputs and outputs of the system. In the specic case the patient is the system, drugs'
doses and stimulus intensity the inputs, and the measurable physiological signals related
to noxious activation the outputs.
TCI of propofol and remifentanil was used for drugs' administration, and therefore
SS of drugs is equivalent to SS of eect-site concentration estimated by the PKPD models. A simple arithmetic rule was initially applied to obtain input drugs' SS [174176],
113
considered.
(4.3)
for more than one minute than SS for the input drugs was
dierent drugs and therefore dierent units, with no physical meaning, nevertheless the
empirical limit set to detect SS is in accordance with the TCI administration scheme.
Figure 4.3: Steady-state detection in the input drugs' eect-site concentration using the
arithmetic rule.
DIFP &R
f (t) = f0 =
X
iI0
114
c0,i 0,i =
c1,i 1,i +
iI1
kk0
cj,i j,i +
iIj
dj,k j,k +
kkj
cJ,i J,i
J X
X
dl,k l,k =
dj,k j,k
(4.4)
j=1 kki
{z
}
|
High F requency
Components Scale 1 to J
| {z }
Low F requency
Component Scale J
j,i and j,k
l=1 kkl
iIj
where
j X
X
Wavelet Transform
dj,k =
where
0
|dj,k | j
, 1 j J, 0 k Kj
sign(dj,k )(|dj,k | j ) |dj,k | > j
(4.5)
At scale j=1 the WT modulus is dominated completely by noise and the threshold
value
can be assigned as the mean of the modulus maxima. The following threshold
j = 1 2(j1)/2 , 2 j J
(4.6)
Rapid changes in the signal may be detected using WT, since these changes in the
signal are identied as a maximum in the corresponding WT.
Two points, within a dened interval
tp ,
(4.7)
(4.8)
ta
to the left of
p1
115
tb
to the right of
p2
that satisfy
Threshold values
(4.9)
(4.10)
T1
and
T2
4.11.
T1 = 31 w,
where
T2 = w
(4.11)
=1
= 0 for unstable
for SS.
Wj f (t).
W Wj f (t),
To do that the WT on
The calculus of
(t)
f (t).
W WS f (t)
WS f (t)
scale, meaning the proper scale to analyze the WT where the WT represents the process
variations properly (Equation 4.12).
ts
(4.12)
if
then
(t) = 0,
where
Tu
if
then
(t) = 1,
where
Ts
|W WS f (t)| < Tw
then
(t) = 1
where
Tw
116
If
(t) = [(t)]
(4.13)
(4.14)
|W WS f | Tw
0
(|W WS f | Tw )/2Tw |W WS f | ]Tw , 3Tw [
=
1
|W WS f | 3Tw
(t) Tu
0
[(t)] Ts < (t) < Tu
(t) =
1
(t) Ts
where
(4.15)
(4.16)
used.
1
x Ts
(x) =
cos
+1
2
Tu Ts
The thresholds
Tu , Ts
and
Tw
(4.17)
manually selecting SS periods, and performing WT and second-order WT. Then the
standard deviation of the WT modulus
modulus
W Wf
Ts = Wf ,
where
Wf
Tu = 32 Wf ,
Tw = W Wf
(4.18)
Figure 4.4 shows the algorithm steps in the SS detection for both input and output
signals. Although for the collected data in this study, input SS detection as described
in Equation 4.3 is adequate (dashed rectangle in Figure 4.4), to increase the method
robustness and generality for it to be fully applied to dierent data sets, wavelet SS
detection was applied directly to the drugs' Ce and an input SS index was obtained
similar to the output SS combined index previously described [177]. The methodology
was also applied directly to the infusion rates, nevertheless for brevity, results will not
be presented here.
The quadratic spline wavelet function and scaling function described by Mallat and
Zhong [181] were used. Tables 4.10 and 4.11 present the nite impulse response lters
correspondent to the quadratic spline wavelet, and the normalization coecients used to
117
118
based algorithm.
steady-state index as summarized. Dashed line rectangle presents the initially used input detection rule, replaced by the wavelet
concentrations (Ce) as inputs, and output physiological variables entering the wavelet based algorithm to determine the combined
Figure 4.4: Diagram presenting the algorithm steps: the patient is the system, with propofol and remifentanil drugs' eect-site
-3
0,0078125
0,0078125
-2
0,054685
0,046875
0,171875
0,117188
-1
0,125
0,375
-2,0
-0,171875
0,65625
0,375
2,0
-0,054685
0,117188
0,125
-0,0078125
0,046875
j = 1, 0.
0,0078125
j > 5,
1,50
1,12
1,03
1,01
1,00
compute the wavelet transform. For each signal considered as output related to noxious
activation this method may be applied, with the adequate threshold detection.
The
signals analyzed in the previous section (phase I), and found to be related to noxious
activation were BIS, EMG, HR, SBP and PPGA, nevertheless an additional signal was
introduced in this analysis, the respiration rate (RespR). Although ventilation in all
collected cases was mechanically maintained, changes in the ventilation conditions may
occur by the anaesthesiologist choice. This may alter the equilibrium state of the patient
and induce changes in other physiological signals. Also, if the patient starts to recover
from neuromuscular blockade, and moves, this may be translated into spontaneous
ventilation, which may be detected and linked to arousal episodes. The following results
include data for the respiration rate, nevertheless when comparing the response to
precise noxious stimuli, the respiration rate was removed from the analysis, since it
would introduce bias due to the initiation of mechanical ventilation concomitant with
the laringoscopy/intubation stimulus.
A training set of ve patients was used to adjust the wavelet thresholds of the
119
Wf
WWf
Remifentanil Ce
0,08
0,06
0,02
0,00
Propofol Ce
0,04
0,05
0,01
0,00
Stimulus
6,35
0,32
0,05
0,00
BIS
16,27
3,04
2,90
0,78
EMG
23,25
10,49
0,33
0,11
HR
10,71
1,87
0,36
0,12
SBP
34,67
3,85
0,93
0,27
PPGA
1,68
0,18
0,11
0,03
RespR
4,64
0,45
0,22
0,10
SBP, PPGA and RespR. For each data set, the output signals were visually inspected
and sequences considered as in SS, selected to extract the thresholds later used in the
SS index processing, shown in Table 4.12. Following the method was applied to each
input and output signal individually, and individual SS indexes (Signali ,
i = 1, ..., 8)
m = 5)
n = 3)
SS
In (t) =
1X
Signali (t)
n
(4.19)
i=1
Out (t) =
1 X
Signali (t)
m
(4.20)
i=1
To inspect if the method was accurately detecting SS periods, the sequences considered as in SS used to dene the detection thresholds were analyzed. Of the total time
used in the threshold detection 91% was found to be in SS (Signali
(t)
(t) 0.9
for more than one minute than the signals were considered to be in
120
Figure 4.5: Detailed representation of steady-state detection of the systolic blood pressure trend, in one of the collected data sets.
Figure
4.6 shows the developed homeostasis index (Out ) for the same collected data set.
Table 4.13:
zed in this study phase was found to be in steady-state (SS) conditions (data as
meanstandard-deviation).
74,17,8
Propofol Ce
79,56,6
Stimulus
90,43,1
In
59,58,5
BIS
75,09,7
EMG
67,07,6
HR
32,617,1
SBP
40,319,3
PPGA
59,428,6
RespR
88,49,6)
Out
24,314,2
The methodology was applied to all data sets (N=31, Table 2.2), data presented
121
Figure 4.6: Representation of the input and output steady-state indexes. The bottom
trend may be interpreted as an homeostasis index of the subject considering BIS, heart
rate, systolic blood pressure, pulse wave amplitude and respiration rate.
122
set, signicant periods in SS and non-SS conditions were achieved. Table 4.13 shows
the time percentage that each physiological signal analyzed was in SS ((t)
0.9)
and
dynamic conditions, and also of input SS (In ) and output signals SS (Out ). The total
time analyzed, considering the 31 data sets, was of 6684 minutes, and average duration
per case of 21576 minutes.
Input/output simultaneous SS periods may also be extracted using this technique.
Periods lasting more than one minute of input and output simultaneous SS were extracted, and may be used in the construction of dynamic surfaces relating drugs, stimulus
and output measurable eect.
It was observed that input drug's SS was not always followed by output SS. This
is a demonstration that dierent factors interfere with the patient's homeostasis, being
noxious stimulation one of these factors. Also, it should be highlighted that SS does
not necessarily mean the patient is in optimum conditions, but only that the patient is
stable around a determined state of operation.
Table 4.14: Median values of the proposed homeostasis index (HI): average value in the
periods baseline, prior and post-stimulation, for each remifentanil eect-site concentration (RemiCe) group.
Baseline Laringoscopy
Pre
Post
Tetanic
Pre Post
Incision
Pre Post
HI
RemiCe=2.0
54,64
80,73
51,01
76,18
86,33
84,60
78,33
RemiCe=3.0
52,95
75,83
67,71
78,50
74,45
77,56
84,11
RemiCe=4.0
52,89
72,94
65,34
80,72
79,99
75,44
76,73
The proposed homeostasis index was calculated for all data sets, and the response to
precise stimulus in study phase I inspected to evaluate if the proposed index responded
adequately to the introduction of noxious stimulus, and dierent analgesic drug doses.
Tables 4.14 and 4.15 present the median values of the homeostasis index before and
following stimulation of the precise stimuli considered in study phase I, and amplitude
responses to the stimuli. The number of data sets used in this comparison is the same
123
Original
Baseline/Pre-Laringo Laringoscopy Tetanic Incision
HI
RemiCe=2.0
23,7
-24,64
10,22
-3,68
RemiCe=3.0
26,6
-12,22
-4,02
-3,84
RemiCe=4.0
22,0
-10,05
-4,91
5,53
Normalized (%)
Baseline/Pre-Laringo Laringoscopy Tetanic Incision
HI
RemiCe=2.0
46,2
-43,1
16,5
-6,2
RemiCe=3.0
44,4
-20,4
-6,7
-6,4
RemiCe=4.0
42,8
-18,5
-7,5
11,2
as described in phase I in section 4.2, 26 patients (see Table 4.1). Similarly to what
was presented in the phase I results (see subsection 4.2.2), non-parametric tests were
employed in the comparison of amplitude responses and analgesic impact. Statistical
dierences were found between pre and post intubation periods, with a decrease in the
homeostasis index, in a dose-dependent manner.
124
Figure 4.7: Schematic representation of the signal processing steps to obtain the individual and combined steady-state indexes, from data collection, to signal pre-processing
unit, and homeostasis index presentation (individual and combined), as implemented
in the developed tool for on-line homeostasis assessment.
In
and
Out
SS indexes, respectively. It
allows the visualization of each variable value every second, and also a graphical repre-
125
For
example, periods when infusion rate is null and in steady-state, due to a decrease in
target eect-site concentration, does not correspond to steady-state in Ce, since the
drug is being washed-out of the system.
Figure 4.8: Designed software for on-line assessment of the developed steady-state indexes: individually for each input (propofol and remifentanil Ce), output (bispectral
index, electromyography, heart rate, blood pressure, photoplethysmography wave amplitude, respiration rate), and combinations of input and output signals (In and
Out ).
Data presented in this example is from one of the patients of the study, simulating an
on-line assessment.
Figure 4.8 shows the developed tool, using data from one of the collected data sets,
to simulate the on-line implementation. Besides numerical information on the SS index
value, the software also processes the information on the direction of the variable shift,
126
127
Figure 4.9: Combined periods of steady-state (SS) input/output in all data sets: relation
between drugs' eect-site concentrations (Ce) and average heart rate (HR). The stimuli
intensities are represented in dierent colors, with white representing periods without
stimulation (0 in the Rasch scale), and increasing gray intensity with increasing stimulus
intensity.
intensity, based on the drug dose impact observed in the output signals, conditioned
by the stimulation intensity. This is similar to the process described by Bouillon et al.
[167]. Although estimating noxious stimuli intensities through this method would be
more direct, similar to what was described for the pre-intensities estimation, more data
for all stimuli would need to be collected in steady-state conditions, which arises practical and ethical issues. This issue may be surpassed using the developed Rasch scale
in Chapter 3, taking into account anaesthesiologists' experience, that was shown to be
related to the estimated pre-intensities in the pharmacological studies (see Figure 3.12
and Table 3.3).
4.3.4 Discussion
In this section a new measure of patients' homeostasis state is proposed. This measure
may be viewed as a method to evaluate patients' Noc/ANoc balance in response to
changes in drug doses and noxious stimulation.
noxious activation in the previous section were in this section employed in the calculus
128
129
Figure 4.10: Representation of the attenuation and depression eects of the analgesic
drug on a stimulus of known intensity [78, 167], with the introduction of an additional
depression curve.
An hierarchical approach may be used in the translation of the opioid drug eect
in stimulus attenuation, followed by the hypnotic eect in hypnosis maintenance, under
the incidence of the unblocked stimulus, as noted before [167].
balance it would be optimal if there was a combination of both stimulus attenuation and
130
polated based on the results of the Rasch analysis, presented in Chapter 3, since the
relation between the estimated pre-intensities and the Rasch score was shown to be close
to linear (see Figure 3.12 and Table 3.3). Knowing the stimulus pre-intensities, conditions are gathered to assess the analgesic impact (attenuation) on each noxious stimulus.
Table 4.16 shows the extrapolated values of pre-intensities using the estimated Rasch
scale developed in Chapter 3. Following, estimated pre-intensities (preopioid intensity )
were employed in the assessment of the perceived stimulus, using Equation 4.21, were
of remifentanil, and
=0.72
This step was achieved using both the registered noxious stimuli
throughout the procedures shown in Table 2.5 (scale), and the estimated remifentanil
eect-site concentration (opioid), to obtain a new variable, combination of these two,
the attenuated stimulus (postopioid
intensity ).
postopioid intensity=
preopioid intensity 1
opioid +(opioid
131
opioid
50 preopioid intensity)
(4.21)
Stimulus inten-
sities estimated and translated in the Rasch scale, as well as extrapolated values for the
pre-intensities, considering the linear relation shown in Figure 3.12. Table entries presented in bold correspond to the stimuli that have pre-intensities estimations; original
values presented between parenthesis.
Code (scale)
Rasch Scale
Extrapolated Pre-Intensities
0,00
0.00
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
4,58
4,99
5,55
4,76
4,49
4,17
7,11
7,20
4,81
6,43
5,27
6,16
6,81
5,07
5,07
4,41
5,36
5,16
4,81
2,35
4,56
4,64
6,15
3,80
3,88
3,37
4,41
4,46
5,65
5,07
5,21
3,74
4,64
4,32
0,99
1,08 (1,25)
1,20
1,03 (0,80)
0,97
0,90
1,53
1,55
1,04
1,39
1,14
1,33 (1,57)
1,47 (0,83)
1,09
1,09
0,95 (1,00)
1,16
1,11
1,04 (0,80)
0,51
0,98
1,00
1,33 (1,50)
0,82
0,84
0,73
0,95
0,96
1,22
1,09
1,12
0,81
1,00
0,93
132
preopioid intensity ,
T SS = preopioid intensity
Remif entanil Ce
3 ng/ml
(4.22)
where the relative weights of TSS were determined so that the clinical range of remifentanil matched the surgical range of stimulation in their study.
The estimated
postopioid intensity
the original Rasch scale stimulus intensity trend, and the remifentanil Ce in one of the
patients of the study. It may be observed that the new
postopioid intensity
and TSS
incorporate information on the stimulus intensity and attenuation provided by remifentanil. Although the trend is similar using both approaches, since the TSS is a simple
arithmetic rule based on a normalization of the remifentanil doses by the average expected, it allows the articial introduction of depression, evident when the administration
of remifentanil is initiated, and there is no stimulus.
evaluate which one of these descriptors better correspond to the physiological signals
objectively evaluated in the patient.
Depression information is important if one imagines a structure to model and predict
the patient's state in a timespan (advisory system), trying to predict the physiological
variables future value (see Figure 4.12), and an estimate of the perceived stimulus,
which may be seen as an estimate of the Noc/ANoc balance, since the objective of the
anaesthesiologist is to obtain maximum stimulus attenuation, without depression.
The impact that the hypnotic drug has on the dynamical relations represented in
Figure 4.12 should be highlighted.
the hypnotic only acts on the hypnosis state, suering interference of the attenuated
stimulus, the hypnotic and analgesic drugs exhibit a synergistic relation [167], and the
133
Figure 4.11: From top to bottom: stimulus intensity trend, using author's annotations
and intensities estimated by the Rasch analysis;
preopioid intensity
estimated based
postopioid intensity
preopioid intensity
and remifentanil attenuation; total surgical stress. Data from one of the patients in the
study.
134
Figure 4.12: Structure representing the impact of anaesthetic drugs and noxious stimulus on the physiological measurable eects.
Figure 4.13:
Obtaining the dynamical physiological model is the rst step to develop an estimator for the perceived stimulus, based on the measurable physiological variables linked
to noxious activation, and drug doses (see Figure 4.13). The approach followed in this
thesis was to consider dynamic relations between variables linked to noxious activation,
and provide a population model that may be used in the estimation of the expected
135
postopioid intensity
4.24 [182].
(4.23)
(4.24)
where Ts is the sampling time (1 s). Matlab was used for system identication, with
automatic order selection.
To evaluate model tness, the mean square error (M SE ) and error of t (F it
Error)
M SE =
n
X
(yh(i) y(i))2
i=1
F it Error =
n
X
|yh(i) y(i)|
i=1
136
(4.25)
(4.26)
yh
u, y
signal length.
To evaluate the adjusted models tness, correlations between the modelled TSS or
postopioid intensity ,
postopioid intensity
It should be highlighted that the model structure here presented implies causality,
and the construction of the model as here described is anti-causal (output precedes the
input response).
Knowing
postopioid intensity .
using bispectral index (BIS), frontal electromyography (EMG), heart rate (HR), systolic
blood pressure (SBP), pulse photoplethysmography wave amplitude (PPGA), and propofol eect-site concentration (PropCe). Data presented as meanstandard-deviation.
Model 1 (TSS)
Input Variables Fit Error
MSE Correlation
0,220,06
0,090,05
0,840,11
0,240,07
0,110,06
0,810,11
0,280,08
0,140,08
0,730,19
0,060,02
0,090,03
0,730,16
0,070,03
0,100,03
0,700,13
0,070,03
0,100,03
0,660,12
Table 4.17 shows the values of the model tness, using dierent combinations of input
and output signals (Model 1 and Model 2). The best order selected for the state-space
model was one, both for Model 1 and Model 2. It may be observed that correlations
are higher using TSS as the output signal, indicating perhaps TSS is translating the
perceived stimulus better than
postopioid intensity ,
on the physiological signals linked to noxious activation, tends to provide better results
in following TSS rather than
postopioid intensity .
Figure 4.14 shows the results for the individually adjusted model, in one of the
collected data sets for both TSS and
postopioid intensity .
137
(a) TSS
Figure 4.14: Modelled perceived stimulus, based on the stimulus intensity and attenuation provided by the analgesic remifentanil (output), and bispectral index, frontal electromyography, heart rate, systolic blood pressure, pulse plethysmography wave
amplitude, and propofol eect-site concentration as inputs. (a) Individually adjusted
state-space model of order 1 to TSS. (b) Individually adjusted state-space model of
order 1 to
postopioid intensity .
although both models follow the output trend the modeled TSS (correlation 0.90) tends
to give a more close relation to the output trend than
0.82).
postopioid intensity
(correlation
indicator of the patient's position on the Noc/ANoc balance, based on the available
information of both stimulus intensity and analgesic dose.
information is incomplete, and it may be observed that although the model follows the
trend of the estimated perceived stimulus, in some portions it is not capable of adequate
adjustment to the output original signal. This is due to a variety of factors: patient
individual response to noxious stimulation and drug attenuation, which may vary due
to several intrinsic and external factors, or to the poor estimation of the output signal.
It should be highlighted that although the t error statistics is in some of the data
sets high, the correlation between modelled trend and the TSS trend is also high. Figure
4.15a shows the result for the data set with higher correlation (correlation of 0.97).
There are however cases in which the correlation is below 0.7 (2 cases), demonstrating
that the trend describing the perceived stimulus may not be accurately describing the
patients' state. Figure 4.15b shows the results for the data set with lowest correlation
between modelled and original TSS (correlation of 0.51).
138
Figure 4.15: Modelled perceived stimulus, based on the stimulus intensity and attenuation provided by the analgesic remifentanil, using total surgical stimulus (TSS) as
output signal, and bispectral index, frontal electromyography, heart rate, systolic blood
pressure, pulse plethysmography wave amplitude, and propofol eect-site concentration
as inputs. (a) Individually adjusted state-space model of order 1, in the data set with
higher correlation to the output signal TSS (correlation of 0.97). (b) Individually adjusted state-space model of order 1, to the data set with lowest correlation (correlation
of 0.51).
Dierent sets of input signals were employed in the search of the dynamical model
translating the perceived stimulus: the rst set composed of BIS, EMG, cardiovascular
signals (HR, SBP, PPGA) and propofol Ce; the second composed of cardiovascular
signals and propofol Ce; and a third set composed only of cardiovascular signals. This
was done to evaluate the information contained in the signals found to be related to
noxious activation, in the continuous translation of the dynamic relations and Noc/ANoc
balance. EEG derived indexes, as referred previously contain information on noxious
activation, but this is usually found to be transient in time. In the adjusted models it
was found that incorporating information derived from the EEG monitor, cardiovascular
variables and propofol Ce interference provided the best results in terms of tness and
correlation to the original trend. Although a multivariate approach leads to the better
results, for practical reasons it should be highlighted that the fewer signals employed in
the determination of the Noc/ANoc balance, the more practical and general the method
is. It may be observed that correlation is decreased with the removal of the dierent
components of the input signals, and also thet the t error and MSE statistics increase.
139
This was done to obtain the best model for the Noc/ANoc
balance in the collected data, using the hospital's standard monitoring and drugs' TCI,
and two simpler models to be valid in dierent clinical setups (without BIS monitoring,
and without TCI administration).
data, and TSS was used as the output signal, since it was demonstrated that better
results are obtained when comparing to
postopioid intensity .
this analysis.
To reduce inter-patient variability, and set the start point of every signal to the same
origin (0), each signal (signal(t)) had the baseline value removed (signal(t)). Also,
due to observed dierences in the amplitude responses, the baseline value was also used
to normalize the response amplitude (signalN (t)), as given by Equation 4.27.
signalN (t) =
where
t=30 s
signal(t) mean(signal(t))
mean(signal(t))
(4.27)
corresponds to the rst 30 data points in the beginning of the data set.
140
Model 1 (TSS)
Input Variables
Fit
Baseline Removed
MSE Correlation
0,680,38
0,780,96
0,260,41
0,690,37
0,800,89
0,190,41
0,710,50
1,021,83
0,280,37
Baseline Normalization
BIS, EMG, HR, SBP, PPGA, PropCe
0,670,37
0,750,87
0,280,41
0,680,37
0,77085
0,240,41
0,710,49
1,001,75
0,320,36
The model adjusted using all collected data was simulated, and individual adjustment evaluated. The model did not provide adequate results when following the trend
of the individual data, nevertheless the model adjusted to the normalized data provided
better results. This is due to the inter-patient variability and inadequate translation
of the patients' state to the TSS, when regarding the amplitude responses that vary
between patients, even following normalization.
and correlations obtained in the population data for each method here described, using
baseline removed data and normalized data by the baseline value.
Figure 4.16 shows the results using the adjusted model to the merged data. Figure
4.16a shows the best individual result to the merged data model, and Figure 4.16b
shows the worst individual adjustment, considering correlation between modelled trend
and the original estimated TSS. Although correlation is lower using the population
model than the individually adjusted model, as expected, a similar trend is observed
between modelled and estimated TSS in almost all data sets, similar to what is shown in
141
Figure 4.16: Modelled perceived stimulus, based on the stimulus intensity and attenuation provided by the analgesic remifentanil, using total surgical stimulus (TSS) as
output signal, and bispectral index, frontal electromyography, heart rate, systolic blood
pressure, pulse plethysmography wave amplitude, and propofol eect-site concentration
as inputs. (a) Data set with higher correlation to the output signal TSS, considering
the merged data model. (b) Data set with lowest correlation, considering the merged
data model.
Figure 4.16b, with increasing and descending phases consistent with the TSS estimated,
nevertheless with drift, leading to low correlation between original and modelled trends.
Table 4.19 presents the average, minimum and maximum values observed for all data
sets considered in this phase of the analysis, for the merged data model, with the input
set of BIS, EMG, HR, SBP, PPGA and PropCe. The maintenance phase of anaesthesia
was considered in this analysis (Start 2nd Phase to Post-Operative Analgesia in
times
vector). The results presented in Table 4.19 correspond to the expected values of the
merged model. It was observed that the patients are kept with values of the perceived
stimulus below zero, meaning with a slight depression considering baseline values prior
drug administration.
adequately responds to the real trends in the patient, since following induction the
variables linked to noxious activation are usually depressed, and in practice tend to
stay below these values, since an increase in these variables may be linked to noxious
activation and arousal. Knowing this, the values observed and shown in Table 4.19 may
be regarded as the typical values, prior to depression and prior to arousal, since the data
collected with this clinical protocol only allowed for data collection in these conditions.
142
Data presented as
meanstandard-deviation.
Minimum
AverageStandard-Deviation
Maximum
Minimum
Average Maximum
-2,41
-1,76
-1,36
-0,960,55
-0,690,49
-0,330,42
-0,08
0,35
0,59
Table 4.20: Parameters of the state-space models using the merged data, considering
the three input sets.
A
B
C
D
A
B
C
D
A
B
C
D
-6,68E-08
3,77E-07
-1,71E-06
2,60E-09
-9,59E-08
0,0760
-0,3194
0,6402
-0,0025
-0,0632
609,7999
0,0386
-1,66E-06
2,59E-09
-2,17E-07
0,7033
-0,0024
-0,0545
612,4551
-0,4651
9,69E-07
8,55E-10
0,9878
-0,0041
-634,0687
-0,1743
Table 4.20 presents the parameters estimated for the merged data models, considering the three dierent input data sets.
143
Figure 4.17:
indexes, individually or for combinations of input and output signals (In and
Out ),
and the estimator of the perceived stimulus, based on the physiological model for the
population. Data presented in this example is from one of the patients of the study, for
a simulation of an on-line assessment.
The physiological models were implemented for an on-line assessment of the patients' perceived stimulus, in the previously described software for the homeostasis index
calculation. Figure 4.17 shows the nal version of the developed software that, based
on passive measures linked to noxious activation, provides homeostasis indexes, both
for input and output variables, and an estimate of the patients' perceived stimulus
(Noc/ANoc balance location).
4.4.4 Discussion
Dynamic models were adjusted to the individual data in the search for the best descriptors of patients' location regarding the Noc/ANoc balance. In order to be able to
estimate the patients' state, or perceived stimulus, the annotations kept by the author
during data collection, and translated in stimulus intensity trends as described in Chapter 3, were employed, using extrapolation to the
144
preopioid intensity
as described in
postopioid intensity
provide dierent trends, since the rst approach is based on the attenuation provided
by the analgesic, and the second is a simple arithmetic rule.
only provides information on the attenuation provided by the analgesic dose, while the
second method articially introduces the depression impact of the analgesic drug, as
shown in Figure 4.10. This may account for the superior results when modelling the
perceived stimulus using both approaches as output signals, indicating that the TSS
better translates the shift in patients state regarding the Noc/ANoc balance.
The model structure employed in this study implies causality, nevertheless the system under identication is anti-causal. Although this was the case, it was decided that
to produce models for on-line implementation the data were assumed to be causal, disregarding the evident anti-causal nature of the process. In fact, the modelled trend for
the individually adjusted models, followed satisfactorily the original output signal.
Regarding the order of the model, due to the characteristics of the output signal, the
best order estimated was of one, nevertheless, the output signal presents limitations,
since it only gives an indication of the patients state, disregarding the obvious variations
in stimulation intensity during the procedure. Also, although the TSS provided the best
results, the amplitude responses provided by stimulus and analgesic introduction may
not translate the real patients state.
Three dierent sets of input signals were considered in the analysis. This was done
to inspect the impact of dierent signals in the translation of the patients' Noc/ANoc
balance state. The best results obtained were with the model adjusted using all input
signals found to be related to noxious activation, and also the hypnotic impact. It was
observed that removing input signals deteriorates the simulation results, with higher
t error and MSE and lower correlation to the output trend. Nevertheless, to be able
to apply these results to more general clinical setups, the three approaches were here
considered.
The output TSS and order 1 model structure were chosen in the search of the merged
model to translate the patients' state regarding the Noc/ANoc balance, valid for the
population in the study. Due to the well known intra and inter-patient variability, different normalization methods were applied to improve comparability between patients,
in the search for the best model structure.
145
4.5 Summary
Collected data under general anaesthesia, following the clinical protocol described in
Chapter 2, allowed the inspection of the most informative physiological variables on the
noxious activation process.
146
4.5 Summary
BIS, SD BIS, EMG and SD EMG. These variables responded adequately to the introduction of a noxious stimulus, and to the analgesic drug attenuation.
The variables linked to noxious activation, and drug attenuation, were used in the
search of possible measures translating the patients' Noc/ANoc balance. Two dierent
perspectives on this subject were employed, the rst an homeostasis index, based on
wavelet analysis, and the second a model translating the patients' perceived stimulus.
Wavelet analysis was employed to the physiological signals linked to noxious activation, allowing the calculus of individual steady-state indexes, calibrated for each signal
characteristics. It was demonstrated that the proposed homeostasis index adequately
responds to precise stimuli and analgesic drug dose. Besides providing an estimate of
inputs and outputs steady-state, with detection of arousal or depression episodes, the
technique was also applied to evaluate if the designed clinical protocol was ecient in
the production of steady-state and dynamic conditions. It was observed that the protocol was ecient in the production of data in steady-state conditions, ideal to produce
steady-state valid models, and in the production of data in dynamic conditions, ideal to
produce dynamic models relating the physiological variables linked to noxious activation, later used in the modeling process. The steady-state index described allowed the
detection of periods of input/output simultaneous conditions, that may be used, when
more data is available, to model the relations of drugs and stimulus intensity on the
physiological signals linked to noxious activation.
Dynamic models were adjusted to assess the perceived stimulus (Noc/ANoc balance
related), taking as output the perceived stimulus based on the analgesic drug dose, and
the estimate of the stimulus intensity registered by the author (Rasch analysis). The
model takes into account the hypnotic eect since it interferes with the physiological variables linked to noxious activation, and allows the estimation of the patient's perceived
stimulus. The model used to describe the Noc/ANoc balance, using as inputs signals
linked to noxious activation, adequately responded to precise dierent noxious stimulus
and analgesic doses in the individually adjusted models. The population models obtained in this study need further enhancements, in order to translate the individualized
response (adaptation to the individual tolerance/sensitivity).
A software was designed, for on-line application, taking as inputs the physiological
signals linked to noxious activation and the input drugs' doses, allowing the on-line
calculus of inputs and outputs steady-state indexes individually, and as a combination of
these variables. Besides information on the patients' homeostasis state, the physiological
147
148
Chapter 5
Evoked Potentials - Active
Nociception Measures
It starts with a
state of the art on evoked potentials, with special focus on the somatosensory evoked
potentials application.
chapter is reported, describing its design, approval, informed consent management and
data collection in a volunteers' study. Finally, the collected data is analyzed to assess
the feasibility of this method to translate the Noc/ANoc balance in uncommunicative
individuals [183].
5.1 Introduction
Pain is one of the main concerns of human beings. It becomes known to all from a very
early age, and throughout life fear of pain, avoidance of pain, and dealing with pain
becomes a major concern experienced by all. Pain from a physiological point of view is
149
Understanding pain
perception and pathways from the sensing organs to processing in the brain is of great
importance when trying to develop tools to assess nociception.
Recent technological developments allowed researchers to literally see the active
brain areas using fMRI [184187], for dierent stimuli and interfering factors. Neuroimaging studies allowed the description of brain areas involved in the noxious stimulus
activation, and a better understanding of the processes triggered in the nociceptive
system.
The nociceptive system comprises several structural areas in the brain, responsible
for dierent processes related to pain perception.
activated are the primary and secondary somatosensory cortices (SI and SII), the posterior, mid and anterior insula, the anterior cingulate and the prefrontal cortices. Activity
in subsortical areas such as the periaqueductal gray, hypothalamus, amygdala, hippocampus, and cerebellum, have also been related to pain experience [187].
Nociceptive stimuli are conducted to the brain by C and A- bers, for later central
processing. The way this information is conducted and post-processed is the basis of
the use of evoked potentials to evaluate nervous conduction. This technique has been
used for many years by neurophysiologists for clinical diagnosis of neurological diseases.
In the 1970s, 1980s and 1990s, researchers explored the possibility of using somatosensory evoked potentials to study, and eventually quantify, pain/nociception. Studies
using somatosensory evoked potentials have linked evoked waves characteristics to the
intensity of stimulation or to the presence of anaesthetic drugs [188]. Understanding
the relations between wave characteristics, stimulation intensity, and drug eect, may
lead to a direct measurement of nociception.
Based on this principle, we decided to investigate, in adult volunteers, the use of
somatosensory evoked potentials to study pain/nociception, namely to see the eect
of dierent controlled stimulus intensities, and to evaluate the eect of anaesthetic
drugs on pain and evoked responses.
150
(AEP) and
use of evoked potentials has been of great importance in understanding the sensing and
brain mechanisms in response to outer stimuli both in human and animal subjects, and
also for diagnostic procedures mostly in neurological diseases.
Evoked potentials are produced in response to a stimulus in the nervous system,
by sensory, electrical, magnetic or cognitive stimulation; sensory evoked potentials are
evoked potentials produced by stimulation of the sensory system.
These responses
arise from action potentials or graded polysynaptic potentials during the propagation
of an electrical impulse from the periphery to the central nervous system, and can be
recorded over the scalp, as well as at various sites along the anatomic pathway, using
surface electrodes (e.g. Ag/AgCl electrodes) or subdermal needles [189].
Figure 5.1: Schematic representation of the stimulation pathway from stimulus site to
higher processing [190].
The use of evoked responses in anaesthesia has been explored mostly using middle
latency auditory evoked potentials to evaluate its hypnotic component state [8]. The
use of evoked potentials during anaesthesia, namely technical aspects of the evoked
151
natural sensation, producing a relatively short duration pain sensation, readily time-
152
(a) Cortical
(b) Cervical
Figure 5.2: Somatosensory evoked potential representation with latency, amplitude and
respective vertex potential marks (positive and negative): (a) cortical site; (b) cervical
site.
locked to the EEG. This type of stimulus is easily quantied and varied, allowing the
determination of the subjects' pain threshold.
For the selection of the stimulus intensity required to obtain an evoked response,
153
Figure 5.3: Equipment to obtain somatosensory evoked potentials, and process data
with time-locked stimulus [190].
154
lower stimulus' thresholds than men, and also that men were more tolerant to pain; this
suggests that gender may be one of the factors that condition the way pain is perceived.
Also ibuprofen only had signicant analgesic impact in men, increasing the stimulus
threshold.
mediated responses, in both animal and human subjects. It was found that there was an
increasing number of well-controlled experiments analyzing gender related dierences
with opioids administration, and also that gender dierences existed, but the direction
and magnitude of this dierence depended on many interacting variables related to the
drug (e.g. amount and pharmacology) and to the subject (e.g. genetics and age); for
example, Smith and Gray [196] explored the age-related dierences in response to opioid
155
due to laterality eects, since after stimulation waves are the same size or larger on
the contralateral side of stimulation, and stimulation of both sides of the body lead
to increased wave amplitudes, and decreased latency. Age and gender impact on the
evoked responses is an issue of debate, suggesting that the eects are not large when
comparing to population variability, however the general trend for all responses is of
increasing latency with age.
multiple sclerosis and other demyelinating diseases. Since these diseases have selective
action in the A- bers, C-bers remain intact, which may result in the absence of late
vertex potential and emergence of ultra-late response transmitted by C-bers [190].
The stimulus characteristics are also important in the evaluation of nociceptive
pathways. Carbon dioxide and argon lasers are used to induce thermal pain, however
the infra-red carbon dioxide laser causes a rapid increase in skin temperature, which
may result in supercial skin lesions, and the local inammation modies the nociceptive
pathway, making this kind of stimulus unfeasible for multiple stimulations. The visible
argon laser stimulus does not produce this kind of side-eects and can be applied for
cutaneous stimulation [190]. In order to develop a tool for pain assessment, the painful
stimulus intensity should be related to the amplitude of the evoked response, and also
the response should be attenuated by analgesic drugs. In previous studies researchers
evaluated the use of the late component of the vertex response (A- bers) and found
156
157
158
potentials.
Figure 5.4: Table extracted from [189], summarizing a review of studies evaluating the impact of anesthetic drugs on evoked
stimulation of the median nerve, with a 100 s constant current pulse generator, set
to 150% of motor threshold at 2 Hz for 256 stimuli. Collection electrodes were placed
over the contralateral hand area and over the fth cervical vertebra, both referrred
to a midfrontal reference. The authors did not nd dierences between latencies and
amplitude responses after drug administration in any of the collected evoked potentials,
stating that these drugs may allow the preservation of evoked responses when it is
necessary to assess ichaemia, nevertheless other studies demonstrated dierences on
evoked potentials' amplitude and latency, as shown in Figure 5.4.
Arendt-Nielsen and Bjerring [202] compared the use of topical and local injection of
analgesic, using laser-induced pain and SEPs. The stimulus was applied to the dorsum
of the left hand using an argon laser 200 ms duration, power adjustable 0.005-2.9 W
and wavelengths 0.488 m (blue) and 0.515 m (green). The stimulus was applied to
dierent sites to avoid sensitization and the time intervals between stimuli randomized
10-30s.
lidocaine produced a total sensory block almost immediately following injection (absence
of response), and following topical cream application both sensory and painful thresholds
were increased with evoked responses reecting the changes in perception.
Arendt-Nielsen et al. [203] evaluated alfentanil ecacy using SEPs in six healthy
volunteers. The stimulus used was an argon laser, with adjustable output power from
159
SEPs amplitude
was reduced signicantly by the nitrous oxide, when compared to the isourane. The
cervically recorded waves had no signicant dierences between drugs, meaning that
nitrous oxide does not appear to antagonize nociceptive transmission at that level.
Granovsky et al. [205] used a
(CHEPS),
stating that this stimulus modality provides a larger area stimulation than the laser
stimuli; nevertheless to avoid habituation it has to be moved around during the tests.
In this study the evoked responses by CHEPS were evaluated using only EEG derived
signals, 24 healthy volunteers. The stimulus temperatures were applied randomly, with
dierent latencies according to the stimulus temperature (time to heat and time to
return to baseline values). Each stimulus block was applied to the same body region,
and then a 15 minutes interval, and the thermode was moved in a random manner in
a 68 cm area. Evoked potentials were recorded from Fz, Cz, Pz, C3, C4, T7 and T8
(Easycap), the average of all selected channels was used as the reference. The authors
found positive correlation between evoked potentials amplitude and both stimulus intensities, and pain magnitude, being the second the main contributor to the evoked
brain response.
Ohara et al. [207] explored electrical activity of the brain using subdural electrodes
in response to laser evoked stimuli. The stimulus was provided using a Thulium YAG
160
Figure 5.5:
laser, with dierent energy levels (300-800 mJ), applied to the dorsum of the hand, recording the patients evaluation of pain perception in a numerical rating scale 0-10. The
electrical activity recorded from the subdural electrodes (
Electrocorticogram,
ECoG),
was amplied, ltered with a bandpass 0.1-300 Hz for the laser evoked potentials and
30-300 Hz for the somatosensory evoked potentials, with sampling rates above 1000 Hz.
The researchers found evoked responses over the SI, parasylvian and the medial frontal
cortex, and also the subjects' pain ratings were correlated to the intensity of the stimulus. The evoked peaks amplitude were also correlated to the intensity of the stimulus,
meaning that these areas may be involved in the stimulus intensity encoding.
In more recent studies further improvements in technology allowed the production
of functional images of the brain [187].
areas responsible for pain processing, pain sensing, pain intensity discrimination and
attention.
brief radiant impulses of thulium near infra-red radiation (wavelength 1.96 m, diameter
5mm, pulse duration of 1 ms). This laser allows a penetration depth of 360 m into the
human skin, without damaging the epidermis or aecting the subcutaneous tissue, and
causing temperature to rise in the supercial skin fast enough to elicit activation of
thinly myelinated A- and unmyelinated C-bers. The stimuli were applied to the back
of the hand in random intervals (8-12 s) to avoid the anticipation of stimulus, and the site
161
Oxygen Level-Dependent
Blood
(DLPFC) and
In both
162
(RIII).
Micalos et al. [119] found a relation between RIII threshold and the painful threshold,
indicating that the identication of the RIII threshold may be a valid reliable methodology in experimental pain studies. France et al. [210] proposed the use of EMG to
identify nociceptive exion reex, and von Dincklage et al. [209] designed a tool to continuously track the RIII threshold, that was demonstrated to be related to anaesthetic
doses, and response to noxious stimulation, with comparable prediction probability to
the BIS index in a study with twenty male volunteers [120].
Possible
limitation of the method is the use of muscle relaxants during general anaesthesia,
blocking EMG responses used in the assessment of the RIII.
In summary, SEPs have been used with success to assess nociception. Electrodes'
placement, stimulus' characteristics and site of stimulation diered among the dierent
studies. SEPs' stimulus should be reproducible, easy to quantify, well dened in time,
with little adaptation, only activating bers responsible for the pain processes, without
causing tissue damage. The stimuli most used were electrical, laser and contact heat.
The use of electrical stimulation activates motor responses, not reecting a pure reaction
to pain [211], nevertheless some studies refer the use of intradermal needles to apply the
stimulus, decreasing the motor interference in the evoked responses [190, 211]. Electrical
stimulation is an easy and repeatable stimulus to apply when studying evoked responses,
and poses as an attractive stimulus option. Laser stimulation has the advantage of only
activating C and A- bers, although there is the problem of adaptation; some authors
propose the change of site of stimulus between stimulations, preventing skin lesions and
habituation. Recent reports on the use of somatosensory evoked potentials refer the use
of contact-heat stimulators, to apply the stimulus. This stimulus has the advantage of
only activating bers responsible for the pain conduction, not evoking motor responses.
The attention factor may also interfere with SEPs, since when the subject is expecting
the stimulus there is a bias in the response due to stimulus anticipation. This may be a
problem in awake subjects, but it does not seem to pose as a problem for unconscious
subjects, nevertheless, changing the site of stimulation and randomizing stimulation
timing may be the solution for this issue.
163
Identify sensitive, motor and painful thresholds with correspondent sensory evoked
responses, in each volunteer.
Identify the most adequate EEG channel (electrodes positioning) to obtain the
somatosensory evoked potentials, in the translation of noxious stimulus intensity
and drugs impact.
164
An anaesthesiologist and an
software RugloopII
c (Demed).
Labgrab2.03
Nasal cannula;
BIS Vista
T M CR (Nicolet-Vyasis);
R
- monitoring and automatic ventilation;
T M monitor;
165
BIS sensors;
Propofol 1% Lipuro
Remifentanil Ultiva
R
Infusomat;
R
R
R
/Fresenius
in B.Braun
Perfusor syringe 50 ml;
R
R
in B.Braun
Perfusor syringe 50 ml;
R
Asena GH MKIII.
5.4.4 Methods
Figure 5.6: International 10-20 system for electroencephalogram collection. The letters
used stand for: F-frontal, T-temporal, C-central, P-parietal, O-occipital, and Z referring
to the electrodes positioned in the midline.
166
Channel 2: Neck (C5S, cervical on the spinous processes of the cervical vertebra
C5) - Scalp (Cz).
EEG data were collected continuously, with stimulus synchronization and evoked
potentials trends extracted. The monitor used to stimulate and obtain the SEPs was an
Endeavor CR (Nicolet-Vyasis), with stimulation rate of 3.1 Hz (around 180 stimulus/min),
with 0.5 ms duration per stimulus. The monitor allows the adjustment of the electrical
stimulus current, so that individual tuning is possible.
Before drugs' administration, all baseline variables were registered.
Then drugs'
administration was conducted using TCI of propofol, with Schnider PKPD model [50]
and remifentanil, with Minto PKPD model [39].
EXPERIMENTAL GUIDE:
1. Start stimulation according to recommendations, adjusting electrical stimulation
in each volunteer in current steps (0.1 mA) until reaching sensitive, motor and
painful thresholds for each individual. After threshold detection, a battery of 5
tests should be taken according to the scheme:
2. Execute the stimulus battery and obtain baseline SEPs, accompanied by volunteer's pain evaluation in a numerical rating scale (NRS) varying from 0 (no pain)
to 10 (worst possible pain).
SEPs.
3. Start remifentanil infusion using eect-site concentration steering. Set remifentanil eect-site to 0.5 ng/ml, after reaching the target, stimulate with the PT and
1.3PT stimulus, obtaining SEPs and volunteer self-evaluation in the NRS (2
minutes interval between stimulation).
4. Increase remifentanil eect-site concentration in 0.5 ng/ml steps and repeat stimulation; terminate increasing steps of remifentanil if arterial pressure or heart
rate decrease 20% of the baseline, or by clinical recommendation.
167
7. Decrease propofol dose to a light sedation level (1.2 g/ml) and obtain SEPs and
patient pain evaluation.
8. Restart remifentanil increasing steps of 0.5 ng/ml, obtaining SEPs and NRS evaluations in each step. Terminate the study as soon as the volunteer loses response
to verbal stimulus, but still responds to mechanical stimulation (OAAS=2).
Following the procedure above described, volunteers rested for a period of one hour
in the recovery room, receiving express indication not to drive a car, operate machinery
or precision mechanisms during that day.
SEPs waves' amplitudes and latencies were obtained as follows:
SEPs, the amplitude considered was the dierence between the positive and negative
peaks, and the latency the dierence between the latencies of the positive and negative
peaks (interpeak latency), this was done hypothesizing that wave morphology suered
an attenuation and expansion in time, besides an increase in latency in response to the
stimulus; for the cervical SEPs the amplitude considered was the amplitude of the only
peak and respective latency, as represented in Figure 5.2.
Remifentanil doses used in the study were dened according to recommended dose
ranges for spontaneous breathing patients as described in Appendix C.
In order to analyze the collected data and assess relations between stimulus intensity,
SEPs wave characteristics, pain evaluations in the NRS and drug doses, non-parametric
Spearman rank correlation coecient and paired sample Wilcoxon rank sign test were
used. P<0.05 was considered statistically signicant.
168
5.5 Results
picted in Figure 2.2. However, this study was conducted in volunteers, and therefore,
beyond the reach of usual clinical practice with increased costs for the hospital. Due to
the use of hospital's resources the clinical protocol was also evaluated by the Hospital's
Financial Department to assess study expenses.
5.5 Results
Figure 5.7: Clinical setup used in the volunteer's study data collection: A) SEPs monitor and stimulation device Endeavor
electrodes placed on volunteer median nerve; C) Aisys monitor and ventilator; D) BIS
sensor and SEPs needles montage.
After obtaining hospital's approval, data were collected during three Saturday mornings. All volunteers arrived with fasting over 6 hours, and following the experiment
169
5/5
Age (years)
29.59.1
Weight (kg)
63.08.9
Height (cm)
171.47.2
Table 5.2: Propofol and remifentanil eect-site concentration targets and interruption
rules applied in the study: the study began with remifentanil administration only, in increasing steps as described in the text, until one of the interruption rules occurred; after
interruption the drug doses were reduced to the rst step considered in the next phase,
and drug doses were increased again according to the scheme.
sequent schemes, rst remifentanil only, followed by constant remifentanil and propofol
increasing steps, and nally constant propofol and remifentanil increasing steps.
Remifentanil Only
Remifentanil and
Propofol
Propofol and
Remifentanil
Propofol
Remifentanil
Propofol
Remifentanil
Propofol
Remifentanil
Ce (g/ml)
Ce (ng/ml)
Ce (g/ml)
Ce (ng/ml)
Ce (g/ml)
Ce (ng/ml)
1.2
1.2
1.2
1.2
2.5
2.5
2.5
1.2
increasing steps
increasing steps
increasing steps
Ten healthy adult volunteers (see Table 5.1 for demographic data), were enrolled
170
5.5 Results
in the study. One subject was excluded from the analysis due to the impossibility of
obtaining SEPs evaluation for the PT stimulus.
Figure 5.8: Collected cortical somatosensory evoked potential waves, with dened peaks
by the neurophysiology technicians.
SEPs waves were punctually collected for each drug combination, and delimited by
experienced neurophysiology technicians (Neurinbloc). Due to the nature of the process,
and being the volunteers forcefully fully awake and freely moving, data were prone to the
occurrence of outlier values, specially during painful stimulation. Figure 5.8 presents a
pair of cortical SEPs waves collected during the experiment; observing the waves while
5.8a is very well dened, similar to the wave outlined in Figure 5.2, 5.8b is not as well
dened (contamination).
Remifentanil and propofol administration was done by stepwise increments as outlined in Table 5.2, using TCI. Interruption rules described in Table 5.2 include a 20%
drop in BP and HR from baseline values, loss of response to verbal command with
response to mechanical stimulation (OAAS=2, Table 1.1), or interruption by clinical
indication.
Drug increments were stopped every time the volunteer was unable to respond to
verbal command, while being capable of responding to mechanical stimulation; this was
achieved in all volunteers. Brief periods of apnea occurred, all spontaneously reversed
requiring only neck extension and drugs' administration interruption.
BIS minimum values reached were 52.38.6 (see Figure 5.9).
NRS varied from 0 to 8, and 1.3PT stimulation currents varied between 6 mA and
52 mA (see Figure 5.10).
171
172
5.5 Results
Figure 5.10: Sensitive (ST), motor (MT), painful (PT) and 1.3 painful (1.3PT) threshold currents of stimulation for each volunteer.
Volunteers evaluated pain in a NRS both when the PT was achieved, and for every
1.3PT stimulus applied subsequently. As soon as the stimulus was enough to produce
pain, the NRS evaluations were higher or equal to 5 in all volunteers except for one.
Figure 5.12 presents the individual evaluations for the PT and 1.3PT stimuli applied
prior to drug administration. From the PT to the 1.3PT stimulus, the evaluations in
the NRS were maintained or increased.
5.5.2 Numerical Rating Scale Evaluations and Remifentanil and Propofol Concentrations
Remifentanil administration, at increasing concentrations, resulted in a reduction in
pain evaluations in the NRS. The same happened when propofol was administered. Table 5.3 presents the NRS evaluations for the maximum drugs' combinations achieved
in each administration scheme (no drugs, increasing remifentanil and no propofol, increasing propofol with constant remifentanil, and nally increasing remifentanil with
constant propofol). Although a decrease in the NRS evaluations was observed for all
administration schemes, in the last two combinations, the volunteer either rated pain as
0, or was unable to rate pain due to becoming unresponsive (lack of verbal response).
173
174
SEP Amp
SEP Amp0
(5.1)
5.5 Results
Figure 5.12: Numerical rating scale pain evaluations (0-10) for each volunteer both for
the painful threshold (PT) and 1.3PT stimuli.
Table 5.3: Numerical rating scale (NRS) evaluations in each volunteer for the baseline
and maximum drug doses combination achieved in each administration scheme (NE no evaluation).
NRS Evaluations
Volunteer
Increasing
No
Increasing
and Constant
Remifentanil and
Drugs
Remifentanil
Remifentanil
Constant Propofol
(1.0 ng/ml)
(1.2 g/ml)
5
NE
0
NE
NE
NE
NE
NE
NE
NE
0
0
NE
0
NE
NE
2
0
1
2
3
5
6
7
8
9
10
where
Increasing Propofol
SEP Amp
8
6
8
8
8
8
9
9
7
2
4
3
6
1
2
0
2
0
SEP Amp0
175
and
RN orm
as
R=
where
SEP Lat
(5.2)
(5.3)
RN orm =
where
SEP Amp
SEP Lat
SEP Lat
SEP Lat0
(5.4)
5.5.4 Cortical Somatosensory Evoked Responses and Numerical Rating Scale Evaluations
Table 5.4: Spearman correlation coecient (rho) for each volunteer between the normalized cortical somatosensory evoked potentials ratio (RN orm ), numerical rating scale
evaluations, and remifentanil eect-site concentration in the rst administration scheme,
where remifentanil is administered in ascending steps, without propofol (N is the number of points used to determine the correlation coecient, and * stands for signicant
correlation with P<0.05).
Volunteer
0,4564
14
-0,6347
0,4343
13
0,1078
0,5934*
18
-0,5593
10
0,4543
-0,7885*
0,2592
11
-0,0904
0,5465*
21
-0,2325
16
0,2285
16
-0,1734
15
0,3898
18
-0,4869
13
10
0,5007*
17
-0,6242
10
MeanStandard-Deviation
0,42920,1214
176
-0,38690,3002
5.5 Results
Table 5.5:
cortical somatosensory evoked potentials normalized ratio (RN orm ) and the numerical rating scale evaluations, and drug's eect-site concentration in each administration
scheme (where N is the number of points used to determine the correlation coecient).
N Variables
0,1678
0,0500
137
-0,1762
0,0860
96
Remifentanil vs
-0,1251
0,4076
46
-0,0231
0,9000
32
NRS vs
RN orm
RN orm
RN orm
vs RN orm
The relationship between SEPs and reported pain is shown in Figures 5.13 and 5.14
as the result of the ratio and normalization of SEPs characteristics (1.3PT stimulation)
versus the NRS evaluations. A decrease in cortical SEPs amplitude with decreasing pain
ratings in the NRS (see Figure 5.13) was observed.
inter-variability was reduced, while the relations were maintained (see Figure 5.14). All
volunteers exhibited the same direction of relation, conrmed by the Spearman rank
correlation coecients: in all volunteers the correlation coecients were positive, but
signicant in only three, the ones with the larger number of evaluation points (see Table
5.4).
Since after normalization the obtained ratios were comparable between subjects and
dimensionless, data from all volunteers was aggregated to obtain a population correlation coecient (see Table 5.5). The correlation coecient between SEPs normalized
ratio and the evaluations in the NRS exhibited a positive trend, although not statistically signicant.
177
volunteer and possible outlier interference (in fact the remifentanil only administration
scheme has the larger number of evaluation points).
Spearman rank correlation coecients between cortical SEPs normalized ratio and
the remifentanil eect-site concentrations were calculated for each volunteer. Almost all
volunteers exhibited the same direction in the relation between increasing remifentanil
doses and the SEPs ratio, however statistical signicance was achieved in only one
volunteer (see Table 5.4). Table 5.5 presents the Spearman rank correlation coecients
for the population data between the cortical normalized ratio and drugs' eect-site
concentration in each administration scheme (decrease of the ratio with increasing drug
178
5.5 Results
The evoked potential in this case was dierent from the evoked
179
180
5.5 Results
ratings did not show a direct relation (see Figure 5.16); also, a relation between these
variables and the drugs eect-site concentrations was not observed (see Figure 5.17).
Correlations were obtained, demonstrating that the relations are not regular: for
example, correlation between NRS evaluations and the SEPs ratio obtained for each
volunteer presented dierent directions, with average value of 0.130.25.
The same
happened for the SEPs ratio versus remifentanil Ce, with average value of -0.090.40.
181
Figure 5.17: Normalized cervical SEP ratio versus remifentanil eect-site concentrations
for each volunteer.
Figure 5.18: Boxplot of the normalized SEP amplitude/normalized SEP latency ratio
(RN orm ) for the maximum drug doses combination achieved in each volunteer in each
administration scheme: 1 - No drugs; 2 - Remifentanil only; 3 - Propofol increasing steps
and constant remifentanil (1.0 ng/ml); 4 - Remifentanil increasing steps and constant
propofol (1.2 g/ml).
182
5.5 Results
The non-
may lead to the production of linear relations between the analyzed variables.
The assessment of the normalized ratio, possibly with a logarithmic transformation
seems to be an adequate method to assess the Noc/ANoc balance.
183
It was related
to changes in drug doses and to the reported pain by the volunteers in the NRS. It
produces comparable measures between subjects, dimensionless and normalized for the
individually adjusted SEP in the painful threshold.
Figure 5.20 presents the original data points of each combination of propofol and
remifentanil, versus the SEPs normalized ratio (RN orm ) obtained during each combination, for all volunteers.
doses (eect-site concentration, that in the case was in equilibrium with the plasma
concentration) and the measurable eect
RN orm ,
Noc/ANoc balance.
184
5.6 Discussion
Figure 5.20: Collected data points from all volunteers considering propofol and remifentanil eect-site concentrations (Ce) as inputs, and the normalized SEPs ratio (RN orm )
as output measurable eect.
More data, with a dierent study design (during general anaesthesia, with unconscious patients) is needed to adjust a dynamical model to be used to assess the relation
between propofol, remifentanil and output measurable eect
RN orm ,
in other words to
5.6 Discussion
In this study a new method, combining both amplitude and latency of cortical SEPs,
is proposed as an objective indicator of the Noc/ANoc balance.
Propofol and remifentanil increasing doses lead to cortical SEPs wave depression
with amplitude reduction, increased latency, and simultaneously to decreased NRS ratings. These observations corroborate results from previous studies [188, 189], indicating
a depression in wave amplitude, and increase in latency, with increasing doses of propofol and remifentanil. In the present study the information contained on SEPs' wave
morphology was assessed, in cooperative volunteers, in response to dierent stimulus intensities, and in response to hypnotic and analgesic drugs, allowing for a more complete
analysis of SEPs applicability in the assessment of Noc/ANoc balance.
185
It
was observed that this information was additive, and that the relation between NRS,
drug doses and the ratio were maintained. Following the same approach as before, the
ratio was improved by using the normalized SEPs amplitude and latency, which led to
the construction of a dimensionless measure, comparable between subjects.
SEPs may be used in the future to translate the nociceptive/drug attenuation balance not only in conscious subjects, but most importantly in individuals not capable
of communicating verbally or in other ways, due to disease, altered mental or consciousness states, or young age, since the results demonstrate that the wave depression is
accompanied by a decrease in conscious pain perception (decreased NRS ratings).
No relations between cervical SEPs wave characteristics and both NRS ratings and
drug doses were found.
collected prior higher processing and therefore, is only representative of the stimulus
conduction before higher processing.
milar in all drug steps, may indicate that propofol and remifentanil do not antagonize
nociception transmission at this level, and similar wave characteristics are exhibited
[204].
186
5.6 Discussion
It must be highlighted that the pain evaluations are in a NRS, and therefore have
limited application. These scales do not guarantee linear relations between entries, they
are just ordinal scales [142, 144], and therefore the linear regressions presented in the
gures are just indicative of the direction of the relations. The correlation coecient
used to express the relation was the non-parametric Spearman rank correlation coecient, which does not imply a linear relation in the data, solely that the direction of
increase is maintained (increase in
y,
not necessarily in a linear fashion). Also note that the number of points to determine
the correlation coecients individually is, in many cases, low and therefore with limited
meaning. To strengthen the results presented in this study the number of volunteers
should be increased.
Although statistical signicance was reached in the collected data, the eect of outliers may not be disregarded. Besides the inter-volunteer variability, volunteers were
conscious throughout most of the procedure, and were not paralyzed, so that movement
was present and a possible cause of artifacts. Although minimization of the incidence
and eect of outliers was a concern, the SEPs measurements obtained correspond to
punctual moments, and therefore may be contaminated or biased. This may dicult
data interpretation, and more data should be collected to minimize this eect.
Ne-
vertheless, signicant results were obtained, leading to the assumption that the SEPs
may translate the nociception/anti-nociception balance.
The way pain is perceived is conditioned by many factors, eliciting noxious responses with known precise stimulus, adjusted to individual needs, and observing evoked
responses after central processing is a direct way of measuring and understanding the
nociceptive phenomenon. SEPs amplitudes are depressed by the use of analgesic and
hypnotic drugs, but more important the SEPs amplitudes are related to the pain ratings
of the volunteers, showing that this may congure an objective method of measuring
nociceptive balance in uncommunicative patients.
The fact that the hypnotic drug had an impact on the SEPs wave amplitude, as
well as the analgesic drug, should be reinforced.
analgesic drug should be the only one blocking the stimulus, in fact observations show
a decrease in wave amplitude accompanied by a decrease in the NRS ratings provided
by the volunteers, corroborating also the decrease in the SEPs amplitude in response to
propofol, which may be interfering with pain perception. It was clearly observed that
the addition of propofol resulted in SEPs depression accompanied by a decrease in pain
187
The
fact that the SEPs waves characteristics are related both to stimulus intensity and drug
attenuation, responds to the basic requirements of a Noc/ANoc index.
The proposed index, based on the ratio of the normalized SEPs amplitudes and
latencies, presents itself as an easy to obtain measure, since the PT and 1.3PT may
be easily obtained to calibrate and produce individually adjusted measures, normalized
to an dimensionless index (RN orm ).
comparability between subjects, since it was observed that the inter-volunteer variability
decreased and extreme values were brought together to the main direction observed in
the population data.
In order to continue exploring the potential value in this method, studies in patients
under general anaesthesia with BIS guided and controlled hypnosis should be conducted,
in order to evaluate the impact both of dierent outer stimulation intensities, and of
dierent analgesic levels on the
RN orm
188
propofol and remifentanil, as represented in Figure 5.20, the number of data points is
small, as well as the dispersion and number of drugs' combinations. This was the case
due to protocol features, which was designed to evaluate the impact of dierent drugs'
doses on pain sensation and SEPs wave characteristics, to inspect if SEPs translated
the Noc/ANoc balance. Now, with the results in the volunteers study pointing in the
direction of SEPs as an objective tool to assess the Noc/ANoc balance, a new study
should be designed to improve technical aspects of data collection and processing, clinically validate the tool and identify and adjust the dynamical model relating Cp to Ce
and Ce to measurable eect.
Figure 5.21: Representation of a possible patch to place on the arm for example, to
administer the somatosensory stimulus, with multiple sources of contact heat or lasers,
that would be activated randomly, and irregularly in time.
One issue that in this phase of the study did not assume great importance, since the
objective was to inspect the feasibility of the method, was the fact that the SEPs waves
were obtained punctually, not in continuous mode.
The
A solution for
this problem would be to have dierent sites of stimulation, that would be activated
189
5.8 Summary
In this chapter the use of evoked potentials with special focus on somatosensory evoked
potentials, has been reviewed. Evoked potentials are elicited responses to a well dened
and precise in time stimulus. The use of evoked potentials in clinical practice is fully
described with special interest in the evaluation of sensory conduction in high risk
procedures, and as diagnostic tests in some pathologies aecting sensory conduction.
Sensory evoked potentials wave characteristics are aected by several factors including genetic predisposition, leading to wide interindividual variability in sensing thresholds and amplitude responses. Anaesthetic drugs interfere with evoked potentials'
waves, with dierent impacts according to the drug site of action and eect.
The use of evoked potentials in anaesthesia is not a new concept, since auditory
evoked potentials have been used to assess the hypnosis state of a patient during general
anaesthesia, with a commercially available monitor translating the patient's hypnosis
state; some authors have initiated the study of somatosensory evoked potentials in
the search for a tool to assess nociception, nevertheless these studies were made with
dierent technological means and never proceeded to the nal step of designing a tool,
fully describing the balance between nociceptive activation and drugs attenuation.
Understanding the way noxious stimulus are processed by the human body allows a
more direct assessment of nociceptive responses, since the stimulus conduction and evoked responses are directly related to the stimulation intensity, and drugs' attenuation.
The use of SEPs in the assessment of noxious stimulation and drugs attenuation in
healthy adult volunteers was investigated. A clinical protocol was designed and submitted for institutional approval. Following institutional approval and written informed
consent, data were collected and subsequently analyzed.
It was observed that both remifentanil and propofol had impact on SEPs amplitude
and latencies, with a correspondent change in the NRS evaluations. The results suggest
190
5.8 Summary
that the SEPs waves contain information on the balance between noxious stimulation
and drugs' attenuation.
A great interindividual variability in stimulation intensity and response to individually adjusted thresholds were also observed. After analyzing the data the amplitude
and latencies were normalized using the observed SEP in response to the PT with no
drugs in the system, which led to a decrease in interindividual variability.
A new method combining both the amplitude and latency of the SEPs is proposed, taking in advantage the known inverse proportionality of the two measures with
increasing drug doses. Normalization also introduces improvement since the ratio interindividual variability is decreased.
Somatosensory evoked potentials wave amplitude and latency were related to the
individual nociception and drug attenuation balance, and may be used to translate the
individual Noc/ANoc balance in uncommunicative individuals.
191
192
Chapter 6
On Nociception Control
This chapter addresses the Noc/ANoc balance control, and discusses the benets and
drawbacks of automatic administration of anaesthetic drugs, as well as possible methodologies to approach this problem's specic demands. It is also proposed a nonlinear
control technique to be applied to the hypnosis control, since it poses as a parallel problem to the Noc/ANoc balance control, and it already has validated monitors
[55, 218, 219].
6.1 Introduction
The use of automation in medicine is gaining increasing interest due to the potential
improvement it may bring to patient care. Although the application of these techniques
in life sciences is not recent nor a novelty [220, 221], technological advances have facilitated its application in clinical environments such as an operating room, to improve
performance in complex tasks [220].
193
6. ON NOCICEPTION CONTROL
Figure 6.1: Schematic representation of the analogy between the automatic pilot systems
in the airplane, and the anaesthesiologists' action during general anaesthesia: induction,
maintenance, noxious stimulation regulation and recovery.
Anaesthesiology is one of the medical specialties that has taken an interest in this
issue, specially for the automatic control of drugs' administration during general anaesthesia, in the search of the optimum drug dose titration, which may lead to cost
reduction and shortened post-operative recovery phases [222]. The introduction of new
monitoring devices, new faster acting and excreted anaesthetics, as well as the development of fast and controllable actuators for the administration of these drugs (e.g.
remotely controlled infusion rate in electronic syringe pumps), has allowed the development of automatic control systems for the administration of anaesthetic drugs [121, 127].
Researchers have long addressed the issue of closed-loop control in anaesthesia, proposing dierent techniques for the administration of muscle relaxants [124126], hypnotics [55, 121123, 127] and analgesics [21, 127]. Nevertheless, the introduction of these
techniques carries safety and ethical issues, that need to be analyzed with care before
a fully automatic system may take over the control of drugs' administration [129].
The idea of closing the loop to provide a more safe and accurate anesthesia is
appealing, both for health professionals and patients, but machines are blind to a variety
of unexpected events that may occur, and often do, in the daily clinical practice, and
may not be disregarded. Anaesthesiologists' action is usually compared to piloting an
airplane with the automatic pilot systems. This is so mostly because of the known fact
that both airplane piloting and anaesthesiology are at the top when it comes to safety
regulations (check lists), safety protocols, and error issues, because both imply heavy
technology and complex decision processes taking in account many variables.
Also
because of the similarities between a ight and anaesthesia: take o of a plane is like
induction of anaesthesia, the lack of control over one's life in a ight being comparable
194
6.1 Introduction
to loss of consciousness; as an airplane passenger knows of the risks, however remote
of crash of the aircraft, so the patient submitted to anaesthesia is aware of the risks
of never regaining consciousness.
anaesthesia, with many variables potentially causing turbulence and requiring action to
maintain stability. Landing is like the recovery phase of anaesthesia and regaining of
consciousness, a smooth landing being comparable to rapid and stable ROC. Airplane's
pilots are responsible for the lives of many people on each ight, anaesthesiologists are
responsible for the life of just one.
the sky (loss of consciousness), maintaining the plane in ight height (maintenance of
anaesthesia), leading the safe landing of the airplane (recovery of consciousness), may
be extended by introducing the air 'bumps' during ight to translate noxious activation
in response to noxious stimulation, or unexpected events such as blood loss, which act
as interferences in the human system altering pharmacokinetics for example (see Figure
6.1). Figure 3.11 in Chapter 2 is representative of these interferences, varying both in
intensity and duration; the anaesthesiologist needs to anticipate/react in accordance,
to keep patient's homeostasis.
ago and is currently widely accepted. Although researchers have developed and used
experimentally automatic administration of anaesthetics in the clinical setting, not a
single automation device has been commercialized.
The rst question that arises in this discussion is: will the
195
6. ON NOCICEPTION CONTROL
network, that could reproduce the anaesthesiologist's decision processes responding to
every possible scenario with the best solution.
resides in the consideration of every possible scenario, and in the response to particular
conditions, not considered in the design phase. Nevertheless, this is a problem both with
automation and with the original human controller, since anaesthesiologists rely on the
experience and learning to deal with similar cases, and are therefore prone to (human)
error. The problem with a full automatic control resides in the danger of ignoring important factors that may be present in the scenario and, to achieve the minimum error
possible, to over or sub-dose the patient in that specic condition. A straightforward
solution is to never allow the system to exceed the maximum or minimum drugs' recommended doses, nevertheless this solution may not be sucient to guarantee patient
safety, due to wide variability in individual responses to drug administration.
From manual control to fully automatic drug administration, dierent levels of utility
and feasibility may be considered:
1.
Advisory system:
Advisory/automatic system:
the physiological monitored data and drug consumption, but the anaesthesiologist
may override the system, to manually control the drug doses, and switch back (or
not) to automatic control.
3.
Automatic system:
These three levels dier in concept and control methodologies. The Advisory System
will only serve as an advisor to recommend drug adjustments and direction's to take to
the anaesthesiologist; it does not require the anaesthesiologist to follow the advice, and
it is prepared to simply continue on predicting the best solution to lead the patient into
optimum conditions (reference target).
The Advisory/Automatic System level is conceptually dierent, since it starts as an
automatic control system, but it may be overriden by the clinician if he/she nds it
196
6.1 Introduction
necessary; the concept is to allow the clinician to adjust drug doses according to his/her
clinical perception/decision, with the system working during this period as an advisory
system.
Resulting from this intervention, some technical problems may arise: when
the clinician takes over control, no problem should arise nevertheless, the system must
be robust enough to admit the return to the automatic control mode.
The problem
is that conceptually the anaesthesiologist considers that the automatic control was not
satisfactory enough (changes in the patient conditions not considered in the controller
design and patient modelling), and he/she overrides the system to lead the patient to
a new optimum condition. This is feasible, the problem residing in the return to the
automatic control, since the controller is blind to the conditions update and may present
an abrupt reaction to this change, over or sub-dosing the patient to guarantee a rapid
return to the target conditions. This problem may be solved using three strategies:
not allowing the return to the automatic control, and solely allow for manual
control from that point on;
limiting the control action when returning to the automatic control, in order to
avoid sudden control actions (this solution has limited application, since the controller will eventually lead the patient to the previously dened state of operation);
changing the state of operation (controller target), guaranteeing that the the system will not overshoot or undershoot when returning to the automatic control
mode, maintaining the new route dened by the anaesthesiologist.
far more ecient than the original control agent, the anaesthesiologist, before it can be
implemented in clinical practice.
So far the generic problems with automatic control systems in the context of this
application were discussed:
anaesthesia genesis that need to be approached in the design of a robust controller regarding the patient as a whole. The next subsection will focus on the concept of general
anaesthesia, the proposed methodologies to control the administration of anaesthetic
drugs, and the future of control systems in the eld of anaesthesia.
197
6. ON NOCICEPTION CONTROL
Some authors have followed this route in the search for a multi-variable control of general
anaesthesia [127, 221, 222, 226].
First there is the known dynamic relation between the analgesic and hypnotic drugs,
and their synergism, leading to lower hypnotic consumptions when in presence of an
analgesic (see Equation 1.7). Next, the impact that the hypnotic drug may have in the
noxious responses, which was observed in the volunteers study, with the decrease in
the pain NRS evaluations in response to increasing propofol doses. This has also been
referred in some pharmacological studies [34, 46, 216, 217].
xant, which may be controlled independently from the remaining general anaesthesia
components, but may interfere in the data collection quality for the monitoring of the
remaining components, as for example, in a paralyzed patient, where the risk of outliers
occurrence in data collection is diminished, meaning the risk of obtaining biased erroneous values due to muscle activity and artifacts is decreased (e.g. frontal EEG derived
indexes contamination by frontal EMG activity) [7]. Figure 6.2 is a schematic representation of the possible interactions between the considered anaesthesia components and
corresponding drugs, be it on each other (direct interference) or as indirect interference,
through the introduction of erroneous measurements in the monitored physiological
variables used to assess the patient's state.
the only component that may be controlled independently from the other two, since
those two do not interfere directly or indirectly in the monitoring and control of muscle
paralysis. Analgesia and hypnosis, may both be contaminated by movement artifacts,
depending on the variables used to asses the patient's state concerning these general
anaesthesia components, and also they interfere with each other. Attempts to control
the hypnosis component independently from the analgesia have been done, using spinal
analgesia, and guaranteeing independence from noxious activation interferences [227].
198
199
hypnosis and analgesia components, and nally dash-dot arrow represents the impact that the hypnotic drug may have on the
dashed arrows represent the impact that the paralysis state of the patient has on the monitoring devices used to assess the
indirect interference on each other: line arrows represent the synergy relation between the hypnotic and the analgesic drugs,
Figure 6.2: Schematic representation of the general anaesthesia triad, with corresponding anaesthetic drugs, and their direct or
6. ON NOCICEPTION CONTROL
Another way to minimize outer interferences is to have the patient totally paralyzed.
This was unpractical due to longer recovery times, leading to unnecessary drug and
time consumption, but recently a new drug capable of rapid full reversal of deep levels
of muscle paralysis has been introduced (sugammadex). It may be used in the future in
daily clinical practice to allow maintaining patients fully paralyzed until the end of the
procedure, and rapidly reverse paralysis. This might allow more reliable hypnosis and
analgesia measurements, and improvements on the way these components are assessed
and controlled.
The control of each component of anaesthesia separately is motivated by the lack
of an adequate measure for the analgesia component, and also of the description of the
dynamic interactions between the three components, which needs more research to be
fully explained, although some attempts have been made in the multi-variable control
of anaesthesia [127, 220, 221].
well to assess muscle paralysis and hypnosis. For muscle paralysis, EMG, mechanomyography (MMG) and acceleromyography (AMG) are used to translate the patients state
in response to an electrical stimulus, EMG being the less prone to mechanical interference in signal collection [222], which is important from the control engineer's point
of view.
For hypnosis, monitors using EEG and AEP are currently widely accepted
into clinical practice, and validated in the translation of the patient's state regarding
this component. Several studies used these monitors to close the loop in the hypnotic administration. For the analgesia component research in the monitoring devices is
currently active, and the search for a tool to adequately translate the patient's state is
on. Nevertheless, as referred before, variables associated to the Noc/ANoc balance have
been used to assess the patients state, and even control the analgesic administration
[21, 127, 224, 228]. However much work needs to be done: rstly in the validation of
such a tool, and secondly in the search of the dynamic relations between drug dose and
eect, as well as the dynamical relations between the general anaesthesia components.
In Chapter 1 a controller structure to control the general anaesthesia triad is proposed in Figure 1.18, taking into consideration the possible direct or indirect interferences
between components. Although conceptually this approach is appealing, there is much
work to be done in the description of the system dynamics, and of the relations between
the components and of possible monitoring interferences.
The Noc/ANoc balance control is conditioned by events of noxious stimulation that
the anaesthesiologist usually anticipates by adjusting drugs' doses, minimizing respon-
200
201
multiple-input and multiple-output (MIMO) approach for the general anaesthesia triad.
Figure 6.3: Schematic representation of an automatic control system for the administration of general anaesthetics, using a
6. ON NOCICEPTION CONTROL
ses associated with the introduction of the stimulus.
stimulation phases may also be taken into account, using some predened stimulus intensities for the usually expected surgical and anaesthesia related noxious stimuli (see
Table 3.2). This information may be manually introduced into the system, much like
what the anaesthesiologist does in current practice, altering the conditions and guaranteeing anticipated response; otherwise the analgesic adjustment will only be done
a posteriori, in response to the noxious activation, which is in many cases undesirable
and may lead to arousal episodes (intense noxious stimulation).
the MIMO controller approach, taking into account the dynamic relations and interference between components. This controller administers drugs and maintains patient
homeostasis, considering the general anaesthesia triad, and possible inputs provided by
the anaesthesiologist (e.g.
stimulation intensity).
To develop a fully automated system for general anaesthesia administration and
control, the rst task is to understand and model the relations from drug dose to nal
eect, taking into consideration possible drugs' interactions. Pharmacokinetic models
are currently available for most of the currently used muscle relaxants, hypnotics and
analgesics. Such models describe the relation between administered drug dose and drug
concentration on the organ of eect. However, dynamic models describing the relation
between the eect-site concentrations and the expected monitored value (e.g. Hill equation), were addressed in only few studies. Taking into account possible drugs' synergy
(bivariate Hill equation), what lacks is the in depth study of the analgesia component,
translating the possible dynamical relations and interferences in the monitored variables (see Figures 1.13 and 1.18). Once the analgesia component may be translated by
a good enough model of the population, the controller may be designed inside a MIMO
framework.
202
available for the anaesthesiologist to asses the patient's hypnosis state. These monitors
use the EEG and derived signals, to transform a signal that is dicult to analyze, into a
simple index varying from 0 (isoelectric EEG) to 100 (awake patient) bringing a useful
tool in preventing over-dosage, that might be associated with long-term consequences,
and in preventing traumatic awareness episodes [13].
The hypnosis index used in this study was the SE, from the Entropy Module (Datex-
R
) [18], described in Chapter 1.
Ohmeda
Several models have been used to predict the drugs' eect on the human body
[39, 50, 51, 229], but because of the known interindividual variability, the use of average
models as predictors of the drug eect has limited performance.
use of sliding-mode control techniques seems suitable for the control of the drug administration, maintaining robustness to patient deviations from the average population
model.
The objective of this study was to apply sliding-mode control techniques to the model
structure of the hypnotic drug in the patient system, and consequent repercussion in the
hypnosis index used [128, 230], using a
203
6. ON NOCICEPTION CONTROL
6.3.1.1 Pharmacokinetic and Dynamic Modelling
Over the years studies have been carried out to describe the drugs' course and metabolism in the human system [39, 4951].
subsequent plasma collection and eect analysis, were made to adjust and obtain the
best model structure and parameters, trying to predict the plasma and eect-site concentrations. One of the key studies was performed for the hypnotic drug propofol, by
Schnider and co-workers [50].
ture (see Figure 1.13) to describe drug plasma concentrations (pharmacokinetic phase).
This study presented the relations between individual patient demographics (gender,
age, height and weight) and the ow parameters of the model, as shown in Table 1.8.
The average and the standard-deviation of the estimated parameters
i (i = 1, ..., 11)
The
ke0
parameter was
obtained from the relation between the point where the maximum eect is achieved,
corresponding to the equilibrium between plasma and eect-site concentrations.
The
ke0
k12 k13 k10 k21
k31
0
m 1 (t)
m 2 (t)
k12
k21
0
0
m 3 (t) =
k13
0
k31
0
ke0
0
0
ke0
Ce(t)
V1
m1 (t)
1
m2 (t) 0
m3 (t) + 0
Ce(t)
0
r(t)
(6.1)
204
State Entropy
Average Standard-Deviation
Hill Equation
E0
EC50 (g/ml)
86.89
1.5
1.9
1.5
4.5
0.48
0.37
E(t) = E0 1
where
r(t)
Ce(t)
Ce(t) + EC50
(6.2)
y,
appears;
appears.
x = f (x) + G(x)u
205
(6.3)
6. ON NOCICEPTION CONTROL
y = h(x)
with
x, u, f , G
and
h(x)
(6.4)
m1 (t)
m2 (t)
x=
m3 (t) , u = r(t)
Ce(t)
(6.5)
k12
k21
0
0
x
f (x) =
k13
0
k31
0
ke0
0
0
ke0
V1
(6.6)
1
0
G(x) =
0
0
h(x) = E0 1
(6.7)
x(4)
x(4) + EC50
La b
of
(6.8)
with respect to
by Equation
6.9.
La b =
b
a
x
(6.9)
The system under study has one input, one output, four state variables, and relative
order
r = 2; after the system feedback linearization comes the relation given by Equation
6.10.
(6.10)
and
(6.11)
G (x) = LG (Lf h) =
x
206
h
f
x
G
(6.12)
f (x) =
L2f h
=
x
h
f
x
f
(6.13)
This method may be applied for dierent non-linear control problems, nevertheless
there are some important limitations: it can not be employed to all non-linear systems;
the state must be measured; and robustness is not guaranteed in the presence of parameter uncertainties or dynamics not incorporated in the model. For the last issue,
robust control techniques and adaptive control techniques have been introduced, such
as sliding control, which will be presented in the following section.
In the case
of the drug pathway in the human system, this uncertainty is evidenced by the known
interindividual variability observed in response to drug administration, conditioned by
patients' demographics (age, weight, height and gender), but also by the clinical background, and associated diseases.
Imprecisions found during the modelling process may be classied in two distinct
groups:
The approach methodology of sliding-mode control comes from the perception that
th
order systems.
A notation
simplication is introduced, in such a way the n-order systems may be in fact replaced
by rst order systems, demonstrating that in principle it is possible to obtain optimum
performance. Nevertheless, such performance carries high costs when it comes to control
action, being necessary to nd a compromise between performance and control action
energy.
Sliding-mode control techniques were applied in this simulation study to the hypnosis
component of general anaesthesia assessed by SE. The model parameters were obtained
207
6. ON NOCICEPTION CONTROL
from clinical studies, and the uncertainties described earlier introduced in the controller
design [128, 230].
The problem of sliding-mode control consists in the construction of an attractive
surface, guaranteeing that all trajectories outside the surface point towards it. Dening
the sliding surface by Equation 6.14.
s=
with
d
+
dt
r1
e
(6.14)
e dened by Equation 6.15, a positive scalar specifying the bandwidth, and r=2.
e = y yd
with
yd
(6.15)
s = y (1) yr(1)
where
yr
is computed from
and
(6.16)
yd .
uncertainties given by Equations 6.17 and 6.18, associated to the uncertainties of the
model, where
and
and
f ,
respectively.
On the input
in an additive form.
(x), || D
G (x) = (1 + )G
(6.17)
f (x) f (x)
(6.18)
In the presence of parametric uncertainty, the system control law is given by Equation 6.19.
1 h
i
(x)
u= G
yr(2) f (x) k sgn(s)
with
G (x)
and
(6.19)
208
1, s < 0
0,
s=0
sgn(s) =
1,
s>0
(6.20)
(2)
(1 D)k = + D yr f (x) +
(6.21)
.
In order to avoid excessive control action, the sign function may be substituted in
the control law by the saturation function given by Equation 6.22, where
boundary layer width around the surface
s;
denes the
was
dened as 5.
1, s
s/, < s <
sat(s/) =
1,
s
(6.22)
To obtain the complete control law some of the constants need to be determined.
|s(t = 0)| / ,
with
6= 0)
(input
Initi-
(r 1)/.
awake at the beginning of the infusion (y=86.89), and knowing that for a slow induction
of 200 ml/h loss of consciousness occurs around 2 minutes following infusion start (120s
to reach y=50), we have
Regarding the
= 0.3.
mentally, nevertheless there are values that limit its value: in mechanical systems it
must be lower than the resonance frequency of the system; it depends of possible neglected delays in time
rate
vS
TA
1
3TA
(6.23)
1
S vS
5
(6.24)
vS = 1/5Hz , = 1/30Hz . TA
TA = 10s,
209
6. ON NOCICEPTION CONTROL
monitoring device output (corrupted signal with extended EEG window of analysis),
and regarding the fact that the
ke0 ,
6.3.4 Methods
Figure 6.4: System general structure and feedback for the sliding-mode controller.
yd (t) =
86.89, t < 50
50,
t 50
(6.25)
The controller activity was initialized by the usual infusion rate of propofol used
for slow induction of anesthesia (200ml/h or 556 g/s), and restricted to the interval
[0-556] g/s, in consistency with the system characteristics and limitations.
To test the controller robustness, the designed controller was applied to three theoretical patients, considering a nominal patient, male, 170 cm, 70 kg and 20 years old:
Nominal Patient:
Minimum Patient: Theoretical patient with all the minimum model parameters
expected, with the parameter deviations considered in the controller design.
210
Maximum Patient: The same as before, but with the maximum model parameters
expected.
These patients were dened in this way in order to take the controller to the extremes of the modelled uncertainties, and to verify if it would be robust enough, while
guaranteeing the achievement of the reference target (for more information refer to
Appendix E).
6.3.5 Results
Results of the application of the designed controller are divided according to simulated
patient, and reference target.
Figure 6.5: Control activity for the nominal patient using the sign function: on top SE
and SE reference, below the error sequence, and on bottom the infusion rate (ml/h).
Because of the excessive control action, the sign function was replaced by the saturation function in the control law. The control action is smoothed, which is preferable
for the clinical application. Figure 6.5 presents an example of the control activity for
the nominal patient and a sinusoidal reference.
211
6. ON NOCICEPTION CONTROL
Figure 6.6: Control activity for the nominal patient using the saturation function: on
top SE and SE reference, below the error, and on the bottom the infusion rate (ml/h).
that for the recovery phases (ascending reference), the SE is not able to follow the SE
reference, due to the patient modelled excretion (see Figure 6.6b).
Figure 6.7: Control activity for the minimum patient using the saturation function: on
top SE and SE reference, below the error, and on the bottom the infusion rate (ml/h).
The control action for the minimum patient had a satisfactory behaviour with small
error and small overshoot. The overshoot may have been caused by the patient deviation
212
Figure 6.8: Control activity for the maximum patient using the saturation function: on
top SE and SE reference, below the error, and on the bottom the infusion rate (ml/h).
Figure 6.8a shows the controller action and performance for the maximum patient.
The reference is achieved with a very small error and with no overshoot.
Figure 6.8b represents the response to a sinusoidal reference with small error. Unlike
the nominal and minimum patients, the maximum patient has a high excretion rate,
which allows a good performance following the reference ascending phases.
6.3.6 Discussion
A robust controller was designed for the administration of the hypnotic propofol, using
the SE as feedback variable, taking into consideration the inter-patient variability observed in the population during the modelling process.
213
6. ON NOCICEPTION CONTROL
The excessive control action was smoothed by the saturation function in exchange
from perfect performance, which may in the future allow the implementation in an
automatic syringe pump for hypnotic administration. The control technique performed
satisfactorily showing robustness to deviations from the nominal patient, which is the
most attractive characteristic due to the known interindividual variability.
Although
with the minimum patient a small overshoot occurred when following the step reference,
this overshoot was insignicant when trying to achieve a target range of 40 to 60.
As for the sinusoidal reference response, and the error associated with the ascending
phase of the sinusoid, this exercise was done only to demonstrate the robustness of the
method, showing that depending on the patient's characteristics a longer time may be
required for recovery, independently of the control action. Indeed, the control action
was able to follow the reference SE very well in the descending phases, which is the
most interesting feature from the clinical application point of view.
Figure 6.9: Representation of the patient modelled system, with possible error inputs,
controller, and state observer, for an on-line implementation of the automatic control
of the hypnotic.
Results showed that sliding-mode control presents a ne choice to control the hypnosis component state, possibly resulting in a more secure and stable anaesthesia administration. It was evident that although the controller presented a good performance,
even in the considered extreme patients, other drugs and variables should be introduced, taking into consideration the interactions previously described (e.g.
interaction
Exter-
nal interferences like surgical stimulus may also be introduced in the model, in order
214
Figure 6.10:
between the administered analgesic drug dose and the measurable eect, when a tool
for this eect is fully validated in the translation of the nociception/anti-nociception
balance.
Although the controller was designed for the administration of the hypnotic propofol, this was done to demonstrate the technique applicability in anaesthesia.
The
presented technique may also be implemented for the control of the administration of
the analgesic remifentanil, taking into consideration the observed interindividual variability in the PKPD model adjustment (see Table 1.7) [38, 39], and of the Noc/ANoc
balance, as soon as a monitor is fully validated to translate the patient's state concerning this general anaesthesia component, and proven to be related to changes both
in stimulation intensity, and analgesic drug dose. Only when such a monitor becomes
validated to be capable of translating patient's Noc/ANoc balance, it will be possible to
construct the dynamic model from the plasma concentration to the measurable eect,
taking into account population deviations from the average patient model. Although
for remifentanil the translation from Cp to Ce is described by Minto et al [38, 39], the
ke0
parameter was obtained using an EEG derived measure to translate the opioid ma-
215
6. ON NOCICEPTION CONTROL
ximum eect. This may well not be the best measure to translate the opioid maximum
eect in the Noc/ANoc balance. Actually, recent research has deviated from the EEG
for the assessment of the Noc/ANoc balance. Moreover, dierent measures or monitors
incorporate dierent delays in the patient's state assessment, which should be taken
into account in the modelling phase. Figure 6.10 presents the steps necessary to obtain
a model describing the relations between administered drug and measurable eect for
the analgesia component.
Because this simulation study was only performed to evaluate the robustness of
sliding-mode control techniques in anaesthesia, the controller was not further developed,
nevertheless it should be highlighted that in order for the technique to be applied in the
administration of anaesthetic drugs an observer of the patients state must be designed.
The two
216
6.5 Summary
clinical expertise to dene a set of rules, that would reproduce the anaesthesiologist's
action, and aid the clinician in the fast identication of the patient state, advising on
the action to take. Such tool would be in the context of an advisory system, as described
earlier in this chapter.
Besides
the dierences in the genesis of the proposed methods, passive or active, the last differs from the others because in this case it is possible to dene the baseline state of
operation, and measure the response to a known intensity stimulus.
6.5 Summary
The automatic control of general anaesthetics, have been under the anaesthesiology
scientic community attention since the early 1980's, with proposed strategies for the
hypnotic and muscle relaxant administration.
The application of automatic control to general anaesthesia must not be seen as
an attempt to substitute the anaesthesiologist, but rather as an assistant tool to the
clinician, which may allow him/her to direct his/her attention to other demanding tasks
in patient care, as well as an improvement in patient care.
The future of automatic control in anaesthesia lies on a MIMO approach, considering
the relations between drug's eects, and the possible monitoring interferences associated
with the patient's state. The approach is to regard the patient's state as a whole, while
217
6. ON NOCICEPTION CONTROL
taking into account each component individually and possible outer interferences such
as blood loss.
From the present to the development of a fully automatic control of the general
anaesthesia triad, more studies on the subject are required.
population trend and the assessment of the impact of possible synergistic eects, and
monitoring interferences, need to be developed.
A robust controller was developed for the hypnosis component.
Control of the
This
system may be regarded as an advisory and alarm system, alerting the anaesthesiologist
to a change in the patients' state, improving detection of arousal or discomfort of the
patient, anticipating a possible unwanted scenario, and advising a drug change to lead
the patient back to homeostasis.
218
Chapter 7
Conclusions and Future Work
Anaesthesia has three main components: hypnosis, analgesia and paralysis (immobility).
From these three components, analgesia is the only one for which an objective assessment
is not available.
Patients were
subjected to dierent noxious stimuli (anaesthesia related and surgery related) as well
as to stepwise changes in analgesic drug concentrations (remifentanil).
In a second study, a survey was designed to obtain a tool to be used in the analysis
of the collected data in the clinical study with patients. This survey was conducted with
219
relationship between the measured physiological variables, the intensity of the noxious
stimuli, and the eect of the anaesthetics propofol and remifentanil, a new multi-variable
method was developed. This method diers from the existing, because it allows an objective assessment of the patient's global state and of each variable taken individually.
The new method consists of a homeostasis index: if the drugs are in steady-state, a
change in the homeostasis index reects a change in the nociception/anti-nociception
balance. This method was applied to data collected in the clinical study for 31 patients,
to extract the periods of simultaneous steady-state for both the input and the output.
This allowed a distinction between situations when the patient was in steady-state or in
a dynamic condition, leading to a new way of assessing the nociception/anti-nociception
balance. With this method a new index of anaesthesia homeostasis was developed.
The next step in the work done in this thesis was the development of a model of the
perceived stimulus during anaesthesia and surgery, dened as any event leading to a
change in the patient. This model related the input drugs and stimulus intensity to the
observed responses in the physiological variables associated to noxious activation. It allows for an estimation of the patients perceived stimulus (nociception/anti-nociception
balance), considering the physiological variables connected to noxious activation as inputs.
The two methodologies responded adequately to the introduction of precise noxious
stimuli, and dierent analgesic levels.
220
mulus estimator provide complementary information considering the nociception/antinociception balance of the patient, and were implemented in a system for on-line assessment.
The objective assessment of nociception using evoked potentials was addressed in
this thesis and a new index was developed. A clinical study in healthy volunteers measured the same physiological variables of the study in patients, and also somatosensory
evoked potentials (SEPs), while volunteers were subjected to painful stimuli before and
during dierent combinations of anaesthetic drugs propofol and remifentanil. It was observed that wave amplitude decreases and latency increases with increasing drug doses,
accompanied by a decrease in reported pain. A new index, based on wave amplitude and
latency values normalized for the individual painful threshold, was proposed as a new
objective active method of nociception/anti-nociception balance, producing normalized
measures, comparable between subjects.
The nal work in the thesis was the development of a new controller for the administration of anaesthetic drugs.
increased the interest in the scientic community for the automatic control of each component of general anaesthesia. Regarding to the analgesia component, some attempts
have been made for the resolution of this problem, nevertheless this is dependent on
the existence of a fully validated monitor for the analgesia component. Robust control
techniques should be applied in the future, taking into account both the known inter
and intra-individual variability observed in the population. It was demonstrated that
robust control techniques may be implemented in the control of general anaesthesia
components, by comparison to the parallel problem of the hypnosis component control,
incorporating the known patient model deviations observed in the population.
The search for objective methods for nociception assessment is still an open research
eld. The development of a tool capable of translating the nociception/anti-nociception
balance during anaesthesia is of great importance in order to provide the patient with
the best care, adequately titrating the dierent anaesthetic drugs. Hypnosis monitors
are nowadays validated and available for clinical use, aiding in the administration of
the hypnotic, preventing awareness and over-dosing and possible adverse eects.
If
one imagines that the patient may be conscious during the surgical procedure, the
scenario may become much worsened if the patient is also perceiving pain.
In this
221
The methods developed in this thesis aim at answering some of the open
developed may aid in a more adequate control of anaesthesia, leading to better patient
care, better recovery and possibly better long term outcomes (e.g.
of post-operative chronic pain).
lower incidence
thesis adequately responded to dierent stimulus intensities and drugs' eect, necessary
premises to be considered as nociception/anti-nociception balance indicators.
Also a
software was developed with the passive nociception measures, providing information
on the patients homeostasis and location regarding the nociception/anti-nociception
balance, which may aid anaesthesiologists in a faster detection and better control of
nociceptive responses.
222
223
224
List of Publications
The research done, during the course of this doctoral program, resulted in a number of
scientic communications and publications, listed below.
Awards
The developed work was recognized in the scientic community, and received two distinctions:
Publications
Ana Castro, Pedro Amorim, Catarina S. Nunes, Modelling State Entropy of the
EEG and Auditory Evoked Potentials - Hypnotic and Analgesic Interactions, 29th
Annual International Conference of the IEEE Engineering in Medicine and Biology
Society (EMBS), New York, NY, USA, IEEE. (ISBN: 978-1-4244-0787-3), 2007
225
. LIST OF PUBLICATIONS
national Conference of the IEEE Engineering in Medicine and Biology Society
(EMBS), New York, NY, USA, IEEE. (ISBN: 978-1-4244-1814-5), 2008
Diana Afonso, Ana Castro, Eduarda Amadeu, Pedro Amorim, Avaliao objectiva
da dor: condutncia elctrica cutnea, Dor 2010; 17: 5-8
Ana Castro, Pedro Amorim, Catarina S. Nunes, Fernando G. Almeida, Development of a scale to assess noxious stimuli intensity during anesthesia and surgery
using Rasch analysis, European Journal of Pain
Ana Castro, Pedro Amorim, Catarina S. Nunes, Fernando G. Almeida, Somatosensory evoked potentials may objectively translate nociception/anti-nociception
balance with anesthetics, Journal of Clinical Monitoring and Computing
Communications
Abstracts of scientic communications presented in international conferences are listed
here. Most were published in supplements of the Anesthesiology, European Journal of
Anaesthesiology, and the Journal of Neurosurgical Anesthesiology.
226
Ana Castro, Eduarda Amadeu, Ftima Martins, Paula S Couto, Pedro Amorim, Catarina S. Nunes, Propofol and Remifentanil Drug Interaction on the State
Entropy od the EEG: Surface Model, Total Intravenous Anesthesia - TCI World
Congress, Oral Presentations, p. 98, 2007
Ana Castro, Eduarda Amadeu, Paula S Couto, Pedro Amorim, Catarina S. Nunes, Bispectral Index and State Entropy Simulation with Propofol and Remifentanil Eect-Site Steering, Journal of Neurosurgical Anesthesiology, 20(4):
340,
Ana Castro, Eduarda Amadeu, Paula S Couto, Pedro Amorim, Catarina S. Nunes, Educational Software for Bispectral Index and State Entropy Simulation,
Anesthesiology A686, 2008
227
. LIST OF PUBLICATIONS
Diana Afonso, Ana Castro, Maria Eduarda Amadeu, Pedro Amorim, Avaliao
objectiva da dor: Conductncia Elctrica Cutnea, 8 Convnio da ASTOR, 2010
Ana Castro, Pedro Amorim, Catarina S. Nunes, Fernando Gomes de Almeida, Development of a scale to assess noxious stimuli intensity during anesthesia and surgery using Rasch analysis, Journal of Neurosurgical Anesthesiology and Anesthesiology 2011 (accepted for publication)
228
Appendix A
Remifentanil Recommendations in
Portugal
Remifentanil is an approved drug by INFARMED (Autoridade Nacional do Medicamento e Produtos de Sade, I. P. - National Authority of Medicines and Health Products) for clinical use in Portugal. The following text presents a summary of the guidelines of remifentanil use in Portugal, with the recommended dosages and potential side
eects [34, 35, 37].
Remifentanil is indicated as an analgesic for induction and maintenance of general
anaesthesia, and also as an analgesic for patients in the Intensive Care Unit, under
mechanical ventilation.
Because this is a very potent drug it must be only administered according to the
recommendations, in facilities equipped with cardiovascular and respiratory functions
support and monitoring, and by trained personnel in the use of anesthetic drugs, and
cardio-respiratory resuscitation maneuvers.
Remifentanil administration may be with a continuous perfusion of remifentanil,
with the infusion velocities according to the patients' body weight and the procedure.
The administered doses should be titrated according to the patients' response. Remi-
229
Table A.1: Recommended doses of remifentanil in co-administration with other anaesthetic agents for adult patients [35, 37].
Indication
Bolus Injection
(g/kg)
Anaesthesia Induction
1 (administered during a
Initial Rate
Limits
0.5-1
period of 30 s or more)
Maintenance of anaesthesia in ventilated
patients
N2O (66%)
0.5 to 1
0,4
0.1 to 2
0.5 to 1
0,25
0.2 0.05 to 2
0.5 to 1
0,25
0.3 0.05 to 2
0.5 MAC)
Propofol (initial dose
100 g/kg/min)
Tables A.1, A.2 and A.3 show the recommended doses of remifentanil for a coadministration with other anaesthetic agents in adult patients, the infusion velocity and
the equivalent plasmatic concentrations by approximations of several stabilized manually perfusion of remifentanil when using target controlled infusion. When administered
as bolus, it should be administered in a minimum period of 30 seconds.
In the recommended doses, remifentanil decreases signicantly the hypnotic agent
need in order to maintain adequate anaesthesia. Knowing this the concomitant administration of hypnotic drugs and remifentanil must be balanced.
230
Table A.2: Recommended infusion rates of remifentanil according to body weight [35,
37].
Infusion Rate
(g/kg/min)
10
20
30
40
50
60
0.0125
0.188
0.375
0.75
1.125
1.5
1.875
2.25
0.025
0.375
0.75
1.5
2.25
3.75
4.5
0.05
0.75
1.5
4.5
7.5
0.075
1.125
2.25
4.5
6.75
11.25
13.5
0.1
1.5
12
15
18
0.15
2.25
4.5
13.5
18
22.5
27
0.2
12
18
24
30
36
0.25
3.75
7.5
15
22.5
30
37.5
45
0.3
4.5
18
27
36
45
54
0.35
5.25
10.5
21
31.5
42
52.5
63
0.4
12
24
36
48
60
72
for simultaneous administration of propofol or isourane. Due to the rapid start and
short duration of the remifentanil action, the administration should be titrated in increases of 25-100%, or decreases of 25-50% of the actual dose, every 2-5 minutes, in oder
to obtain the desired level of -opioid response.
Remifentanil decreases the need of other medications such as benzodiazepines, inhaled agents or hypnotic drugs. Because of the short duration of action, the post-operative
analgesia should be prepared in time with a long duration analgesic. Remifentanil is
not recommended for the post-operative analgesia in spontaneous breathing patients,
only for the immediate post-operative period until the long duration analgesia reaches
its maximum eect.
The bolus injection of remifentanil, due to its properties and side eects, is not
recommended in patients in spontaneous breathing.
231
Infusion Rate
Plasmatic Concentration
(g/kg/min)
of Remifentanil (ng/ml)
0.05
1.3
0.1
2.6
0.25
6.3
0.4
10.4
0.5
12.6
25.2
50.5
muscle rigidity.
Like in other
opioids this eect is related with the dose and administration velocity. Knowing this a
bolus should be administered for a period of 30 seconds at least. Muscle rigidity observed
should be treated according to the clinical state of the patient, if it occurs during the
induction of general anaesthesia, this condition should be treated using muscle relaxants
232
sion,
respiratory depres-
what demands the use of remifentanil only in places with the right equipment
to monitor and treat this condition. Special care should be taken with patients with
respiratory insuciency. Signs of respiratory depression may be abolished with a temporary suspension of the drug administration or with a decrease of 50% of the infusion
rate. Remifentanil did not display recurrent respiratory depression, however many factors may aect the post-operative recovery, and it is important to make sure that the
patient has fully recovered and is in spontaneous breathing before taking the patient to
the recovery room.
The
hypotension and bradycardia that rarely may evolve to assystole and cardiac arrest,
may be reduced by decreasing the remifentanil infusion velocity, decrease the dose of
the concomitant anesthetics or by administrating intravenous uids, vasopressure agents
and anticholergenics, according to the situation.
older patients may be more sensitive to the cardiovascular side eects of remifentanil.
As in all opioids, remifentanil may cause
dependency.
Residual drug in the death space of the intravenous catheter may be sucient to
cause respiratory depression or apnea, and/or muscle rigidity, if the same catheter is
used to administer other drugs or solutions.
rapid ux catheter or an exclusive catheter for the administration of remifentanil that
is immediately removed after the discontinuation of the drug administration.
Like other opioids, remifentanil decreases the need of inhaled or intravenous anesthetics, and benzodiazepines to maintain general anaesthesia. If the central nervous system
depressants doses are not reduced there is an increased risk of incidence of adverse side
eects associated with these drugs. The adverse eect of remifentanil may also be exacerbated by the concomitant use of cardiac depressor drugs such as beta blockers and
233
ded when the potential benet is greater than the potential malecent. It is unknown if
remifentanil is excreted in the mother milk, however substances related to remifentanil
were found in female rat milk experiments, and the mother should be advised t not to
breast feed in the next 24 hours after the remifentanil administration. During labour,
remifentanil crosses the placenta barrier, and may cause respiratory depression in the
child.
The
more common adverse eects of remifentanil derive directly from the phar-
macology of the -opioids agonists, these adverse eects disappear minutes after the
interruption or decrease in the administration rate of the drug. The following list presents the adverse eects incidence, classied as very frequent (>1/10), frequent (>1/100
and
<10/10), little frequent (>1/1000 and <1/100), rare (>1/10000 and <1/1000) and
Cardiopathy Frequent:
Gastrointestinal diseases
Very frequent:
Little frequent:
obstipation.
itching.
234
remifentanil, the side eects are limited to the immediate period after administration.
The drug interruption leads to the return to the initial state in 10 minutes. In the case
overdosage occurs, the drug administration should be interrupted, the airway should
be accessible for possible need of assisted or controlled ventilation, and to keep the
adequate cardiovascular function. In the case respiratory depression is associated with
muscle rigidity, the administration of neuromuscular blockers may be needed to facilitate
patients ventilation. Other uid and vasopressors may be administered intravenously
to treat hypotension.
The intravenous administration of an opioid antagonist like naloxone may be given
as a specic antidote to control serious respiratory depression and muscle rigidity. The
respiratory depression duration after the overdosage is unlikely to exceed the opioid
antagonist action.
-opioid
-opioid
235
potency. In man, the research indicates that all pharmacological activity is associated
with the original compound. The activity of this metabolite has no clinical consequence.
The metabolite half-life in adult healthy patients is of about 2 hours. In patients with
normal renal function the time for 95% of renal elimination of the primary metabolite
of remifentanil is of about 7 to 10 hours. Remifentanil is not a substrate for plasmatic
cholinesterases.
The rapid reversion of the post-analgesia action of remifentanil is not aected by
the state of the
renal function.
hepatic in-
suciency waiting for hepatic transplant, or during the anhepatic phase of the surgery
of leaver transplant. Patients with hepatic insuciency may be more sensitive to the
eect of respiratory depression of remifentanil.
In
pediatric patients
phase are increased in younger children, reaching the values for adult healthy patients
at the age of 17. The pharmacokinetics of the metabolite acid carboxylic in pediatric
patients age 2 to 17 years, is similar to the observed in adults, after correction of the
body weight dierences.
236
elderly patients the remifentanil clearance is slightly reduced (>65 years), when
electroencephalogram that is 50% of the tonger patients dose, therefore the initial dose
of remifentanil should be reduced in 50% in elderly patients and then carefully adjusted
according to the patient needs.
237
238
Appendix B
Informed Consent Form
239
Estudos de investigao
Consentimento Informado
CONSENTIMENTO INFORMADO
Data [Ano/ms/dia].
Data [Ano/ms/dia].
DEFI Modelo de Consentimento Informado para Estudos de Investigao Edio 1 / Verso 1 Aprovado em CA em 15-03-2007
Estudos de investigao
Consentimento Informado
Estudos de investigao
Consentimento Informado
O estudo ter uma primeira fase antes do incio da cirurgia, mas j depois de o doente estar
anestesiado. Nessa primeira fase usaremos um procedimento habitual que usamos com frequncia
no dia-a-dia e que consiste na aplicao de um estmulo elctrico, cutneo ao nvel do pulso
(estmulo tetnico), analisando a resposta dos sinais clnicos atrs referidos. Se houver alterao de
algum dos sinais, a dose do medicamento analgsico ser aumentada; no havendo alterao, d-se
incio cirurgia.
Neste estudo, as nicas diferenas na anestesia em relao ao que seria a anestesia, caso
no fosse aplicado o protocolo do estudo, so as j descritas variaes na dose do medicamento
analgsico e a escolha da dose inicial de analgsico. Para esta dose inicial os doentes sero
distribuidos ao acaso para receber uma de trs doses pr-definidas. Qualquer uma dessas doses
est dentro das doses clnicas recomendadas, sendo que os doentes que recebem a dose mais
baixa, naturalmente recebero uma dose mais elevada do medicamento hipntico.
Aps a cirurgia, o doente ser monitorizado de acordo com os procedimentos habituais,
sendo registados os consumos de medicamento analgsico, bem como a avaliao do doente numa
escala de dor, durante o perodo em que se encontrar no recobro.
DEFI Modelo de Consentimento Informado para Estudos de Investigao Edio 1 / Verso 1 Aprovado em CA em 15-03-2007
Appendix C
Remifentanil Recommended Doses
site concentration targets with correspondent bolus dose and infusion rates were evaluated, considering a male patient, 70 kg, 170 cm, 60 years old. Recommended remifentanil
doses in co-administration with propofol are:
35-70 g;
3.5-140 g/min.
Simulating the induction targets for each patient using TCI, the following bolus are
administered, in consonance with the recommended doses:
2.0 ng/ml
bolus of 34 g;
3.0 ng/ml
bolus of 51 g;
243
4.0 ng/ml
bolus of 68 g.
Simulating the maintenance doses of remifentanil administered using TCI, one must
consider the minimum 1.5ng/ml eect-site concentration allowed in the study protocol,
for the second phase of the study. Considering the simulated patient this target is translated into a continuous administration of remifentanil at an infusion rate of 3.84 g/ml,
and therefore within the recommended range.
244
Appendix D
Informed Consent Form Volunteers' Study
245
Estudos de investigao
Consentimento Informado
CONSENTIMENTO INFORMADO
Data [Ano/ms/dia].
Data [Ano/ms/dia].
DEFI Modelo de Consentimento Informado para Estudos de Investigao Edio 1 / Verso 1 Aprovado em CA em 15-03-2007
Pgina 1 de 2
Estudos de investigao
Consentimento Informado
DEFI Modelo de Consentimento Informado para Estudos de Investigao Edio 1 / Verso 1 Aprovado em CA em 15-03-2007
Pgina 2 de 2
248
Appendix E
Simulations and Control
The PKPD models of Schnider[50] and Minto[38, 39], for propofol and remifentanil
respectively, were implemented in Simulink, Matlab
R
.
have a three compartments structure (Figure 1.13) and the block diagram is shown in
Figure E.1.
Each pharmacokinetic and dynamic model was obtained from a set of individuals,
and adjusted to build the average patient model[38, 39, 50]. Pharmacokinetics and
dynamics vary between patients according to age, weight, height and gender.
These
variables were introduced in the modelling process, and model parameters depend on
them either directly (age, height and weight), or through LBM (gender, height, weight).
To assess inter-patient variability considering the patient data, PKPD models were
simulated for dierent combinations of age, weight, height and gender as follows:
249
Figure E.1: Simulink block diagram for the three compartments pharmacokinetic and
dynamic model.
The
kij
ce
massi
= 1, ..., 3).
Figures E.2 and E.3 present the results of the simulations for dierent combinations
of age, weight, height and gender, with LBM limitation, for Schnider and Minto models
250
(a) Male
Figure E.2:
(b) Female
251
(a) Male
(b) Female
Figure E.3: Step response simulations for the Minto model, considering dierent combinations of age, weight, height and gender.
sed structure to control on-line the infusion rate based on the State Entropy monitored
values in the patient, producing an adequate infusion rate (ml/h) to maintain or lead
the patient to the reference target concerning the control variable, in this case the EEG
derived index translating the patient's hypnosis state.
Inter-patient variability observed is resultant both from dierences in the patient's
data, age, weight, height and gender, and due to variations from the population modelled
trend, seen in the variability reported in the modelled parameters. Much of the variability observed in settling times of the simulations is consequence of the non-symmetry
between ow rate constants. We also observe that compartments 2 and 3 may be seen as
perturbations of the central compartment (parallel dynamics), and translation between
Cp and Ce presents itself as a series dynamic.
252
253
controller, used in the simulations for the nominal, minimum and maximum patients.
R
, with the system representation and feedback for the sliding-mode
254
Representation of the diagram block for a possible on-line implementation, with the input incoming from the
actuator.
monitored values of State Entropy (SE) in the patient, and with the output infusion rate (ml/h) directed to the syringe pump
Figure E.5:
Table E.1: Considered variation intervals in the simulations, regarding the reported parameters and dispersion measures of the pharmacokinetic and dynamic model, reported
in [50].
Minimum
Maximum
V1
3,714
4,826
V2
14.240-0.251*(age-53)
23.560-0.531*(age-53)
168,2
307,8
[1.772+(weight-77)*8 +(LBM-
[2.008+(weight-77)*8 +(LBM-
If age>53:
If age>53:
[1.066-0.034*(age-53)]/4.826 Else:
[1.514-0.014*(age-53)]/3.714 Else:
[1.066-0.014*(age-53)]/4.826
[1.514-0.034*(age-53)]/3.714
k13
0.748/4.826
0.924/3.714
k21
If age>53:
If age>53:
[1.066-0.034*(age-53)]/V2max Else:
[1.514-0.014*(age-53)]/V2min Else:
[1.066-0.014*(age-53)]/V2max
[1.514-0.034*(age-53)]/V2min
0.748/307.8
0.924/168.2
V3
k10
k12
k31
255
256
References
Miller's Anesthesia.
(Cited on
page 1.)
[3] F. D. Moore.
Anesthesiology,
Annals of Surgery,
229(2):187196,
Engnieering.
Introduction to Biomedical
(Cited on pages 3
and 39.)
[6] L. C. Jameson and T. Sloan. Using eeg to monitor anesthesia drug eects during
surgery.
(Cited
Acta Anaesthesiologica
257
REFERENCES
[8] J. Bruhn, P. S. Myles, R. Sneyd, and M. M. R. F. Struys. Depth of anaesthesia
monitoring: what's available, what's validated and what's next?
British Journal
Journal of
Anesthesiology, 87(4):808815,
Lancet,
363(9423):17571763, 2004.
during anesthesia.
[15] USA Aspect Medical Systems. Monitoring for pain - new insight from the eeg,
2007.
258
REFERENCES
[16] S. D. Greenwald and C. Rosow. Bis and emg variability increase before somatic
responses during surgery.
Description of
Acta
[20] E. R. J. Seitsonen, I. K. J. Korhonen, M. J. van Gils, M. Huiku, J. M. P. Ltjnen, K. T. Kottila, and A. M. Yli-Hankala.
49:284292, 2005.
(Cited on
remifentanil based on the dierence between response entropy and state entropy.
98(6):785791, 2007.
259
REFERENCES
Denitions and Notes on Usage. Classication of Chronic Pain. IASP Task Force
on Taxonomy. IASP Press, 2nd edition, 1994.
[23] W.
F.
Gannong,
editor.
Lange
Medical
(Cited on pages 9,
Elsevier
(Cited on
their technology and use - a novel automated von frey device improved testing for
hyperalgesia.
(Cited on
page 16.)
[28] M. E. Wewers and N. K. Lowe. A critical review of visual analogue scales in the
measurement of clinical phenomena.
Anesthesiology, 34(1):
260
REFERENCES
in patients with chronic pain: The spouse response inventory.
The Journal of
Pain,
A unidimensional pain/disability
69(3):269278, 1997.
(Cited on
(Cited on pages 19
and 77.)
[33] F. Varoli and V. Pedrazzi. Adapted version of the mcgill pain questionnaire to
brazilian portuguese.
17(4):328335, 2006.
(Cited on
Goodman and
(Cited on pages 20, 21, 22, 23, 24, 188, 198 and 229.)
[35] INFARMED.
FARMED at 30-08-2006.
ca/drugs/DB00899/structure.
[37] INFARMED.
Approved by IN-
http://www.drugbank.
261
REFERENCES
codynamics of remifentanil. i. model development.
(Cited on pages 22, 25, 29, 30, 43, 51, 215 and 249.)
[39] C. F. Minto, T. W. Schnider, and S. L. Shafer. Pharmacokinetics and pharmacodynamics of remifentanil. ii. model application.
(Cited on pages 22, 25, 29, 41, 43, 49, 50, 51, 165, 167, 203, 204, 215 and 249.)
100(1):
Deeper level of
(Cited on
Deeper level of
103(3):633640, 2006.
(Cited on pages 22
and 113.)
ca/drugs/DB00818/structure.
http://www.drugbank.
(Cited on
262
REFERENCES
the 7th International Symposium Memory and Awareness in Anaesthesia, 100(6):
868P869P, 2008.
Anesthesiology,
80(1):104122,
The pharma-
Anesthesiology, 88:1170
(Cited on pages 24, 25, 29, 30, 31, 41, 51, 165, 167, 203, 204, 249
and 255.)
Academia Press, Belgium, 1st edition, 1991. (Cited on pages 24, 25, 203 and 204.)
263
REFERENCES
[53] C. F. Minto, T. W. Schnider, T. G. Short, K. M. Gregg, A. Gentilini, and S. L.
Shafer. Response surface model for anesthetic drug interactions.
92(6):9801002, 2000.
Anesthesiology,
European
and 41.)
Society, 2008. EMBS 2008. 30th Annual International Conference of the IEEE,
pages 5810 5813, aug. 2008.
doi:
10.1109/IEMBS.2008.4650535.
(Cited on
Presentation.
Oral
[57] A. Castro, P. Amorim, and C.S. Nunes. Modeling state entropy of the eeg and
auditory evoked potentials - hypnotic and analgesic interactions. In
Engineering
in Medicine and Biology Society, 2007. EMBS 2007. 29th Annual International
Conference of the IEEE, pages 1949 1952, aug. 2007.
4352699.
doi: 10.1109/IEMBS.2007.
[58] P. M. Schumacher, T. W. Bouillon, D. Leibundgut, V. Hartwich, and M. Luginbhl. Time-based online display of a noxious stimulus response index on pharamacological data.
Anesthesiology,
page 28.)
264
ASA Meeting.
(Cited on
REFERENCES
[59] T. W. Bouillon, P. M. Schumacher, D. Leibundgut, S. L. Shafer, and A. M. Zbinden. A novel mechanistic model based on the mac reduction paradigm describes
hypnotic-opioid interaction for suppression of response to stimulation.
ASA Meeting.
Anesthesi-
A novel
Anesthesiology,
112
Anesthesiology,
page
Quantifying uncer-
the IEEE Engineering in Medicine and Biology Society (EMBS), pages 786789,
2004.
Anesthesiology,
89(3):557559,
265
British
Somatosensory event-
REFERENCES
related potentials to painful and non-painful stimuli: eects of attention.
38(3):303312, 1989.
Pain,
22(3):503507, 1970.
The
(Cited on
Journal of Neurophysiology,
Neirozi-
Pain,
61(3):435439, 1995.
(Cited on
Anesthesia and
266
Sex-specic responses to
REFERENCES
opiates:
2008.
107(1):8395,
[75] M. Pereira, B. Rehberg-Klug, A. Castro, and P. Amorim. Inuence of diuse noxious inhibitory control, temporal summation and expectation on pain perception
in healthy volunteers. In
Veterinary
Novel multipara-
96(3):367376, 2006.
(Cited on
[78] M. Huiku, K. Uutela, M. van Gils, I. Korhonen, M. Kymlinen, P. Merilinen, M. Paloheimo, M. Rantanen, P. Takala, H. Vierti-Oja, and A. Yli-Hankala.
Assessment of surgical stress during general anaesthesia.
(Cited on pages 33, 34, 35, 68, 72, 77, 79, 88, 92,
267
REFERENCES
surgical stress in gynaecological laparoscopic day-case surgery.
British Journal of
ASA Meeting.
Anesthesiology,
(Cited on
page 35.)
103(4):586593, 2009.
(Cited on
268
104(5):587895, 2010.
REFERENCES
[87] M. Bloom, A. Jurmann, G. Cu, and A. Bekker. Bis variability reects analgesia.
72 and 92.)
Va-
riability comparison of the composite auditory evoked potential index and the
bispectral index during propofol-fentanyl anesthesia.
[89] I. Takamatsu, M. Ozaki, and T. Kazama. Entropy indices vs the bispectral indextm for estimating nociception during sevourane anaesthesia.
British Journal
Facial muscle
anctivity, response entropy, and state entropy indices during noxious stimuli in
propofol-nitrous oxide or propofol-nitrous oxide-remifentanil anaesthesia without
neuromuscular block.
(Ci-
(Cited
on page 37.)
Rocuronium dose-dependently
102(5):667672, 2009.
(Cited on
page 37.)
[93] M. Valjus, A. R. Jokela, and K. Kortila. Response entropytm is not more sensitive
than state entropytm in distinguishing the use of esmolol instead of remifentanil
269
REFERENCES
in patients undergoing gynaechological laparoscopy.
[94] G. Weil, S. Passot, F. Servin, and V. Billard. Does spectral entropy reect the response to intubation or incision during propofol-remifentanil anesthesia?
Anesthe-
[95] V. Bonhomme, V. Llabres, P. Y. Dewandre, J. F. Brichant, and P. Hans. Combined use of bispectral indextm and a-linetm autorregressive indextm to assess antinociceptive component of balanced anaesthesia during lumbar arthodesis.
British
[96] E. Weber, P. Gambus, M. Jospin, X. Borrat, and S. Cortes. Prediction of nociceptive response by auditory evoked potentials during ultrasonographic exploration.
ASA Meeting.
[98] A. Cividjian, J.Y. Martinez, E. Combourieu P. Precloux, A. M. Beraud, Y. Rochette, M. Cler, L. Bourdon, J. Escarment, and L. Quintin. Beat-by-beat cardiovascular index to predict unexpected intraoperative movement in anaesthetized
unparalyzed patients:
a retrospective analysis.
270
REFERENCES
ovascular index, cardean: A prospective randomized assessment of its utility for
the reduction of movement during colonoscopy.
765772, 2010.
Cardean, a
DelSys Inc.
[106] H. Storm. Skin conductance and the stress response from heel stick in preterm
infants.
(Cited on
page 39.)
[107] H. Storm. The development of a software program for analyzing skin conductance
changes in preterm infants.
(Cited
on page 39.)
[108] Med-Storm. Med-storm stress detector - user manual, 2008. Med-Storm Innovation AS Vers. 1.0.
271
REFERENCES
[109] H. Storm. Changes in skin conductance as a tool to monitor nociceptive simulation
and pain.
21(6):796804, 2008.
(Cited on
Pain,
97
[111] A. C. Gjerstad, H. Storm, R. Hagen, M. Huiku, E. Qvigstad, and J. Raeder. Comparison of skin conductance with entropy during intubation, tetanic stimulation
and emergence from general anaesthesia.
(1):815, 2007.
Anesthesiology, 113(1):175182,
(Cited on
91(3):347352, 2003.
(Cited on
Latency of
272
REFERENCES
pupillary reex dilation during general anesthesia.
97(2):725730, 2004.
ASA Meeting.
[118] F.-X. Donnette, E. Bourgeois, I. Murat, and I. Constant. The mac of sevourane
allowing inhibition of pupillary response to nociception in children.
page 1633, 2007. ASA Meeting.
Anesthesiology,
105(1):5562, 2009.
(Cited
[120] F. von Dincklage, M. Hackbarth, R. Mager, B. Rehberg, and J. H. Baars. Monitoring of the responsiveness to noxious stimuli during anaesthesia with propofol and
remifentanil by using riii reex threshold and bispectral index.
British Journal of
(Cited on
[122] N. Liu, T. Chazot, B. Trillat, G. A. Dumont, and M. Fischler. Closed-loop titration of propofol guided by the bispectral index.
273
REFERENCES
[123] K. Volyanskyy, W. M. Haddad, and J. M. Bailey.
tion control for compartmental systems with application to introperative anesthesia inuenced by hemorrhage and hemodilution eects.
International Journal of
23(1):129, 2009.
[124] T. J. Gilhuly, S. R. Hutchings, G. A. Dumont, and B. A. M. Struys. Development and pilot testing of the neuromuscular blockade advisory system.
Computer
(Cited on pages 42
and 194.)
2007.
(Cited on pages 42
and 194.)
[126] D. G. Mason, J. J. Ross, N. D. Edwards, D. A. Linkens, and C. S. Reilly. Selflearning fuzzy control of atracurium-induced neuromuscular block during surgery.
(Cited
274
REFERENCES
closed-loop control anesthetic drug infusion devices.
Anaesthesiologica Scandinavica,
53(8):975985, 2009.
Acta
and 72.)
[132] M. Malik. Heart rate variability - standards of measurement, physiological interpretation, and clinical use - task force of the european society of cardiology the
north american society of pacing electrophysiology.
1996.
Circulation, 93(5):10431065,
[133] J. M. Hyman.
4(4):645654, 1983.
SIAM
(Cited on
page 66.)
Extracta Mathematicae,
74(368):829836, 1979.
(Cited
on page 66.)
275
REFERENCES
for the clinician.
(Cited on
page 72.)
Acta Anaesthesiologica
[141] J. M. Linacre.
1994.
Annals
Da-
Expanded edition,
(Cited on
[143] K. Pesudovs and B. A. Noble. Improving subjective scaling of pain using rasch
analysis.
70
Proc.
276
REFERENCES
[146] G. Rasch. A mathematical theory of objectivity and its consequences for model
construction.
Branch, 1968.
Psychometrika,
[149] D. Andrich.
42
Psychometrika,
(Cited on
page 76.)
[151] D. Andrich.
dit scoring.
Psychometrika,
47(2):
1(4):363378, 1988.
(Cited on
page 76.)
Rasch Measurement
Rasch Meascurement
277
REFERENCES
enhance measurement precision.
(3):278288, 2005.
an
European Journal of
rehabilitation research.
Journal
[162] R. D. Hayes, H. Liu, K. Spritzer, and D. Cella. Item response theory analyses of
physical functioning items in the medical outcomes study.
3238, 2007.
278
Medical Care,
45(5):
REFERENCES
[163] J. A. Teresi, K. Ocepek-Welikson, M. Kleinman, K. F. Cook, P. K. Crane, L. E.
Gibbons, L. S. Morales, M. Orlando-Edelen, and D. Cella. Evaluating measurement equivalence using the item response theory log-likelihood ratio (irtlr) method
to assess dierential item functioning (dif ): applications (with illustrations) to
measures of physical functioning ability and general distress.
[164] L. S. Walker, J. E. Beck, J. Garber, and W. Lambert. Children's somatization inventory: psychometric properties of the revised form (csi-24).
Journal of Pediatric
[165] G. Velde, D. Beaton, S. Hogg-Johnston, E. Hurwitz, and A. Tennant. Rasch analysis provides new insights into the measurement properties of the neck disability
index.
104(6):
Anesthesiology,
100(6):13531372, 2004.
(Cited on pages 77, 78, 79, 88, 113, 128, 130, 131, 133 and 145.)
Anesthesiology,
279
80(2):261267, 1994.
(Cited on pa-
REFERENCES
[169] T. Kazama, K. Ikeda, and K. Morita. Reduction by fentanyl of the cp50 values of
propofol and hemodynamic responses to various noxious stimuli.
87(2):213227, 1997.
Anesthesiology,
(Cited on
[171] M. E. Ausems, C. C. Hug, D. R. Stanski, and A. G. L. Burm. Plasma concentrations of alfentanil required to supplement nitrous oxide anesthesia for general
surgery.
101(1):161
[174] A. Castro, F.G. Almeida, P. Amorim, and C.S. Nunes. A wavelet based method
for steady-state detection in anesthesia. In
Society, 2009. EMBC 2009. Annual International Conference of the IEEE, pages
954 957, sept. 2009. doi: 10.1109/IEMBS.2009.5334037.
and 113.)
280
Journal of
REFERENCES
[176] A. Castro, F. G. Almeida, E. Amadeu, P. Amorim, and C. S. Nunes. A wavelet
based algorithm for steady-state detection: application in bis guided anesthesia.
Anaesthesia steady-
state index: a combined measure of bis, blood pressure and heart rate.
European
[179] Z. Szab.
27
Journal of
Computer Methods
Pattern
Journal of Neu-
281
REFERENCES
frontal, insula and somatosensory cortex: a single-trial fmri study.
13261336, 2002.
Brain, 125(6):
ous eeg and fmri and their correlation with evoked pain.
(8):8, 2008.
BMC Anesthesiology, 17
Nature
British
and 188.)
British Journal
and 186.)
282
51(1):3542,
REFERENCES
[192] L. Arendt-Nielsen and P. Bjerring. Reaction times to painless and painful
argon laser stimulation.
1988.
co2
and
[196] M. A. Smith and J. D. Gray. Age-related dierences in sensitivity to the antinociceptive eects of opioids in male rats.
445453, 2001.
Psychopharmacology (Berl),
156(4):
[198] A. Carmon, J. Mor, and J. Goldberg. Evoked cerebral responses to noxious thermal stimuli in humans.
(Cited
on page 157.)
Experimental
[200] A. Carmon, Y. Friedman, R. Coger, and B. Kenton. Single trial analysis of evoked
283
REFERENCES
potentials to noxious thermal stimulation in man.
(Cited
on page 159.)
Anaesthesia,
42(2):195198,
British Journal of
Objective correlate of
The Journal of
Amplitudes
284
Pain, 110(1-2):318328,
REFERENCES
[208] E. Kochs, E. Scharein, O. Mllenberg, B. Bromm, and J. am Schulte Esch. Analgesic ecacy of low-dose ketamine: Somatosensory-evoked responses in relation to
subjective pain ratings.
Brain Research,
[210] C. R. France, J. L. Rhudy, and S. McGlone. Using normalized emg to dene the
nociceptive exion reex (nfr) threshold: further evaluation of standardized nfr
scoring criteria.
Algesimetry:
2(1):
American
285
REFERENCES
and P. Bjerring. Subhypnotic doses of thiopentone and propofol cause analgesia to
experimentally induced acute pain.
1991.
[217] T. C. O'Connor and S. E. Abram. Inhibition of nociception-induced spinal sensitization by anesthetic agents.
Anesthesiology,
82(1):259266, 1995.
(Cited on
Congress, 2009.
Oral Presentation.
Propofol automatic
26(45):34, 2009.
(Cited on
[220] N. Ty Smith, M. L. Quinn, J. Flick, Y. Fukui, R. Fleming, and J. R. Coles. Automatic control in anesthesia: a comparison in performance between the anesthetist
and the machine.
63(8):715722, 1984.
(Cited on
[221] G. Schils, F. Sasse, and V. Rideout. Automatic control of anesthesia using two
feedback variables.
and 200.)
286
pages 1
REFERENCES
[223] A. R. Absalom and G. N. C. Kenny. Closed-loop control of propofol anaesthesia
using bispectral index
controlled propofol and manually controlled remifentanil infusions for minor surgery.
[224] C. Schaniel, A. Gentilini, M. Morari, C. Bieniok, R. Wymann, and T.W. Schnider. Perioperative closed-loop control of analgesia in humans. In
Engineering in
Medicine and Biology Society, 2001. Proceedings of the 23rd Annual International
Conference of the IEEE,
and 200.)
[225] S. Yelneedi, S. Lakshminarayanan, and G.P. Rangaiah. Advanced regulatory controller for automatic control of anesthesia. In
In
Closed-loop
Anaesthesiaa, 53
[228] C.W. Frei, M. Derighetti, M. Morari, A.H. Glattfelder, and A.M. Zbinden. Improving regulation of mean arterial blood pressure during anesthesia through estimates of surgery eects.
1456 1464, nov. 2000.
287
REFERENCES
index as measures of the electroencephalographic eects of propofol.
Applications.
288
Anesthesia