New Indications for Peripheral Pulse Wave Acta Universitatis Tamperensis 2408

JARKKO HARJU New Indications for Peripheral Pulse Wave Acta Universitatis Tamperensis 2408

JARKKO HARJU

New Indications for Peripheral Pulse Wave AUT 2408 JARKKO HARJU

New Indications for Peripheral Pulse Wave

ACADEMIC DISSERTATION To be presented, with the permission of the Faculty Council of the Faculty of Medicine and Life Sciences of the University of Tampere, for public discussion in the auditorium of Finn-Medi 5, Biokatu 12, Tampere, on 14 September 2018, at 12 o’clock.

UNIVERSITY OF TAMPERE JARKKO HARJU

New Indications for Peripheral Pulse Wave

Acta Universitatis Tamperensis 2408 Tampere University Press Tampere 2018 ACADEMIC DISSERTATION University of Tampere, Faculty of Medicine and Life Sciences Tampere University Hospital, Department of Anaesthesia Finland

Supervised by Reviewed by Professor Arvi Yli-Hankala Docent Vesa Kontinen University of Tampere University of Turku Finland Finland Professor Niku Oksala Professor Tarmo Lipping University of Tampere Tampere University of Technology Finland Finland

The originality of this thesis has been checked using the Turnitin OriginalityCheck service in accordance with the quality management system of the University of Tampere.

Cover design by Mikko Reinikka

Acta Universitatis Tamperensis 2408 Acta Electronica Universitatis Tamperensis 1918 ISBN 978-952-03-0821-6 (print) ISBN 978-952-03-0822-3 (pdf) ISSN-L 1455-1616 ISSN 1456-954X ISSN 1455-1616 http://tampub.uta.fi

Suomen Yliopistopaino Oy – Juvenes Print

Tampere 2018 441 729 Painotuote To Eeva, Anna and Olli

List of original publications

This thesis is based on the following four original publications, referred to in the text by their Roman numerals (I-IV):

I Matthias Gruenewald, M.*, Harju, J.*, Preckel, B., Molnár, Z., Yli- Hankala, A., Roßkopf, F., Koers, L., Orban, A., Bein, B. and the AoA- study Collaborators. Comparison of Adequacy of Anesthesia monitoring with standard clinical practice during routine general anesthesia: an international, multi-center, single-blinded RCT. Submitted II Harju, J., Kalliomaki, M.L., Leppikangas, H., Kiviharju, M. & Yli- Hankala, A. (2016). Surgical pleth index in children younger than 24 months of age: a randomized double-blinded trial, British journal of anaesthesia, 117(3), 358-364. III Harju, J., Vehkaoja, A., Kumpulainen, P., Campadello, S., Lindroos, V., Yli-Hankala, A. & Oksala, N. (2018). Comparison of non-invasive blood pressure monitoring using modified arterial applanation tonometry with intra-arterial measurement. Journal of Clinical Monitoring and Computing, 32(1), 13-22. IV Harju, J., Vehkaoja, A., Lindroos, V., Kumpulainen, P., Liuhanen, S., Yli-Hankala, A. & Oksala, N. (2017). Determination of saturation, heart rate, and respiratory rate at forearm using a Nellcor™ forehead SpO2- saturation sensor", Journal of Clinical Monitoring and Computing, 31(5), 1019-1026.

*) Contributed equally

The original publications are reprinted with the kind permission of the copyright holders. Abbreviations

CO2 carbon dioxide BIS Bispectral index BSR Burst Suppression Ratio CI confidence interval CVI Composite variability index EEG electroencephalogram ECG electrocardiogram EMG electromyography ETAC end tidal anesthetic control HR heart rate ICU intensive care unit LoA limits of Agreement NIRS Near-infrared spectroscopy NMBA neuromuscular blocking agent NoL Nociception level MAC minimum alveolar concentration

O2 oxygen ORI Oxygen Reserve Index PACU post anesthesia care unit PONV post-operative nausea and vomiting PPG photoplethysmography PPGA pulse photoplethysmographic amplitude PPGAnorm normalized pulse photoplethysmographic amplitude PVI Pleth variability index RE Response entropy RMSE root-mean-square-error RR respiratory rate RRI RR interval RRInorm normalized RR interval SD standard deviation SE State entropy

SaO2 arterial oxygen saturation

SpO2 arterial plethysmographic oxygen saturation SPI Surgical Ppleth Index

StO2 tissue oxygen saturation SSI Surgical stress index TOF Train-of-four VAS Visual analogue scale Abstract

The peripheral pulse wave contains a lot of important information that can be used in patient monitoring. It is easily accessible but contains several sources of potential error in measurement. The solutions using peripheral pulse wave include measurement of nociception, saturation, heart rate, respiratory rate and blood pressure among several other parameters. Surgical Pleth Index (SPI) is a novel algorithm for measuring intraoperative nociception. It combines the normalized plethysmographic amplitude and heart rate as one number indicating nociception and has been used intraoperatively to guide opioid administration. Although several studies have shown the reactivity of SPI to nociceptive stimulus, the evidence in guiding opioid administration is still scarce. Nor have such studies been accomplished on very small children. A typical site to measure SPI and other peripheral pulse wave parameters is the fingers. However, the measurement at the fingers is susceptible to artifacts caused by movement, vasoconstriction or hypothermia. A watch-like device at the distal forearm would offer a better and more convenient fixation for the measurements. Furthermore, there is some evidence that the forearm might be a better place for respiratory rate measurement using the same technology. There is at present no reliable method to measure blood pressure using plethysmography, but a wrist device with a tonometry sensor added would be one option to add that information, too. The aim of this thesis was to study new ways to utilize the peripheral pulse wave. The technologies used were plethysmography (Studies I, II and IV) and applanation tonometry (Study III) Study I was a randomized multi-center controlled trial with 494 patients. Patients were anesthetized using either standard monitoring (control group) or standard monitoring with additional monitoring of the depth of anesthesia (EntropyTM) and nociception (SPI) (test group). In this study we found no difference in the rate of signs of inadequate anesthesia. At the secondary endpoints, the time to eye-opening showed a trend in favor of the test group. In Study II the reactivity of the SPI was studied in thirty children aged less than two years. Children were randomized and double-blinded into two groups with either functional peripheral block or a placebo injection during surgery. SPI and its components were recorded blinded using a software and the time points of interest were analyzed post hoc. SPI was found to be reactive in small children and the reactivity was blunted by the use of peripheral local anesthetic block. However, the reactivity of SPI was rather small and there was marked inter-individual variability in reactions. The data for Studies III-IV was collected simultaneously. Thirty patients were monitored during postoperative care for two hours as an observation study. The data were collected using a study device attached around the distal forearm. Study III reported the measurement of non-invasive blood pressure using tonometry compared to invasive blood pressure monitoring. The blood pressure readings were found to be unacceptably inaccurate. Furthermore, the rate of failed measurements was rather high (22%). Factors affecting rate of failure consisted of movement and peripheral arterial disease. In Study IV the reliability of heart rate, respiratory rate, and saturation measurement at the distal forearm using a plethysmography sensor was compared to standard monitoring from finger plethysmography (saturation and heart rate) or impedance pneumography (respiratory rate). There was a small bias in heart rate, respiratory rate, and saturation measurements when compared to standard monitoring. However, the accuracy described by root-mean-square-error (RMSE) was unacceptably high for respiratory rate and saturation while heart rate was detected with good accuracy. Movement was associated with higher RMSE. In conclusion, the monitoring of SPI and Entropy was not associated with better management of anesthesia in a large multicenter study. Time to eye- opening was slightly shorter in the test group when compared to control group. In small children SPI seems to be similarly reactive as in adults in spite of wide inter-individual variability in reactions. A higher baseline and smaller amplitude of change when compared to adults suggest that a modification of the algorithm might be needed before introducing the index for use with small children. The measurement of blood pressure, heart rate, and saturation at the distal forearm was unacceptably inaccurate, while respiratory rate yielded slightly better results. Blood pressure measurement especially seems to be highly sensitive to movement, also at the distal forearm. Accuracy therefore needs to be improved before adopting the technologies for patient monitoring.

Tiivistelmä

Perifeerisen pulssiaallon sisältämää tietoa voidaan käyttää potilaan seurannassa monin tavoin. Pulssiaalto on helppo saada mitattua, mutta mittaaminen on altis useille virhelähteille. Mitattavia suureita ovat olleet muun muassa kipuärsyke, kudoshapetus, sydämen syke, hengitystaajuus, verenpaine sekä useat muut elintoiminnot. Surgical Pleth Index (SPI) on leikkauksen aikaisen kipuärsykkeen mittaamiseen tarkoitettu lukuarvo. Se yhdistää muutokset pulssiaallon suuruudessa sekä sykkeen vaihtelussa yhdeksi leikkauksen aikaista kipuärsykettä kuvaavaksi lukemaksi. Useissa tutkimuksissa on osoitettu sen vaihtelu suhteessa aiheutettuun kipuärsykkeeseen. Kuitenkin sen näyttö opiaattien annostelua ohjaavana suureena on vielä vähäinen. Mittarin toimivuutta pienillä lapsilla ei ole myöskään määritetty. Tyypillisesti perifeeristä pulssiaaltoa hyödyntävät mittarit, kuten myös SPI, on optimoitu toimimaan sormessa. Mittauspaikkana tämä on kuitenkin hyvin altis liikehäiriöille, suonten supistumiselle ja lämpötilan vaihteluille. Siksi ranteessa toimiva mittari saattaisi tarjota paremman ja soveltuvamman paikan sensorille. Hengitystaajuuden mittaamisessa on olemassa rajallista näyttöä, joka puoltaisi käsivarren käyttöä mittauspaikkana. Plethysmografia ei mittaustapana vielä mahdollista verenpaineen mittausta, mutta herkän paineanturin yhdistäminen laitteeseen saattaisi mahdollistaa myös sen mittaamisen samalla laitteella. Tämän väitöskirjan tarkoituksena oli tutkia uusia tapoja hyödyntää perifeeristä pulssiaaltoa. Tutkimuksissa käytettiin hyödyksi plethysmografiaa (osatyöt I, II ja IV) sekä herkkää paineanturia (osatyö III). Osatyö I oli satunnaistettu, kontrolloitu monikeskustutkimus, johon osallistui yhteensä 494 potilasta. Tutkimuksessa unen syvyyden (Entropia) ja kipuärsykkeen (SPI)(testiryhmä) hyötyä verrattuna perinteiseen monitorointiin (kontrolliryhmä) arvioitiin kahden satunnaistetun ryhmän avulla. Tutkimuksessa ryhmien välillä ei todettu merkittävää eroa leikkauksen aikaisissa haittavaikutuksissa. Toissijaisissa muuttujissa testiryhmässä silmien avaaminen tapahtui hieman kontrolliryhmää nopeammin. Osatyössä II, SPI:n toimivuutta tutkittiin 30:llä, alle kaksivuotiaalla potilaalla. Tutkittavat satunnaistettiin saamaan toimiva puudutus tai lumelääke toimenpiteen ajaksi. Tutkittavat suureet kerättiin sokkoutetusti tietokoneohjelman avulla. SPI arvo ei vaikuttanut anestesian suorittamiseen. Tutkimuksessa SPI:n todettiin reagoivan myös pienillä lapsilla ja muutos oli pienempää ryhmässä, jossa lapsilla oli toimiva puudutus. Muutokset SPI:ssä olivat kuitenkin pieniä ja niiden suuruudessa oli suuria vaihteluita potilaiden välillä. Tutkimusaineisto osatöihin III ja IV kerättiin samanaikaisesti. Tutkimussuureet mitattiin 30:ltä potilaalta kahden tunnin heräämöseurannan aikana ranteen ympärille asennetulla tutkimusmittarilla. Osatyössä III paineanturilla kajoamattomasti mitattuja verenpainearvoja verrattiin kajoavaan mittaukseen. Tutkimuslaitteen mittaamat lukemat todettiin epätarkoiksi ja eri syistä johtuvien epäonnistuneiden mittausten määrä oli varsin korkea (21.6%). Epäonnistuneen mittauksen todennäköisyyttä nostivat liike mittauksen aikana sekä perifeerinen valtimonkovettumatauti. Osatyössä IV sydämen syke, hengitystaajuus ja saturaatio mitattiin ranteesta plethysmografia-sensoria käyttäen ja sitä verrattiin joko sormesta mitattaviin (syke, saturaatio) tai rintakehältä mitattaviin (hengitystaajuus) arvoihin. Tutkimuslaitteen ja verrattavan monitorin mittaamien sykkeen, hengitystaajuuden ja saturaatioarvojen todettiin eroavan keskimäärin vain vähän. Mittalaitteiden tarkkuus kuvattuna root-mean-square-error (RMSE) lukemalla oli kuitenkin huono hengitystaajuuden ja saturaation osalta. Sykkeen osalta myös tarkkuus oli hyvä. Mittauksen aikainen liike heikensi mittaustarkkuutta. Tutkimuksen johtopäätöksinä voidaan esittää, että SPI:n ja Entropian lisääminen ei parantanut nukutuksen laatua laajan monikeskustutkimuksen perusteella. Silmien avaaminen vaikutti olevan hieman nopeampaa testiryhmässä. Pienillä lapsilla SPI vaikuttaa yhtä lailla reaktiiviselta kuin aikuisilla, mutta yksilöiden välillä todettiin isoja vaihteluita. Aikuisiin verrattuna korkeamman perustason ja kokoluokaltaan pienemmän muutoksen perusteella mittauslaskennan muutos voi olla pienillä lapsilla tarpeen. Verenpaineen, hengitystaajuuden ja saturaation mittaus ranteesta antoi liian epätarkkoja mittatuloksia. Hengitystaajuuden mittaus oli hieman luotettavampaa. Erityisesti verenpaineen mittaus oli hyvin herkkä liikkeen aiheuttamille häiriöille, joten kaikkien mittausten virheen sietoa täytyy saada parannettua ennen niiden soveltamista potilaskäyttöön. Contents

List of original publications ...... 4 Abbreviations ...... 5 Abstract ...... 7 Tiivistelmä ...... 9 1 Introduction ...... 15 2 Review of the literature ...... 18 2.1 The physiology of photoplethysmographic waveform measurement 18 2.2 Adequacy of anesthesia ...... 19 2.3 EEG-based depth of anesthesia monitoring ...... 21 2.3.1 Background ...... 21 2.3.2 Entropy ...... 23 2.3.3 EEG-based depth of anesthesia monitoring – Limitations ...... 24 2.4 Measurement of relaxation ...... 26 2.5 Detection of nociception ...... 26 2.5.1 Surgical Pleth Index ...... 27 2.5.2 Surgical Pleth Index – Clinical studies ...... 29 2.6 Measurement of blood pressure ...... 32 2.6.1 Applanation tonometry...... 33 2.6.2 Finger cuff method ...... 35 2.6.3 Continuous blood pressure measurement using pulse transit time 36 2.7 Measurement of saturation ...... 37 2.7.1 Transmission photoplethysmography ...... 39 2.7.2 Reflectance photoplethysmography ...... 39 2.7.3 Near-infrared spectroscopy ...... 40 2.8 Measurement of heart rate ...... 41 2.9 Measurement of respiratory rate ...... 42 2.9.1 Estimation of respiratory rate using plethysmographic waveform ...... 42 2.10 Measurement of fluid responsiveness...... 46 2.11 Technologies based on multiple wave measurement ...... 47 2.12 Summary of the theoretical background ...... 47 3 Aims of the study ...... 49 4 Patients and methods ...... 50 4.1 Patients ...... 50 4.2 Methods ...... 51 4.2.1 Study designs ...... 51 4.2.2 Data collection ...... 53 4.3 Statistical analysis ...... 54 4.3.1 Sample size estimation ...... 54 4.3.2 Data analysis ...... 55 4.4 Ethical statement ...... 55 5 Summary of the results ...... 57 5.1 Adequacy of Anesthesia (I) ...... 57 5.2 Reactivity of SPI in small children (II) ...... 66 5.3 Comparison of modified arterial applanation tonometry with intra- arterial pressure measurement. (III) ...... 69 5.4 Detection of vital signs using plethysmography (IV) ...... 72 5.4.1 Comparison of reflection type saturation measurement at the distal forearm with finger measurement using transmission mode. (IV) 72 5.4.2 Comparison of plethysmographic heart rate measurements (IV) 72 5.4.3 Comparison of tonometric heart rate with intra-arterial measurement (IV) ...... 72 5.4.4 Comparison plethysmographic respiratory rate measurement to impedance pneumography (IV) ...... 74 6 Discussion ...... 75 6.1 Adequacy of anesthesia (I)...... 75 6.2 SPI in small children (II)...... 77 6.3 Measurement of blood pressure using modified applanation tonometry (III) ...... 78 6.4 Indications for photoplethysmography ...... 80 6.4.1 Measurement of saturation at distal forearm (IV) ...... 80 6.4.2 Measurement of heart rate (IV) ...... 81 6.4.3 Measurement of respiratory rate (IV) ...... 81 6.4.4 Measurement of vital signs at the distal forearm ...... 82 6.5 Strengths and weaknesses of the studies ...... 82 6.6 Clinical aspects and future perspectives ...... 84 7 Conclusions ...... 86 8 Acknowledgements ...... 87 9 References ...... 89

1 Introduction

The peripheral pulse wave contains a plethora of important information, part of which is still poorly utilized and little known. The most commonly measured vital signs using peripheral pulse wave are probably arterial oxygen saturation and heart rate defined by peripheral oximeter (Korhonen & Yli-Hankala 2009). The recommendations of the American Society of Anesthesiologists guide us to routinely monitor pulse oximetry during any operation or sedation (ASA House of Delegates 2015). In addition to pulse oximetry there are several other clinical implications. Typically, the measurement consists of a light source illuminating the tissue and a detector that collects the incoming light. It is often used non- invasively to take measurements through the skin surface. The changes in the light intensity that is recorded consist of a pulsatile and non-pulsatile slowly varying background component. Both can and have been used to measure vital signs (Allen 2007). Among the other implications, peripheral pulse wave has been used as an interpreter for nociception to achieve adequate anesthesia. The stress response to nociceptive stimulus is known to cause changes in heart rate, blood pressure, and blood circulation by activation of sympathetic, neural, and humoral pathways (Weissman 1990). Some of the humoral responses are still poorly understood, but different approaches may be required when trying to achieve analgesia, prevent hemodynamic changes or stress response (Wolf 2012). The control of nociceptive reaction aims to reduce the stress response and balance the administration of anesthesia (Chen et al. 2010). The measurement of intraoperative nociception has traditionally been based on observations of heart rate (HR), blood pressure changes, and detection of patient’s movement and muscle tension. Several attempts have thus been made to objectively and specifically measure the nociception, and of these the Surgical Pleth Index (SPI) is probably the most studied. It combines information on changes in HR and plethysmography waveform as a simple numerical index describing the amount of surgical stress (Huiku et al. 2007). There is increasing evidence on the performance of SPI in different clinical scenarios (Bergmann et al. 2013; Chen et al. 2010; Won et al. 2016), but the role

of SPI as a part of ‘balanced anesthesia practice’ is still little researched. The evidence from children is also very limited (Kallio et al. 2008; Park et al. 2015). As a second aspect of the present thesis, the peripheral pulse wave can be used to measure common clinical signs. Basic vital signs are strongly associated with in-hospital mortality. Presence of any of the six events: decrease or loss of consciousness, hypotension, decreased respiratory rate, decreased saturation or marked tachypnea, were associated with a 6.8-fold increase in risk of mortality (Buist et al. 2004). Earlier research has reported cardiopulmonary arrest to be typically preceded by clinical antecedents (Schein et al. 1990). Some of these may be prevented by setting predefined limits for basic vital signs (Jones et al. 2011; Tirkkonen et al. 2014). The measurement of vital signs from a peripheral location would offer an easily accessible option for monitoring and there is currently extensive research on the development of such devices (Khan et al. 2016). Arterial oxygen saturation, along with pulse rate, is typically measured at the finger with good sensitivity and specificity. Yet the finger is prone to movement and the measurement is greatly affected by hypothermia and hypotension (Bohnhorst et al. 2000; Jubran 2015). The pulse plethysmography waveform is also subject to modulation by breathing. Several recent reports have defined respiratory rate using plethysmography (Addison, Watson et al. 2015; Garde et al. 2014; Karlen et al. 2014). The gold standard for continuous blood pressure measurement is invasive measurement, typically measured at the radial artery. The open bloodline is considered a serious risk on regular wards and needs to be monitored vigilantly, which limits its feasibility in many clinical scenarios (Chim et al. 2015; Slogoff et al. 1983). New ways to measure blood pressure continuously or near- continuously are therefore constantly under development. The use of a pressure sensor placed on a peripheral artery has long been studied (Wolff 1969). Yet only recent developments in technology have achieved a potentially convenient and portable method for measurement (Nair et al. 2008). The development of battery and mobile technology offers a tempting opportunity for near-continuous remote monitoring of a patient’s clinical condition. The use of the distal forearm as a site for measurement would allow the use of watch-like devices with much higher comfortability than multiple sensors at different locations. This thesis aims to study new ways to utilize the peripheral pulse wave expanding its potential in perioperative treatment. The techniques used to

16 measure peripheral pulse wave consist of photoplethysmographic sensor (Studies I, II and IV) and modified applanation tonometry (III). Study I is concerned with the concept of ‘adequacy of anesthesia’, which aims to guide the anesthesia by measuring the depth of anesthesia (Entropy™), relaxation (Train-of-four, TOF) and nociception (Surgical Pleth Index, SPI). Study II aimed to examine SPI reactivity in small children. Studies III and IV focus on measuring vital signs at the distal forearm.

2 Review of the literature

2.1 The physiology of photoplethysmographic waveform measurement

The optical measurement technique used to detect changes in the microvascular bed of tissue is known as photoplethysmography (PPG). The waveform consists of two components: a pulsatile component dependent on heart rate and a slowly varying baseline component, more dependent on respiration, sympathetic nervous system activity, and thermoregulation. A light source to illuminate the tissue and a photo detector are used to measure slight variations in light intensity. Typically, red or near infrared wavelengths which are long and penetrating have been used in the sensors (Allen 2007). In addition, sensors using green and blue wavelengths have been studied or are under development (Lee et al. 2013). In particular, the use of less penetrating wavelengths has been speculated to enhance resistance to background artifacts (Maeda et al. 2011b; Matsumura et al. 2014; C. Zhou et al. 2016). Several factors affect the signal and its quality. The sensor may press the tissue too tightly, impeding circulation and thereby reducing the changes in the waveform. Ambient light and especially movement of the sensor and tissue may cause significant artifacts in measurement. The absorption of light in tissue is calculated using the Beer-Lambert law, where absorption is dependent on wavelength, the absorptivity coefficient of the material and the concentration of the material. However, the law assumes a simple, homogenous tissue, which is usually not true in vivo. This means that the light absorption characteristics are heavily dependent on the placement area (Reisner 2008). Even the velocity of the blood flow has been reported to affect PPG measurement (Visser et al. 1976). Lastly, myoglobin absorbs the same wavelengths as hemoglobin (Arakaki et al. 2010). It has been estimated that 75% of the skeletal muscle blood is located in the venous compartment, which may result in falsely low oxygen saturation readings if saturation is measured at sites with thick muscle layer (Mesquida et al. 2013). This has limited the use of technology at sites where the tissue is thin, such as the fingers for absorption-type measurements (Nitzan et

18 al. 2014). The reflection types of probes are typically used at sites where there is a bony support near surface such as forehead (Nesseler et al. 2012). Photoplethysmography technology is relatively cheap, lightweight, low energy consuming and easy to use (Allen 2007). This has led to a wide variety of devices exploiting such technology. This includes measurement of saturation, heart rate (Nitzan et al. 2014), nociception (Chen et al. 2010), respiratory rate (Folke et al. 2003), fluid volume (Monnet et al. 2005), blood pressure (Hennig & Patzak 2013; Silke & McAuley 1998) and even to guiding the selection of respiratory settings (J. Zhou & Han 2016). The basis of measurement and different devices using plethysmography are aptly described in a review by Allen (Allen 2007). As a measurement of nociception, photoplethysmography has been used as a component in quantifying adequate anesthesia.

2.2 Adequacy of anesthesia

The state of “balanced” anesthesia is a balance between the amount of anesthetic drug(s) administered and the state of arousal of the patient. Traditionally, anesthesia has been guided by signs of the autonomic nervous system such as changes in blood pressure, heart rate or movement. However, these are often insufficient to avoid adverse effects of anesthetics or provide sufficient anesthesia. An analysis of awareness showed that once awareness occurred, it was not reliably indicated by either of hypertension (only 15% of all cases), tachycardia (only 7% of all cases), or movement (only 2% of all cases) (Domino et al. 1999). The challenge to adequately measure intraoperative unconsciousness is as old as the concept of general anesthesia. The first public anesthesia demonstration in 1845 at the Massachusetts General Hospital in Boston by Dr. Horace Wells turned out unsuccessful, with the patient moving and groaning during the procedure conducted under nitrous oxide anesthesia. In the classical ether demonstration soon after by William T. G. Morton, the patient reported feeling first pressure and then having lots of wonderful dreams (Robinson & Toledo 2012). The success of the operation started an era of general anesthesia. The invention of muscle relaxants and their adaptation to anesthesia in the 1940’s markedly increased the incidence of anesthesia awareness. Thus the warning of the possibility of patients being relaxed and awake due to muscle relaxants was first published at 1945 in the Lancet editorial Mainzer Jr &

Mainzer1979). This raised the question of the ability to measure the depth of anesthesia. The MAC concept, originally published in 1965 to compare different volatile anesthetics (Eger et al. 1965), has been used as one estimate for depth of anesthesia (Quasha et al. 1980). MAC stands for minimal alveolar concentration and the median effective dose describes the value at which 50% of patients are expected not to react to surgical incision. Similarly, MAC-awake describes the value at which patients are not expected to react to verbal stimuli (Stoelting et al. 1970) and MAC-Bar the minimum concentration required to block adrenergic responses (Roizen et al. 1981). Although these have been used to guide anesthesia (Avidan et al. 2008; Avidan et al. 2011), they continue to be population-based measures according to the definition. Studies on decorticated animals have also shown that twice as high MAC is needed to prevent movement in decorticated animals when compared to animals with intact spinal cord (Antognini & Schwartz 1993; Rampil 1994). The Observer’s Assessment of Alertness/ Sedation Scale (OAA/S) quantifies a combination of observations of the resting patient (expression, eyes) and of patient’s responses to verbal commands (responsiveness, speech) with increasing intensity, and describes the level of sedation on a numerical scale (Chernik et al. 1990). It has been validated for most benzodiazepines, but a modification of the algorithm Modified Observer’s Assessment of Alertness/Sedation Scale (MOAA/S) allows assessment of deeper levels of sedation with assessment of reaction to painful stimuli by using only the responsiveness component of the original scale. However, the assessment of responsiveness introduces a second problem: the intervention itself changes the level of sedation due to the use of arousal stimuli with increasing intensity. Anesthesia is currently considered to consist of three components: hypnosis, immobility, and analgesia. In other words, this means that the patient will remember nothing, will not move, and that the patient’s pshysiological homeostasis will not be too severely disturbed. The development of EEG indexes helped to understand the physiology of the reactions (Antognini & Carstens 2002). The target of achieving “adequate anesthesia” reflects the intention to adequately measure the depth of anesthesia by measuring relaxation, amnesia, and nociception thereafter reducing unwanted movement, memory recall, and to improve the surgical conditions.

2.3 EEG-based depth of anesthesia monitoring

2.3.1 Background

Sufficient anesthesia is needed to provide the state of amnesia. In modern anesthesia the incidence of intraoperative awareness with implicit recall varies between 0.2% to 1% (Mashour et al. 2012; Mashour et al. 2009; Sebel et al. 2004; XU et al. 2009) depending on the method of assessment and patient population (Mashour & Avidan 2015). In the clinical situation defining a state of amnesia can be challenging, thus the drugs induced to achieve a proper state of amnesia vary widely in efficacy across individuals. One way to guide the delivery of anesthetics is to measure the concentration of volatile anesthetics in the exhaled gas (Chhabra et al. 2016; Punjasawadwong et al. 2014). Direct measurement of intravenous anesthetic concentration would be more challenging. Thus attempts have been made to estimate anesthetic concentration based on estimated concentration with noxious stimulation response index, but this is as yet poorly validated (Luginbuhl et al. 2010). Both these approaches are based on assumption of the desired effect and not measurement of the desired target. EEG measurement has been proposed as a tool to measure depth of anesthesia. It was shortly after its discovery when it was first seen as a potential tool for anesthesia (Gibbs et al. 1937). However, raw EEG is rather cumbersome and hard to use as a clinical monitor. Within the last thirty years several attempts have been made to adequately measure the depth of anesthesia (Bruhn et al. 2006). These include the bispectral index (BIS; Covidien, Boulder, Colorado, USA), State and Response Entropy (SE and RE, GE Healthcare Technologies, Helsinki, Finland), Narcotrend Index (Schiller AG, Baar, Switzerland), Patient State Index (PSI; Hospira, Lake Forest, Illinois, USA), SNAPII (Everest Biomedical Instruments, Chesterfield, Missouri, USA), and the Cerebral State Monitor (Danmeter A/S, Odense, Denmark). Of these the most studied monitor is BIS. In high-risk patients, BIS has been reported to markedly reduce the risk of awareness during anesthesia (Myles et al. 2004). A recent Cochrane review gathered 36 trials and demonstrated odds ratio 0.24 with 95% Confidence Interval 0.12 to 0.48 risk reduction for intraoperative awareness among surgical patients with high risk of awareness (Punjasawadwong et al. 2014). BIS-guided

anesthesia also reduced the need for propofol by 1.32 mg/kg/h and for volatile anesthetics by 0.65 MAC. Time to eye opening was reduced by 1.92 min (95% CI -2.70 to -1.16), time to response to verbal commands by 2.73 min (95% CI - 3.92 to -1.54) time to extubation 2.62 min (95% CI -3.46 to -1.78), and duration of post anesthesia care unit stay by 6.75 min (95% CI 11.20 to -2.31). Concerning awareness, the evidence was inconclusive, stating that end tidal anesthetic control (ETAC) with and audible alarm may be equivalent in protection against intraoperative awareness mostly based on studies by Avidan et al. (Avidan et al. 2008; Avidan et al. 2011). Furthermore, a Cochrane review summarizing the impact of all depth of anesthesia monitoring devices did not prove effectiveness in preventing awareness during anesthesia (Messina et al. 2016). The cumulative duration of low BIS index has been associated with mortality after anesthesia (M. Kertai et al. 2010; Leslie et al. 2010; Lindholm et al. 2009; Monk et al. 2005), while some later and better conducted studies have found no such association (M. D. Kertai et al. 2011; Lindholm et al. 2011). The association may be more related to background factors than to the conduct of anesthesia. Similar findings have been reported in association with the incidence of postoperative cognitive dysfunction or delirium (Chan et al. 2013; Mathews et al. 2012; Radtke et al. 2013) but well conducted studies on strategies for preventing post-operative delirium are still lacking (Orena et al. 2016). At the other end of anesthesia practice, in other words sedation, depth of anesthesia monitors have not achieved correlation with sedation scores (Haenggi et al. 2009). Another approach to guiding anesthesia is based on tight ETAC with an audible alarm and minimum alveolar concentration of 0.7 to 1.3 (Avidan et al. 2011; Avidan et al. 2008). According to these studies, it is beyond question that depth of anesthesia measurement is superior in preventing anesthesia awareness compared to clinical symptoms, but not necessarily when compared to an audible anesthetic concentration alarm (Mashour & Avidan 2015). When intravenous anesthetics are used, the measurement of anesthetic concentration is not possible at present and there is strong evidence supporting the use of monitoring for depth of anesthesia (Mashour et al. 2012; Myles et al. 2004; Punjasawadwong et al. 2014, Zhang et al. 2011). Similar findings suggest that Entropy and Narcotrend have similar effect on quality of anesthesia (Aime et al. 2006; Chhabra et al. 2016; El Hor et al. 2013; Gruenewald et al. 2007; Shepherd et al. 2013).

Most of these studies report lowered consumption of anesthetics a well as shorter intraoperative times. Nevertheless, the reduction in anesthetics has not led to an increase in anesthesia awareness (Chhabra et al. 2016; Punjasawadwong et al. 2014). For this reason, it seems evident that EEG-based depth of anesthesia monitoring can be safely used in order to optimize the conduct of anesthesia and to ensure the quality of anesthesia. This applies especially when muscle relaxants are used and the patient is rendered unable to react to raising awareness or in association to intravenous anesthetics, when no measurement of anesthetic concentration is available.

2.3.2 Entropy

Entropy® consists of measuring three different values, Response Entropy (RE), State Entropy (SE) and Burst Suppression Ratio (BSR). The algorithm has been published in full once Entropy was introduced (Viertiö-Oja et al. 2004). A three-electrode sensor is placed on the patient’s forehead to collect one-channel EEG. The sensors are located unilaterally from the middle of the forehead to the corner of the eye. In the calculation of Entropy, The SE values are calculated from the 0.8- 32Hz-frequency band and RE from frequencies 0.8-47Hz. The higher frequencies from 32 to 47 Hz are thought to reflect more EMG activity, while the lower frequencies reflect more purely EEG activity. The scale for SE is 0-91 and for RE 0-100, where 0 describes very deep anesthesia and 100 increasing awareness. The first publications used a target range for SE from 45 to 65 (Vakkuri et al. 2004; Vakkuri et al. 2005), and the manufacturer currently recommends from 40 to 60. When the depth of anesthesia increases the patterns in the EEG change and the EEG diminishes to a flat line, typically showing a very low amplitude and high frequency recording. This state is called burst suppression. The proportion of burst suppression is then described with increasing percentages from 0 to 100% by the third number BSR and the algorithm for detection has been described elsewhere (Särkelä et al. 2002).

Figure 1. Schematic illustration of the frequencies included in the measurement of Entropy. Strong filtering is present at lowest frequencies as well as close to 50 Hz in order to remove the artifacts. Frequencies between 32-47 Hz are considered more likely to represent EMG.

2.3.3 EEG-based depth of anesthesia monitoring – Limitations

All EEG-based devices measuring depth of anesthesia are at least partially based on pattern recognition in the EEG and certain frequencies that are interpreted as EEG. Thus, any signal close to EEG frequencies that is being recorded can have a significant misleading effect on the analysis. The electrodes are typically placed on the forehead over the facial muscles. Consequently, the EMG activity caused by frontal muscles is shown to corrupt BIS monitoring and cause falsely elevated measurements (Vivien et al. 2003). In Entropy monitoring this is taken into account by dividing the measurement into two different components: SE with frequencies 0.8-32 Hz and RE 0.8-47 Hz. This

24 is in order to reduce contamination by EMG activity. Nevertheless, the EMG has still been reported to corrupt the signal, especially during intubation (Aho et al. 2009), while the effect can be reduced with the use of neuromuscular blockade (Aho et al. 2011). Also, other electrical activity such as thermal blankets or cardiac pacemakers may erroneously cause changes in the index (Gallagher 1999; Hemmerling & Fortier 2002). In addition to neuromuscular blocking agents, the choice of anesthetic regimen may have a marked effect on the signals. The “awareness” monitors all assume GABAergic drug effect; thus anesthetic agents with different routes of effect may result in different reactions in the recording. Ketamine is an NMDA receptor antagonist causing excitatory neurotransmission in the central nervous system. The excitation causes changes in the EEG patterns different to anesthetics of the GABAA type, such as sevoflurane and propofol. This has been shown to paradoxically cause an increase in both BIS and Entropy indexes under otherwise stable anesthesia (Hans et al. 2005; Vereecke et al. 2003) even though the anesthesia actually deepens. The findings on Xenon anesthesia are limited and partly contradictory (Fahlenkamp et al. 2010; Goto et al. 2000; Laitio2008). Similarly nitrous oxide has not shown a marked effect on the indexes (Anderson & Jakobsson 2004; Hirota et al. 1999; Ramesh & Rao 2006; Rampil et al. 1998), while at deeper levels of sedation as an addition to sevoflurane it seems to lower indexes (Hans et al. 2001; Ozcan et al. 2010). At isoflurane concentrations high enough to cause burst suppression, nitrous oxide seemed to have an opposing effect increasing the indexes (Yli-Hankala et al. 1993). With limited evidence, dexmedetomidine seems to work as an adjunct lowering BIS values dose dependently (Kwon et al. 2015; T. Wang et al. 2013). Opioids cause a dose dependent change in the EEG measurement, which may lower Entropy readings (Egan et al. 1996). This was seen in the study by Bouillon et al. on reactions to laryngoscopy under different remifentanil and propofol concentrations (Bouillon et al. 2004). In brief, remifentanil lowered the probability for change in Entropy as a reaction to increasing propofol concentrations, but under concentrations of <8 ng/ml the effect was minor, at least in approximate entropy. Another study showed an effect in certain EEG frequencies at lower remifentanil levels (Ferenets 2007). The same effect has also been shown in other EEG studies (Hoffman et al. 1993; Kortelainen et al. 2008; Kortelainen et al. 2009). This may be in relation to reports on awareness during anesthesia when excess amounts of opioids have been administered without sufficient concentration of anesthetics (Vassiliadis et al. 2007; Yli-

Hankala 2008). In contrast, BIS seems to be more resistant to remifentanil, when no effect on the index was seen even at a remifentanil effect site concentration of 16 ng/ml under constant propofol concentration (Guignard 2006). Furthermore, any condition with an impact on the brain, such as epileptic seizure (Smith et al. 2015) or brain ischemia (Goodman 2009), may cause significant changes in the EEG pattern thus affecting the indexes. Finally, the algorithms may fail in pattern recognition. For example, this is described with Entropy showing very high readings despite very deep anesthesia due to failure to recognize the burst suppression (Hart et al. 2009; McCulloch & Thompson 2010. Arousal may also come in many forms. Most typically the frequency of the EEG increases along with increasing EMG activity (β-arousal). However, sometimes EEG arousal is seen as slowing of the rhythm (δ-arousal), which causes a decrease in the EEG index values contrary to expectations (Aho et al. 2011; Bischoff et al. 1993).

2.4 Measurement of relaxation

Neuromuscular blocking agents (NMBAs) have been used in order to enhance endotracheal intubation and to provide better conditions for surgery (Schreiber 2014). NMBAs prevent muscular contractions thereby preventing movement and providing better surgical conditions. However, the use of NMBAs intraoperatively is partly controversial with some evidence suggesting better surgical conditions with a deep neuromuscular blockade during surgery (Dubois et al. 2014; Madsen et al. 2015; Staehr-Rye et al. 2014). In contrast, the problem of postoperative residual curarization is well known, and there is strong evidence advocating quantitative measurement of relaxation when NMBAs are used (Grosse-Sundrup et al. 2012; Lien & Kopman 2014).

2.5 Detection of nociception

Pain is a conscious sensation, which can be assessed using various verbal or visual rating scales (Briggs & Closs 1999; Isik et al. 2011; Pesonen et al. 2009). However, when under deep sedation or general anesthesia there is no pain sensation, but the nociceptive stimulus can still elicit marked reactions. One of

26 the first commercial solutions detecting nociception was the Anaesthesia and Brain Monitor (ABM, Datex, Helsinki, Finland), in which a single channel EEG/EMG display was used to quantify nociceptive reactions. (Andrews et al. 1995). Subsequently several devices have been produced in an attempt to quantify the nociceptive reaction, some of which are described in Table 1 on pages 30-31. Of these, the Surgical Pleth Index (SPI), formerly known as the Surgical Stress Index, is among the best studied. Thus other techniques, especially pupillometry, have gathered some evidence for the effective guiding of opioid administration.

2.5.1 Surgical Pleth Index

SPI is based on variations in photoplethysmographic waveform and heart rate in response to nociceptive stimuli. It combines the plethysmographic amplitude (PPGA) and changes in heart rate (RR interval, RRI) as one number ranging from 0 to 100. Zero indicates no detected nociception and 100 maximal detected nociception. Moreover, both values are normalized against a database value (Huiku et al. 2007). In two other studies RRI and PPGA were added with response entropy (Rantanen et al. 2006; Seitsonen et al. 2005), but the final algorithm contains only the variables from the plethysmography sensor. Several studies have investigated the reactivity of SPI. Struys et al. (2007) first described the reactivity of SPI to tetanic stimulus at various remifentanil concentrations as well as at varying depths of amnesia. SPI showed a better prediction probability for changes after tetanic stimulus than SE, RE, HR or PPG at different remifentanil concentrations, but the probability was still rather low (Struys et al. 2007). Reactivity has likewise been tested under sevoflurane and isoflurane anesthesia (Gruenewald et al. 2009; Mustola et al. 2010) and on children over three years of age (Kallio et al. 2008). In another study, the SPI value diminished as a reaction to opioid bolus, while there was no change in stress hormone levels measured intraoperatively (Ledowski et al. 2010). Ahonen et al. showed that SPI was reactive to surgical stimulus and the reaction was diminished in the group receiving remifentanil when compared to the group receiving short acting β-adrenergic antagonist esmolol (Ahonen et al. 2007). In contrast, when the nociceptive reaction was blocked by local anesthetics, the reactions in the index were lower (Paloheimo et al. 2010) and the reactivity was regained after fading blockade (Wennervirta et al. 2008). In one study change in

SPI was a good predictor of movement in response to a tetanic stimulus (Gruenewald, Herz et al. 2014). A limitation of the measurement is that it is related to heart rate and the use of drugs or devices affecting HR have a marked effect on the measurement (Hocker et al. 2010). In another study the reaction was not totally blunted with the use of beta-blockers (Ahonen et al. 2007). Similarly, trendelenburg posture increased SPI in patients under general anesthesia, spinal anesthesia, and spinal anesthesia with sedation probably in response to increased vascular load. The reaction was consistent for at least 45 minutes (Ilies, Ludwigs et al. 2012). A rapid fluid bolus may also change SPI, but the effect is related to volemic status (Hans et al. 2012). The potential limitations due to the physical and technical artifacts in measurement are well described in an article by Korhonen & Yli- Hankala (2009) and even probe movement or too tense sensor pressure on the skin may cause significant artifacts corrupting the signal (Korhonen & Yli- Hankala 2009). Although the potential changes seen in the PPGA waveform are well known, how these convert into changes in SPI has not been thoroughly researched. Moreover, so far no studies have been reported on patients receiving catecholamines such as noradrenaline infusion, which theoretically would affect the measurement. Most studies conducted with SPI have used propofol as a sedative and remifentanil as an opioid. Typically a target range from 20 to 50 has been used (Ahonen et al. 2007; Bergmann et al. 2013; Gruenewald et al. 2009; Kallio et al. 2008). In one study, intraoperative score of SPI <30 was predictive of lower postoperative pain score (Ledowski et al. 2016) suggesting lower target values than previously used. A rapid change in the value will probably also correlate with clinical signs (Gruenewald et al. 2009; Gruenewald et al. 2015). The optimal target range for SPI is based on rather limited evidence. The usefulness of SPI is limited to the intraoperative period under general anesthesia. Although SPI seems to correlate with experimental pain by cold and heat stimuli in healthy subjects (Hamunen et al. 2012), values obtained in the postoperative period do not seem to correlate either with stress hormones (Chen et al. 2012) or self assessment of pain using Numeric Rating Scales (Thee et al. 2015). A similar finding was seen in fully awake patients under spinal anesthesia where SPI was constantly at a higher level during surgery in the spinal group compared to the general anesthesia group and even slight sedation returned the levels to baseline levels (Ilies et al. 2010). It seems evident that in fully awake patients other factors, such as emotions, affect the index making the

28 interpretation of nociception more difficult. However, this does not exclude the use of SPI for guiding relief from anxiety in order to improve patient satisfaction. So far no studies have been published on this approach. In pediatric population, only two studies have investigated the performance of SPI (Kallio et al. 2008; Park et al. 2015). Both of these were conducted in children aged more than three years of age. Since heart rate is an age-dependent parameter (Daymont et al. 2015) and consists of one third of the SPI calculation (Huiku et al. 2007; Struys et al. 2007) evidence gained from adult studies cannot be directly adapted to small children. So far no studies on small children are available.

2.5.2 Surgical Pleth Index – Clinical studies

The clinical evidence on the use of SPI is still limited and partly contradictory. SPI guidance has resulted in lower remifentanil consumption, more stable hemodynamics, lower incidence of unwanted events and faster recovery (Bergmann et al. 2013; Chen et al. 2010; Colombo et al. 2015). In contrast, under sufentanil anesthesia SPI failed to have any major effect on the measured end points (Gruenewald, Willms et al. 2014). An interesting finding was also reported with oxycodone in patients undergoing elective thyroidectomy, where oxycodone consumption and extubation time were lower in SPI-guided group (Won et al. 2016). In pediatric population, the evidence is also scarce. In a study by Park et al. (2015) the consumption of anesthetics was reduced with SPI guidance. By contrast, postoperatively the number of unwanted effects, such as nausea and vomiting, was increased and interpreted as a failure of the index (Park et al. 2015). This was interpreted as a failure of the SPI, but a more intensive prevention of nausea and emergence delirium using multiple means could have reduced the incidence of these unwanted events in both groups.! In conclusion, there is mounting evidence that SPI is able to detect nociception and is a potential tool to guide opioid administration under general anesthesia. Evidence from multicenter studies, unstable patients and also small children is still lacking.

Table 1. Devices indicated for the measurement of nociception.

ss ,; onds Jr et al. al. et Jr onds Mathews et al. et al. Mathews

;

Edm

Sabourdin et al. 2013; 2013; al. et Sabourdin Dedrie et al. 2016; De 2016; al. et Dedrie

References

Andrews et al. 1995; Wilson1993 1995; al. et Andrews Upton et al 2017, 2017, al Uptonet Ellerkmann et al. 2013 al. et Ellerkmann

Israel et al. 2013; Edry et al. 2016; Martini et Martini 2016; al. et Edry 2013; al. et Israel Takamatsu et al. 2006; Valjus et al. 2006; 2006; et al. Valjus 2006; et al. Takamatsu - (Cividjian et al. 2007; Martinez et al. 2010; Rossi Rossi 2010; al. et Martinez 2007; al. et (Cividjian 2012) al. et (Ben 2015) al. ABM: 1986; al. et Jr Edmonds 1985; CVI: et 2014; al. Sahinovic 2012; RE: 2005; al. et Wheeler et Boselli 2014; al. et Boselli 2013; al. et (Boselli Broucqsault 2015; al. 2015; al. etGall 2012; al. et Jonckheere Je 2014; al. et Jeanne 2015; al. et Gruenewald al. et Ledowski 2012; al. et Guen Le 2016; al. et 2014; 2015) al. et Szental

after fentanyl fentanyl after

: No signs of effect vs. vs. effect of No signs :

operative room operative - Notions

at post at

pain pain

guided. -

Decreased

Predicted intraoperative movements. Reduction of movements during movements of Reduction movements. intraoperative Predicted device. Finapres Uses colonoscopy. on No studies on nociception. react to Seems commercially. Notavailable NoL. with guided anesthesia the affects amnesiaof levelThe noxioustostimuli. Reactivity RE studies. comparative no CVI: measurement. to when compared reactions or movements in anesthesia standard BISor on based analyzed EEG and EMG of Combination esmolol. electrodes. Frontal measurement Entropy andemotion Stress patients. pediatric and onawake evidence Some than nociception, of prediction better on Someevidence affect. might traditional than notbetter morphine With measurements. traditional guidance. ANI administration as as

Skin

Used technique Used Changes in heart rate. heart in Changes cuff Finger HR,PPG HRvariability, amplitude, Skin conductance, fluctuations conductance combined theof facial reactivity The to areaction as muscles nociception. theof HR variability The nociceptive reactionto electrodes. ECG stimuli.

activity activity Device Method or or Method Devices indicated for the measurement of nociception. CARDEAN Nociception (NoL) level and Anesthesia brain (ABM) monitor Composite index variability (CVI), Response (RE) entropy Analgesia nociception index (PhysioDoloris)

phenomena Physiologic Physiologic Cardiac Cardiac baroreflex Combining multiple reactions FacialEMG rate Heart variability Table 1.

Sabourdin Sabourdin

2013;

lminotti et al. al. et lminotti

References ark et al. 2015) al. et ark

)

(Abad Torrent et al. 2015; Aissou et al. 2012; 2012; al. et Aissou 2015; al. et (AbadTorrent 2011; al. et 2010;Ledowski al. et (Ledowski 2010; al. et Chen 2013; al. et (Bergmann (von Dincklage et al. 2009; von Dincklage, Dincklage, von 2009; al. et Dincklage (von et Velten Dincklage, von 2010; al. et Hackbarth 2010) al. 2006; al. et Constant 2014; al. et Connelly Guglie 2013; al. et Guglielminotti 2014; al. et Kantor 2013; al. et 2015;Isnardon 2013; al. et Migeon Behrends2015; & Larson 2013 al. et Rouche 2013; al. et Paulus 2017 alet &Gray Strehle 2015; al. et Solana 2016) al. et derLee van 2012; al. et Valkenburg et Ilies 2007; al. et Huiku 2014; al. et Gruenewald P 2010; al.

olatile tion and morphine morphine and tion operative pain operative - available commercially. available

sensitive to stimuli than HR or blood blood thanor HR stimuli to sensitive Notions

Reduced remifentanil consump remifentanil Reduced

remifentanil anesthesia. Controversial results with v with results Controversial anesthesia. remifentanil - ve of pain. Size of electrodes affects the readings. Poor Poor the readings. affects electrodes of Size pain. of ve Prediction of movement during anesthesia. Not anesthesia. during movement of Prediction VAS postoperative on findings Controversial function. brainstem Assesses on ICU anesthesia, general under stimuli nociceptive to Reacts scores. dilatation. cervix during movement predict to Able patients. awake andon More use. continuous for Notsuitable post reduced pupillometry guided Anesthesia pressure. consumption andopioid pain. chronic in reduction requirement. on Tested affect. may system nervous sympathetic affecting Drugs patients. awake on scores pain with Somecorrelation patients. sedated wake at Reactions the reaction. precludes block Neuromuscular irrespecti hormones. stress to correlation associated effects side Reportson reduced on reactivity. reports Several propofol with function analysis, in ECG Uses opioids. andother anesthetics useon or pacemaker arrhythmias, cardiac with compromised drugs. anticholinergic

glands and skin andskin glands

Used technique Used Electrodes on femoral femoral on Electrodes muscle measuring region contraction in change the Evaluates as diameter pupillary nociception. to response scanner. Handheld of sympathetic Activity affecting system nervous sweat Skin conductivity. electrodes. RRInorm and PPGAnorm indicating onenumber as or clip Finger nociception. sensor. adhesive an

, ex ex

® , ®

® hold)

Storm Storm s - Device Method or or Method NFTS NFTS (Nociceptive refl Flexion Thre Neurolight AlgiScan Med Monitoring Pain system Pleth Surgical Index

reflex reflex -

phenomena Physiologic Physiologic Normalized Normalized RIII threshold Pupillometry Skin conductance of Reactivity andPPG pulse

2.6 Measurement of blood pressure

Blood pressure has conventionally been measured by the inflatable cuff first developed by Riva-Rocci in 1896. A commonly used variation of the technique for rapid measurement is “return-to-flow technique”, where the point when a palpable pulse reappears after cuff deflation marks the systolic blood pressure. In order to measure diastolic pressure, the technique by Nikolai Korotkoff from 1905 is the most commonly used manual method. Audible sounds occur at the distal portion of the artery as the cuff is deflated. The discovery of the sounds resulted in more accurate determination of systolic and diastolic artery pressure (Schordes et al. 2010). H. von Recklinghausen in 1931 described oscillometric changes in blood pressure, leading to the development of automatic blood pressure devices in 1970s, which are now considered standard for noninvasive blood pressure measurement (Sorvoja 2006). While the technology relies on the inflatable cuff occluding the artery, there are significant restrictions. Changes in limb position in relation to heart level may cause changes in blood pressure (Mourad et al. 2003). Prolonged measurement may also cause nerve and skin damage, pain, petechiae, limb edema, venous stasis and compartment syndrome. The method is therefore restricted to a minimum interval of about 3-5 minutes between measurements (Bause et al. 1986; Lin et al. 2001). Quite often the measurement interval is lengthened in stable conditions in order to relieve the patient’s discomfort. Thus rapid changes in blood pressure may go unnoticed. Direct measurement of arterial pressure with invasive arterial cannula, continuous pressure transduction, and waveform display is the reference gold standard for invasive blood pressure monitoring. Several reasons, such as real- time measurement and the option to draw blood samples make it superior to non-invasive measurement (Cockings et al. 1993). The use of an invasive bloodline as such is not without problems. In particular, the use of invasive cannula may cause significant nerve and vascular damage and also subject the patient to a risk of air or thromboembolism (Singleton et al. 1993; Slogoff et al. 1983). Furthermore, technical issues such as underdamping and resonance may seriously impair the accuracy of the arterial line, while these may go unnoticed (Romagnoli et al. 2014). The invasive and non-invasive oscillometric method may also show significantly different readings in critically ill patients or under general anesthesia (Kim et al. 2014; Meng et al. 2013; Ribezzo et al. 2014).

The current standard practices both have their limitations and there is a continuous need to develop improvements. The following sections briefly introduce current technologies in the search for continuous or near-continuous blood pressure measurement.

2.6.1 Applanation tonometry

Arterial applanation tonometry is based on a sensitive pressure sensor placed non-invasively on an easily accessible peripheral artery with sufficient bony support, typically on the radial artery. The sensor may be either fixed at one place or moved continuously in order to achieve the best pressure signal. The force applied is sufficient to partially flatten the artery, but not to occlude it. The measurement can therefore be obtained comfortably and continuously or almost continuously. The pressure recordings are transformed into blood pressure readings using artifact cancellation (Sato et al. 1993; Sorvoja 2006). Currently there are at least three commercially available monitors using applanation technology. The Tensys® T-Line® (Tensys Medical, San Diego, CA, USA) is a distal forearm-based monitor for continuous blood pressure measurement. The sensor is self-calibrating and locates and tracks the best arterial signal using an electromechanical system moving the sensor constantly. In addition, the distal forearm is fixed to an extension in order to stabilize the artery (Janelle & Gravenstein2006). The sensor has been compared to invasive arterial pressure in several studies with a reported clinically acceptable accuracy of 1.4 to 6 mmHg difference in systolic pressure, although the precision for systolic pressure especially has been variable (Langwieser et al. 2015; Meidert et al. 2014; Saugel et al. 2013). However, the mechanical system is energy consuming and so far the system has not been portable. Moreover, in the studies the distal forearm has been fixed in one position, thus compromising the degree of comfort. The second device using modified applanation tonometry is BPro® by Healthstats International (Singapore). BPro® is a watch-like portable device with a mounting strap to secure it in position on the distal forearm allowing moderately free movement of the distal forearm (Figure 2). A pulse-sensing module is pressed against the arterial wall to detect pressure changes (Ng et al. 2004). An array of waveforms is collected and used for blood pressure calculation (Nair et al. 2008). The initial results by Nair compared the device to

a non-automated non-invasive sphygmanometer and reported very good accuracy (1.3 mmHg (SD = 5.55 mmHg)). Similar results were found in a study on patients with diabetes (Theilade et al. 2012). When the artery is not fixed the risk of unsuccessful measurements rises. This was demonstrated in a study with patients undergoing ambulatory 24h blood pressure monitoring. The proportion of successful measurements of all measurements was only 51%. Likewise the difference between the non-invasive brachial blood pressure device and the BPro® device was greater than in earlier studies (mean difference in systolic pressure (-4.5 mmHg, SD = 9.2 mmHg)) (Komori et al. 2013). A high correlation and low bias were also found in a study comparing BPro® measurements to invasive measurement during cardiac catheterization (Ott et al. 2012). Furthermore, the accuracy was reported to be low with a standard deviation of 13 mmHg vs. invasive central systolic pressure. No studies have addressed the use of the device in cases of critical illness or vasoactive medication. Sphygmocor® (Atcor Medical Pty Ltd, Sydney, Australia) provides a point measurement of blood pressure where a pen-like sensor and a display are used to obtain pressure and additional indexes. In this device the tonometric sensor is used to measure pulse wave velocity rather than actual pressure. The blood blessure is calculated from the velocity of the pulse wave. As for the other tonometry sensors, calibration pressure is needed before reliable measurements can be obtained. This means that the type of measurement makes the device incapable of continuous measurement. As with other devices using tonometry, the reliability has been variable with mean differences from non-invasive oscillometric measurement ranging from 2.8 mmHg (Ding et al. 2011) to 15 mmHg (Sarafidis et al. 2014).

Figure 2. An illustration of a sensitive pressure sensor used by BPro. Modified from a picture by Ng et al. (2004).

2.6.2 Finger cuff method

Penaz first described the finger cuff or volume clamp method in 1967. The cuff around the finger is pressurized to the same pressure as the artery. Thereafter the changes inblood pressure can be detected using a light source and a detector (Sorvoja 2006). The device is auto-calibrating and changes in arterial diameter and volume lead to loss of the loaded volume and a re-calibration is initiated. The pressure discomfort on the fingers can be reduced using two to three cuffs alternating in measurement (Schmid et al. 2013). A re-calibration may be needed after changes in body position (Ilies, Bauer et al. 2012). There are three commercially available devices using the technology. The oldest device, called Finapres (Finapres Medical Systems, the Netherlands) has shown a very high precision and accuracy in various settings (Silke & McAuley 1998). The same technology is utilized with the brand names Portapres for ambulatory monitoring and Finometer as a standalone monitor. A similar device with the brand name Clearsight (previously known as Nexfin, Edvard Life Sciences, CA, USA) has also shown fairly small bias and reasonable precision. However, it has not reached sufficient accuracy in blood pressure measurement to be interchangeable with invasive measurement, especially in severe hypotension, reduced perfusion or hypothermia (Ameloot et al. 2015).

The CNAP (CNSystems Medizinteknik AG) combines the vascular unloading technique and the VERIFI algorithm. The algorithm inspects waveforms in order to eliminate the vasomotor effect on blood pressure and to enhance the accuracy (Fortin et al. 2013). The device has shown good accuracy in critically ill patients (Smolle et al. 2015), thus the evidence is contradictory (Ilies et al. 2015). The technology uses the fingers as a measurement site. This limits the usefulness of the device for patients when outside operating room or other close follow up, where patients typically remove unpleasant devices. The evidence on critically ill patients is also very limited. In one study the correlation with invasive blood pressure measurements was poor when the fingers were swollen or when patients received norepinephrine infusions. The bias and precision were not so much affected (Hohn et al. 2013). The results may be also limited under severe vasoconstriction (Ameloot et al. 2015).

2.6.3 Continuous blood pressure measurement using pulse transit time

Pulse wave velocity is the speed of the pulse wave in the arteries, whereas pulse transit time refers to the time delay for the pressure wave to travel between two arterial sites within the same cardiac cycle (Mukkamala et al. 2015). The blood pressure is determined using the Moens-Korteweg equation, where thickness of the arterial wall, diameter of the artery, density of the blood and elasticity of the wall affect the measurement. In the equation the blood density is considered constant, while other parameters are constantly measured or calculated (Oksala et al. 2015; Sorvoja 2006). The pulse transit time is known to reflect the changes in blood pressure (Callaghan et al. 1984; Pruett et al. 1988), but several factors such as atherosclerosis, movement, and age affect the measurement (Schiffrin2004; Yamashina et al. 2003). A novel technology using one-point calibration with the cuff-based method has been proposed as a solution. The pulse transit time is measured at the finger using electrocardiogram leads and plethysmography (Gesche et al. 2012). The method is still poorly validated with some evidence showing inaccuracy in the method (Keehn et al. 2014; Patzak et al. 2015). Other approaches have also failed to show acceptable accuracy (Ruiz- Rodriguez et al. 2013). In the applications using ECG wave, the initiation of the systole is defined as the time of the R-wave in the ECG. Thus, the R-wave is not precisely simultaneous with mechanical systolic contraction, which is

36 preceeded by a varying pre-ejection time. This time may vary significantly depending on the patient’s age, height, vessel properties, and state of sympathic activation. This adds a varying error to the measurement of pulse transit time (Hennig & Patzak 2013; Mukkamala et al. 2015). One possible way to prevent the error caused by pre-ejection time could be to measure pulse transit time at two different sites in the same limb, but the correlation with blood pressures has been variable (Douniama et al. 2009). In theory, pulse transit time would offer a very easy way to measure blood pressure with several technologies such as PPG.

2.7 Measurement of saturation

Pulse oximetry, a standard noninvasive technique to monitor arterial hemoglobin saturation (SpO2) is routinely used in all patient monitoring. It has long been a basis for estimating patient oxygenation non-invasively. Typically, two different wavelengths of light are used and the saturation is calculated based on the difference in the absorption of light through the tissue. Two common types of sensor configurations are used; transmission mode and reflectance mode. The transmission mode sensor consists of an optical emitter and detector positioned on opposing surfaces. It can be used in multiple places, where the light can travel through the tissue such as the finger, ear lobe, toe, nose or infant palm. In the reflectance mode, the emitter and detector are located side by side while typically the red and infrared light used is reflected back from the tissue (Agashe et al. 2006) (Figure 3). Even though the use of oximeters is unquestionable and widespread, there still is no evidence showing any benefit on mortality, or reduction in transfers to the intensive care unit (Pedersen et al. 2014). The transfer of oxygen to and from hemoglobin is defined with the oxygen- hemoglobin dissociation curve. The shape of the curve is sigmoid, where a relatively stable phase when the oxygen is tightly bound by the hemoglobin, is followed by a very rapid decline in the values after an initial cut-off point in oxygen saturation measurement (Gomez-Cambronero 2001). Each hemoglobin molecule binds four molecules of oxygen. The phase when all hemoglobin is fully saturated corresponds to a saturation value SaO2= 100. After that point, the partial pressure of oxygen may still rise without it being visible in saturation measurement (Collins et al. 2015).

Most manufacturers claim an accuracy of 2% for measurement of saturation defined by average root mean square error, although the accuracy found in clinical studies corresponds to 3-4% when compared to arterial oxygen saturation (Nitzan et al. 2014). Typically SpO2 is able to detect a sudden 3-4% drop in saturation levels. This is accepted as a reliable parameter for the detection of a significant deterioration in respiratory function. In critical illness, the correlation between arterial saturation and SpO2 may be even poorer, suggesting a higher acceptable level for ventilated patients (Perkins et al. 2003). In addition to low oxygen saturation, the accuracy of saturation measurement may be compromised by multiple factors, such as dyshemoglobinemia, dyes (injected intravenously), low perfusion state, skin pigmentation, anemia, nail polish, hypothermia, and motion artifacts which may severely affect the readings. Thus interpretation of the plethysmographic waveform may help to outrule these changes (Jubran 2015). An interesting new device patented by Oxitone Medical uses two laser light sources and they claim that it reaches an accuracy of 2% in average root mean square error when measuring saturation at the wrist (Eisen et al. 2013). However, there are currently no peer-reviewed publications on the technology.

Figure 3. An illustration of the difference between reflectance and transmissive oximetry.

2.7.1 Transmission photoplethysmography

Saturation is typically measured by detecting the amount of light absorbed through the tissue. This has long been the standard for saturation measurement. The most typical places for measurement are sites where light can travel easily through the tissue, such as the fingers, ear lobes or the palm in infants (Nitzan et al. 2014). The wrist and ankle also seem to be reliable sites for measurement in infants (Phattraprayoon et al. 2012; Safar & El-Dash 2015). The long distance between sensors prevents the use of the wrist as a measurement site in adults.

2.7.2 Reflectance photoplethysmography

Reflectance photolethysmography has attracted more attention with the development of forehead saturation sensors. Earlier studies have reported conflicting performance of the devices and transmission-based methods became more interesting (Clayton et al. 1991; Palve 1992a; Palve 1992b). However, the accuracy seems to depend on the thickness of the tissue (Aoyagi et al. 2007) and a multi-wavelength technique was recommended to solve the problem. Other potential sites studied for using reflection type sensors include the ear canal (Budidha & Kyriacou2014; Budidha & Kyriacou 015), the nose (Rosenberg & Pedersen 1990) and the sternum (Sola et al. 2007), where the distance to the bone is short. The myoglobin in the muscle tissue has the same light absorption spectrum as hemoglobin, but the oxygen level stays higher in myoglobin than in hemoglobin. Measuring from muscle would then cause a different level of saturation than measuring from a site with less myoglobin interaction (Pellicer & Bravo 2011; Spires et al. 2011). This limits the potential sites for saturation measurement. Yet recently the use of forehead sensors and a headband has gained popularity showing better results in critically ill patients when compared to finger measurements (Nesseler et al. 2012; Schallom et al. 2007). The use of two-wavelength photoplethysmography is also under investigation for measuring tissue oxygenation. A study with a reflectance sensor at the proximal forearm suggested lower saturation values when compared to finger readings. The values were even lower when measured with near-infrared technology (Abay & Kyriacou 2015). Furthermore, a study at the forearm comparing reflectance measurement to finger measurement showed a mean difference of

2.47% (SD= 1.7%) in healthy adults, suggesting good performance (Mendelson & McGinn 1991) with similar findings for neonates (Fanconi & Tschupp 1994).

2.7.3 Near-infrared spectroscopy

A modification of reflectance photoplethysmography is called near-infrared spectroscopy (NIRS). The sensors use multiple wavelengths ranging typically from 650 nm to 900 nm measuring tissue oxygen saturation (StO2). Most often several wavelengths are used to estimate oxygen saturation in peripheral tissue. The proportion of venous blood in muscle tissue is typically 75% and therefore the measurements may differ from arterial oxygen saturation (Epstein & Haghenbeck 2014; Hyttel-Sorensen et al. 2014). Different devices also produce different absolute values at the forearm (Hyttel-Sorensen et al. 2011). The most typical site for the light probe is on the thenar muscle with an adhesive sensor (Epstein & Haghenbeck 2014) but other sites such as the soleus muscle have also been used (Stone et al. 2016). NIRS technology is typically used to estimate brain or tissue oxygen saturation. In one study, thenar muscle StO2 measured by an NIRS device predicted mortality in critically ill patients (Donati et al. 2016). It has been hypothesized that StO2 could be used to predict deterioration in tissue oxygenation (Lima et al. 2009). In addition, NIRS devices are used to measure brain oxygenation when placed on the forehead in adults (Davies et al. 2016) or infants (Vesoulis et al. 2016). This aims to prevent brain hypoxia under various circumstances such as during reverse trendelenburg position during abdominal surgery (Sørensen et al. 2016). Thus more information is needed before wider adoption of the technology, especially when estimating tissue hypoxia.

2.8 Measurement of heart rate

Heart rate is one of the basic vital signs and over the decades various reliable methods have been developed to measure it accurately. It can be obtained from electrical activity using ECG leads, variations in pulse by invasive arterial line or plethysmography wave (Bartels & Thiele 2015). There is also research on wearable sensors capable of producing a reliable signal for HR measurement (Khan et al. 2016). HR has long been available for Holter 24 -hour registrations on devices based on an ECG signal (Wimmer et al. 2013) and specific versions for athletes and devices intended for sports with multiple techniques have been developed (Fukushima et al. 2012; Li et al. 2016; Losa-Iglesias et al. 2014; Zhang et al. 2015). In addition, heart rate variability has been studied as one option to measure nociception (Luginbuhl et al. 2007). Multiple locations have been found to offer an appropriate signal for heart rate measurement (Nilsson et al. 2007). For this reason, it is feasible that heart rate can be measured with current technology from several locations while the demand for other vital signs is more limited for the technology used and site of measurement. The latest trend in heart rate monitoring is in the development of mobile devices with wrist based optical heart rate monitoring. The devices typically use green light, which has a shorter penetration length into the tissue. Small penetration into the tissue is associated with a smaller area of illumination, which is considered to cause fewer artifacts in association with movement (Delgado-Gonzalo et al. 2015; Maeda et al. 2011b; Matsumura et al. 2014; C. Zhou et al. 2016). Most of these devices are not capable of beat-to-beat measurement of heart rate, but some reports on their accuracy have been presented (Parak et al. 2015). The drawback for more superficial measurement is the possible effect of peripheral vasoconstriction. The temperature of the skin may seriously affect the measurement (Maeda et al. 2011a). Currently there is not enough evidence to support the use of green light sensors in hypovolemic or hypothermic patients.

2.9 Measurement of respiratory rate

Adequate respiratory activity is among the basic vital signs that might be disturbed during critical illness or by medication. Respiratory failure can build up gradually or become life threatening within minutes. Therefore the ability to adequately measure it should be obvious (Folke et al. 2003). A number of techniques and devices have been developed to adequately measure respiratory rate (RR), some of which are described in Table 2 on page 45, but convincing evidence for the accuracy of most of the devices is still lacking. Measurements from nose or mouth are also often too uncomfortable to be measured continuously. The traditional way of measurement by an observer counting the breaths is time consuming and may involve significant inter-observer variability and is thus not convenient for continuous measurement (Edmonds et al. 2002).

2.9.1 Estimation of respiratory rate using plethysmographic waveform

Several authors have recently reported on the use of PPG signal in order to determine respiratory rate. Respiratory cycles cause multiple changes in the plethysmographic waveform, which can be used to calculate respiratory rate. The changes seen in the plethysmographic waveform can be classified into four kinds of modulations. First, the baseline modulation that is considered to be caused by cyclic fills and drains of the venous bed. Respiratory cycles change the intra-thoracic pressure thereby changing the central venous pressure causing the phenomenon. Second, intra-thoracic pressure is also known to alter cardiac function. The pressures inside the thorax are changed in relation to the phase of respiration changing the stroke volume. This can be seen in changes in pulse amplitude and is called pulse amplitude modulation. Third, a cyclic modulation in heart rate is seen where HR increases during inspiration and decreases during expiration. This is called respiratory sinus arrhythmia (Addison, Wang et al. 2015). Last, the waveform amplitude increases and broadens in accordance with respiratory rate (Oksala & Liuhanen 2015). The changes are described schematically in Figure 4.

Figure 4. A schematic presentation for the change of plethysmographic waveform caused by respiration. a: the baseline modulation, b1 to b2: the amplitude modulation, c1 to c2: respiratory sinus arrhythmia and d1 to d2: waveform modulation.

Although the changes in the waveform are clear, several confounding factors can affect the measurement of RR. The change in limb position in relation to the heart causes changes in the baseline of the PPG waveform. This is caused by hydrostatic pressure affecting the veins (Reisner 2008). Cardiac arrhythmia, especially atrial fibrillation, may cause much more severe changes in pulse waveform than in respiratory rate. Vasoconstriction and vasodilatation caused by pharmacological or physiological confounders can change the amplitude. Also, decreased volume load causes pulse variations. These changes demand strong noise and artifact detection and error removal in the algorithms proposed for RR measurement (Alian & Shelley 2014; Karlen et al. 2014). The relative power of the frequencies corresponding to the heart rate in the power spectrum describing the potential to measure respiratory rate seems to vary across locations. In one study, the forearm was reported to be the most suitable for RR detection. The relative power for the forearm was 42%, whereas for the wrist over the radial artery it was 13% and for the finger 3% (Nilsson et al. 2007). Other studies have likewise reported good sensitivity and specificity rates for respiratory rate detection (Nilsson et al. 2005). Most of the studies conducted use respiratory-induced intensity variations (Leonard et al. 2006; Lindberg et al. 1992; Nilsson et al. 2006) but reports on pulse transit time have also been presented (Johansson et al. 2006). Unfortunately the plethysmographic technique does not detect absence of airflow from the lungs,

thus central apnea and obstructive apnea may go unnoticed. Hence more development work on the technology or combining other sensors such as saturation measurement is needed for adequate monitoring of respiration. Respiration is known to cause several changes in the plethysmographic waveform. The combination of all known changes, simultaneous continuous comparison for disturbances in the measurement, and selection of least disturbed signals might theoretically be able to improve the accuracy of the measurement. This has been done in a simplified way in the algorithm patented by Oksala & Liuhanen (2015).

Table 2. Methods evaluating respiratory rate

Silveira et al. al. et Silveira -

References

(Deitch et al. 2010; Gaucher (Atkins & Mandel2014; Krauss2012; & al. et Hess2007) (Hers et al. 2013) (Addison et al. 2015; Garde 2014) al. et et Tanaka 2013; al. et Ramsay 2014) al. (Hernandez 2005); al. et Lovett 2015; laryngeal

Notions

to use; Continuous, can reduce hypoxic events during sedation; sedation; during events hypoxic reduce can Continuous, annoyance Physical airway measure to Unable Comfort; Continuous; Stationary obstruction; easy Continuous; to Unable parameters; other with combined be Can obstruction airway measure under Studied airflow; Measures Continuous; surgically post and sedation mask, Readily Cheap; Continuous; comfort; use; to Easy Low obstruction; measureairway Unable to available; tachypnea and bradypnoea of detection

parameter

concentrationof Measured CO2 CO2 expiredair from movement Breathing the sensors placed on bed return Venous air of Flow wall thoracic of Distension

Methods evaluating respiratory rate. respiratory evaluating Methods tidal CO2 -

Method/device

End measurement Infrared Passive technology Photoplethysmography detection Sound Transthoracic impedance Table 2 Table 45

2.10 Measurement of fluid responsiveness

Passive leg raising causes a rapid increase in venous return and can be used to mimic rapid change in fluid load. It has been shown to cause changes in plethysmographic waveform amplitude, which could potentially be used as an indicator for vascular load (Delerme et al. 2007). There have been several promising reports to develop a tool for guiding fluid administration based on plethysmography changes (Cannesson 2007; Feissel et al. 2007; Natalini et al. 2006). A commercial solution known as the Pleth Variability Index (PVI®, Masimo, Irvine, USA) tracks dynamic changes in the amplitude of the pulse oximeter waveform taking note of both pulsatile and nonpulsatile components of the waveform. Several studies have investigated the Pleth Variability Index under different clinical situations. The fluid load and serum lactate levels were lower during abdominal surgery (Forget et al. 2010; Yu et al. 2015) and the PVI could predict fluid responsiveness (Zimmermann et al. 2010). Similar findings have also been reported among pediatric patients undergoing neurosurgery (Byon et al. 2013) and in patients after cardiopulmonary bypass (Haas et al. 2012). The reactivity has also been shown during intensive care during mechanical ventilation (Loupec et al. 2011) and the device was also able to predict fluid responsiveness (Feissel et al. 2013). The device is appropriate for measurement at the fingers, but forehead and ear lobe have likewise shown good sensitivity and specificity (Desgranges et al. 2011; Fischer et al. 2015). The cardiac preload and thereafter hemodynamic status may change for several reasons. A false elevation in the PVI was seen in association with insufflation during laparoscopic surgery (DeBarros et al. 2015). Similarly, when the breathing is not controlled, the index may be severely disturbed (Sun & Huang 2014; Wrench et al. 2015; Yokose et al. 2015). Vasomotor tone influences the vascular reactivity in the fingers and medications such as noradrenaline have been shown to alter the reaction in PVI (Biais et al. 2011; Landsverk 2008). These factors suggest that there are major restrictions in the use of PVI. On the other hand, plethysmography index is greatly influenced by volume load, which may severely impair its usefulness for other parameters such as respiratory rate or detection of nociception.

2.11 Technologies based on multiple wave measurement

Recent advances in plethysmography technology have concerned using multiple wave technology. Masimo rainbow SET® technology (Masimo Corporation, Irvine, USA) uses a combination of eight wavelengths to measure hemoglobin (Barker et al. 2016; Frasca et al. 2015; Frasca et al. 2011; Galvagno Jr et al. 2015; Macknet et al. 2010; Miller et al. 2012; Patino et al. 2014), methemoglobin (Annabi & Barker 2009; Barker et al. 2006; Soeding et al. 2010), carboxyhemoglobin (Feiner et al. 2013; Hampson2012; Roth et al. 2011) as well as Oxygen Reserve Index (ORI) (Applegate et al. 2016; Simpao & Gálvez 2016; Szmuk et al. 2016). The accuracy of all these measurements is fairly good, but not exactly the same as with invasive measurements, which are still the standard. However, the measurement technology is continuous, non-invasive, and easily accessible. One possible way to monitor changes in hemoglobin and the prediction ability of lowering oxygen concentration with oxygen reserve index might be extremely useful in clinical practice. It seems evident that the potential of peripheral pulse wave could be better used with more widespread adoption of multiple wavelength technology.

2.12 Summary of the theoretical background

The ability to measure physiological changes in the body does not necessarily mean that it is useful or adds clinically necessary information. As described previously, several devices aiming to adequately measure nociception do indeed exist, but the evidence of their clinical significance is still scarce. Reactivity of SPI is fairly well known in adults while the evidence in small children is very limited. Also, multicenter evidence on the clinical usefulness in varying patient scenarios is lacking. This would prove the concept of measuring the components of anesthesia and guiding the delivery of anesthesia accordingly. Equally evident is the potential of the peripheral pulse wave to provide clinically significant information on clinical signs. Most of the evidence concerns measurement at the fingers using two or more wavelengths of light. Blood pressure can likewise be measured relatively accurately from the fingers. However, this approach is severely challenged when the patients are not under constant surveillance. Measurement at the fingers prevents movement and causes discomfort making the patients typically remove the sensor. The

measurements also show lower accuracy associated with movement. The distal forearm is a potentially convenient measurement site. Furthermore, there is some evidence that basic vital signs could be measured from the wrist region. This warrants further clinical investigation.

3 Aims of the study

The aim of the present study was to explore ways to utilize peripheral pulse wave in different clinical situations. The purpose in the publications was:

1. To study whether the concept of adequacy of anesthesia using measurements of amnesia, nociception, and muscle relaxation can reduce the number of unwanted events and improve postoperative care (I). 2. To study whether SPI can be used to detect nociception in small children (II). 3. To compare the performance of modified arterial applanation tonometry to that of invasive blood pressure measurement during postoperative care (III). 4. To determine whether a readily available photoplethysmography sensor can be used to detect saturation at the distal forearm (IV). 5. To determine whether a readily available photoplethysmography sensor can be used to detect heart rate at the distal forearm (IV) 6. To determine whether a readily available photoplethysmography sensor can be used to detect respiratory rate at the distal forearm (IV).

4 Patients and methods

The studies were divided into four publications based on three clinical studies. The allocation of the studies is summarized in Table 3.

Table 3. Summary of the study plan Study n Groups Randomization Double- blind I 250 Standard monitoring Yes No 250 Entropy & SPI

II 15 Block group Yes Yes 15 Saline group III 30 Observation No No IV 30 Observation No No

4.1 Patients

Study I: Patients undergoing elective surgery under general anesthesia with endotracheal intubation with pre-estimated duration of at least two hours were enrolled in the study. Inclusion criteria were age 18-80 years and ability to provide written informed consent. The exclusion criteria were patients with cardiac pacemakers, subjects with atrial fibrillation, more than five ventricular extra systoles/minute, need for invasive blood pressure measurement, hemodynamics that would be qualified as inadequate anesthesia at the time of baseline measurement and need for prone position or epidural anesthesia during surgery, body mass index > 35, known allergies to anesthetic agents used, high-risk surgery with expected extensive blood loss or known chronic use of opioids. Study II: Patients subjected to elective inguinal hernia repair or open correction of undescended testicle, aged less than two years of age, ASA

50 classification I-III. Patients with cardiac problems or known ECG disturbances were excluded. Studies III-IV: Patients undergoing elective surgery requiring invasive blood pressure measurement and postoperative treatment at the post-anesthesia care unit. Patients with implanted cardiac pacemakers were excluded.

4.2 Methods

4.2.1 Study designs

Study I was designed to evaluate the usefulness of the adequacy of anesthesia monitoring using Entropy and SPI when compared to standard monitoring with non-invasive blood pressure, ECG and pulse rate monitoring. The study was a prospective, randomized, international, multi-center, single-blinded controlled trial. The study was initially designed to be conducted in five different university hospitals, each allocating one hundred patients. However, due to unsolved problems in obtaining approval from the ethics committee at one site (Göttingen, Germany), the study was carried out as a four-center study (Tampere, Finland; Amsterdam, Netherlands; Szeged, Hungary and Kiel, Germany) with a total expected number of 500 patients. Each site agreed to recruit at least one hundred patients and the remaining one hundred were divided between the sites so that the recruitment could be completed as fast as possible. The patients were randomized into one of two groups. In one group the anesthesia was guided using standard monitoring (control group) and in the other standard monitoring was supplemented with a measurement of depth of anesthesia (Entropy) and nociception (SPI)(test group). The patients were anesthetized with target-controlled infusions of remifentanil (Minto, effect site) and propofol (Marsh, effect site). In the test group, after beginning of surgery PECprop was adjusted by 0.5 µg/ml steps to maintain state entropy between 40 and 60, while PECremi was adjusted by 1 ng/ml steps in order to maintain SPI values within a predefined range from 20 to 50. Absolute limits for anesthetics were 2-10 µg/ml for PECprop and 2-15 ng/ml for PECremi. In the control group propofol and remifentanil levels were adjusted according to standard clinical practice alone. The patients were followed for predefined signs of inadequate anesthesia during surgery as described in Table 4. During post-anesthesia care a modified Aldrete score was

recorded every thirty minutes until the patient was discharged to the ward. One day after surgery a post-operative questionnaire was completed with the patients.

Table 4. Criteria for inadequate anesthesia, hypotension or bradycardia (Chen et al. 2010). The duration of the change needed to be clinically relevant to be counted. For heart rate, duration of tachycardia or bradycardia had to be >30s and for non-invasive blood pressure (NIBP), two consecutive measurements (three minutes pause between measurements).

Inadequate Definition anesthesia Hypertension Mean blood pressure (MAP) >120% from baseline or MAP ≥100 mmHg Tachycardia Heart rate >100 beats/min Somatic arousal Coughing, chewing, grimacing, breathing against respirator Somatic response Purposeful movement Hypotension MAP< 80% of baseline or MAP < 60 mmHg Bradycardia Heart rate (HR) < 80% from baseline or HR <45 beats/min

Study II was designed to test reactivity of the SPI in small children. Patients were randomized and double-blinded into two groups. One group received ilioinguinal/ iliohypogastric nerve block immediately after anesthesia induction and an injection of saline at the same region after the surgery and before extubation. The other group received injections in the opposite order. The personnel were blind to the order of the injections by using pre-numbered syringes for the injections. SPI was recorded electronically and the personnel continued to be blinded throughout the surgery. The points of interest were analyzed post hoc. Studies III-IV were designed to test heart rate, saturation, respiratory rate, and blood pressure measurements using a watch-like study device. The device consisted of Nellcor™ OxiMask MAX-FAST plethysmography sensor, accelerometer, infrared temperature sensor, and BPro® applanation tonometry sensor embedded on a single device mounted around the distal forearm. The measurements were compared to standard measurement with a Carescape B650 patient monitor (GE Healthcare, Helsinki, Finland) equipped with oxygen saturation measurement and placed on the finger, ECG, and invasive blood pressure measurement. Similarly, non-invasive blood pressure was recorded

52 once at the beginning and end of each measurement section. The study device was placed on the distal forearm opposite to the standard monitoring. All patients were monitored during postoperative care for two hours and the study personnel verified the accuracy of the data and sensor placement throughout the study period. Figure 5 presents a picture of the study device.

Figure 5. A picture of the study device.

4.2.2 Data collection

Study I: The data were collected on a data collection form by the study personnel during the intraoperative period and by the anesthesia nurse during postoperative care. In the test group the anesthesia team could not see the SPI and Entropy on the monitor and the values were printed out afterwards from the monitor. The nurse in the postoperative care unit and the person interviewing the patient on the day after surgery were always blinded to the group allocation. Patients were interviewed by the study personnel using a

structured form on the day after surgery either on the ward or by phone at home. Studies II-IV: The numerical data of all recorded parameters from standard monitor were recorded on a laptop computer using S5 Collect software (GE Healthcare, Helsinki, Finland) at 5-second intervals. A text file was used to mark the time points on a text file for post study synchronization. Studies III-IV: The numerical data from the study device were recorded electronically and stored on a server via a Wi-Fi connection. All the data were downloaded from the server after the study measurements and synchronized with the data from standard monitor during post hoc analysis.

4.3 Statistical analysis

4.3.1 Sample size estimation

Study I: According to the study by Chen et al (2010), 30 subjects in each group would be needed to show the decrease of unwanted effects from 1/subject to 0.15/subject. However, their study was a single-site study with a great reduction in the number of unwanted signs of inadequate anesthesia. Our goal was to demonstrate the effect in a multi-center study with different organizations and different populations. Therefore the study size was not based on power calculation but the number of patients was increased so as to have a reasonable number of subjects to permit generalization. Four hospitals in four different countries took part in the study. With a total enrolment target of 500 patients, a minimum of 250 subjects in each group would have 80% power for showing a decrease in number of signs of inadequate anesthesia from 1/subject to 0.76/subject, a statistically significant level of 0.05 (two-sided). The 24% decrease can still be considered clinically relevant. Study II: The sample size was calculated with an expected 20-unit difference (SD 14.4). Nine patients in each group would then be needed to achieve the power of 80% (p<0.05). The numbers used in the calculations were based on studies performed on adults and therefore the number of patients in each group was increased to 15.

Study III: According to pilot data from the manufacturer the blood pressure measurement has a mean bias of maximum 3 (SD 3) mmHg when compared to a traditional tourniquet. Assuming similar differences a minimum of 17 subjects per group was required to provide 80% power (p<0.05). Furthermore, the Association for the Advancement of Medical Instrumentation recommends a minimum of 15 subjects and 10 readings/ subject to be recorded (Association for the Advancement of Medical instrumentation 2013). Study IV: The power analysis was calculated with numbers from measurements comparing reflectance measurements at the forehead to absorption measurements at the fingers. In the calculation a mean difference of -1.39% at the reflectance sensor compared to a mean difference of -2.61% in absorption-type measurements and a standard deviation of 1.3% was used. Thus, in total 18 patients were needed to achieve a power of 0.80 (p<0.05). While the device was still in the testing phase the number of participants was increased to cater for greater bias and loss of data.

4.3.2 Data analysis All analyses were performed using SPSS software version 23.0. The data are presented as mean (standard deviation) for parametric and median (interquartile range or minimum and maximum) for non-parametric data depending on the distribution. Statistical tests used were t-test, Wilcoxon signed rank tests, Chi- suare test, Mann-Whitney-U test, and Kruskall-Wallis test. The logistic rank test comparing groups at time to eye-opening and awakening was performed with Kaplan-Meier curve. Studies III-IV: The data for the Bland–Altman plot were calculated using Microsoft Excel 10 as described by Zou (2013). The trending analysis was visualized with four-quadrant plot using MATLAB version 2014b.

4.4 Ethical statement

Study I was approved by the ethics committees of each of the four university hospitals. The ethics committee of Tampere University Hospital granted approval for Studies II-IV. All studies were conducted in accordance with the declaration of Helsinki and registered at ClinicalTrials.gov prior to patient enrolment. All patients gave written informed consent prior to any study measurements except for Study II, where consent was given by parents.

Study I was supported by GE Healthcare Finland. The study sponsor provided the relevant study equipment and was involved in the planning and monitoring of the study and also in data collection and analysis. A medical writing team did the data interpretation and preparation of the manuscript. Otherwise the studies were partly supported by competitive research funding from Pirkanmaa Hospital District, Tampere University hospital, the Finnish Cultural Foundation, Pirkanmaa Regional Fund, the Paolo Foundation and the Finnish Society of Anesthesiologists.

5 Summary of the results

A summary of the demographic data for all studies is described in Table 5. The studies were accomplished as one multicenter trial (I) and two single-center trials divided into three different publications (II, III and IV).

Table 5. Demographic data in Studies I-IV, presented as means (SD) except for age in Study II, presented as median (range) in weeks. Study n Group Age Height(cm) Weight(kg) I 248 Control 48.1 (15.7) years 168.4 (8.7) 72.2 (13.7) 246 Test 47.6(15.1) years 169.2 (9.9) 72.5 (14.7) II 15 Block 21.6 (6-78) weeks 58 (11.9) 5.8 (3.1) 15 Saline 23.7 (7-80) weeks 63.0 (2.2) 6.8 (2.2) III-IV 30 - 67 (13) years 172 (10) 80 (16)

5.1 Adequacy of Anesthesia (I)

The initial target number of patients was 500, however data collection was ended at 494 patients due to problems with the availability of the study medication. The groups were comparable as regards to the patient demographics. The types of surgery and mean durations of surgery are described in Table 6. There were no significant differences between the groups. Regarding site comparison, one site (Kiel) performed 100% ear, nose, and throat surgery, one site (Tampere) 100% gynecological surgery, whereas the others had a mixed profile of types of surgery as described in Table 7. Furthermore, the consumption of anesthetics differed slightly between hospitals as described in Table 8. The concentrations of anesthetics and the characteristics of SPI and Entropy are described in Figure 6. and 7.

Table 6. Types of surgery and duration of procedures. Duration of surgery is calculated from start of surgery to end of surgery. Duration of anesthesia is calculated from intubation to discontinuation of anesthetics. Time to eye-opening is calculated from discontinuation of anesthetics to emergence. Propofol and remifentanil consumption are normalized infusion rate by duration of anesthesia and patient weight. The numbers are described as percentages and numbers or mean and standard deviation. P-value is calculated using student’s t-test or Fisher’s exact test. Control Group Test Group All Subjects (n=248) (n=246) (n=494) p-value Type of Surgery 0.224 Ear Nose Throat 29.8% (74) 30.5% (75) 30.2% (149) Gastrointestinal 3.6% (9) 2.4% (6) 3.0% (15) Gynegological 52.4% (129) 49.6% (122) 51.0% (251) Maxillofacial 6.0% (15) 8.9% (22) 7.5% (37) Minor Trauma 2.0% (5) 0.0% (0) 1.0% (5) Orthopedic 1.6% (4) 2.4% (6) 2.0% (10) Other 4.4% (11) 6.1% (15) 5.3% (26) Duration of Surgery 119.1 (54.9) 118.1 (57.7) 118.6 (56.3) 0.836 (min) Duration of 142.2 (58.5) 140.3 (60.6) 141.3 (59.6) 0.722 Anesthesia (min) Time to Eye-opening 9.6 (7.3) 8.0 (5.2) 8.8 (6.4) 0.005 (min) Propofol (mg/kg/hr) 7.53 (2.9) 6.88 (2.5) 7.21 (2.7) 0.008 Remifentanil 0.20 (0.1) 0.21 (0.1) 0.20 (0.1) 0.617 (µg/kg/min)

Table 7. Type of surgery performed at each site

Site Type of surgery Control Group Test Group All Subjects Kiel n=74 n=75 n=149 Ear Nose Throat 100.0% 98.7% 99.3% Orthopedic 0.0% 1.3% 0.7% Tampere n=75 n=75 n=150 Gynecological 100.0% 100.0% 100.0% Szeged n=48 n=48 n=96 Gynecological 81.3% 91.7% 86.5% Minor trauma 10.4% 0.0% 5.2% Orthopedic 8.3% 8.3% 8.3% Amsterdam n=51 n=48 n=99 Ear Nose Throat 0.0% 2.1% 1.0% Gastrointestinal 17.6% 12.5% 15.2% Gynecological 31.4% 6.3% 19.2% Maxillofacial 29.4% 45.8% 37.4% Orthopedic 0.0% 2.1% 1.0% Other 21.6% 31.3% 26.3%

Table 8. Consumption of anesthetic per site. Described as mean (SD). P-value is calculated using student’s t-test.

Control Group Test Group All Subjects p-value Propofol (mg/kg/hr) Kiel 5.6 (1.3) 5.8 (1.2) 5.7 (1.2) 0.344 Tampere 9.3 (1.9) 8.3 (1.9) 8.8 (1.9) 0.003 Szeged 8.4 (2.5) 7.7 (3.3) 8.0 (2.9) 0.294 Amsterdam 8.6 (4.6) 7.2 (3.6) 7.9 (4.2) 0.105 Remifentanil (µg/kg/min) Kiel 0.29 (0.1) 0.23 (0.1) 0.26 (0.1) 0.002 Tampere 0.15 (0.0) 0.17 (0.0) 0.16 (0.0) 0.008 Szeged 0.14 (0.0) 0.20(0.1) 0.17 (0.1) <0.001 Amsterdam 0.25 (0.1) 0.29 (0.1) 0.27 (0.1) 0.149 Propofol and remifentanil consumption are normalized infusion rate by duration of anesthesia and patient weight.

Figure 6. Changes in predicted effect-site concentration of propofol (A) and remifentanil (B) at certain time points during anesthesia. Described as mean (SD).

Figure 7. Changes in SPI and State Entropy in the AoA group at certain time points during anesthesia. Described as means (SD).

The time fractions of Entropy and SPI values during anesthesia are described in Table 9. During the time between incision and the end of surgery SPI was within the recommended range 75.0% (SD 19.6) of time, while Entropy was within the range 45.0% (SD 33.0) of time. The time to eye- opening is described in Figure 8. The number of cases of inadequate anesthesia was comparable between the groups (Table 10.). The occurrence of an inadequate anesthesia was defined according to the publication by Chen et al. 2010 as described in Table 4 on page 52. The number of movements defined as somatic arousal (grimacing, coughing etc.) or somatic response (purposeful movement) was 71 in the control group and 64 in the test group.

Table 9. Time fractions of actual SPI and Entropy values during anesthesia in the AoA group. Data is for the duration from start of surgery to end of surgery described as mean (SD).

Fraction of time (n=244) SPI < 20 11.6 (16.7) 20 £ SPI < 50 75.0 (19.6) SPI ³ 50 13.4 (16.5)

SE < 20, % 1.7 (5.5) 20 £ SE < 30, % 12.7 (18.9) 30 £ SE < 40, % 39.2 (26.5) 40 £ SE < 60, % 45.0 (33.0) SE ³ 60, % 1.5 (6.7)

20 £ SPI £ 50 and 40 £ SE £ 60, % 34.1 (28.49

Figure 8. The Kaplan-Meier curve representing time to eye-opening. Calculated from discontinuation of anesthetics to extubation. The log-rank p-value comparing the AoA group and control group is 0.0016.

Table 10. Number of unwanted events during the intraoperative period based on the definition in Table 4.

Control Group Test Group Event (N=248) (N=246) p-value Inadequate Anesthesia 235 (0.9) 224 (0.9) 0.936 Hypertension 133 (0.5) 146 (0.6) 0.438 Tachycardia 31 (0.1) 14 (0.1) 0.385 Movements 71 (0.3) 64 (0.3) 0.591 Somatic Arousal 55 (0.2) 43 (0.2) 0.868 Somatic Response 16 (0.1) 21 (0.1) 0.519 Hypotension 509 (2.1) 568 (2.3) 0.486 Bradycardia 301 (1.2) 339 (1.4) 0.822 Total Unwanted Events 1045 (4.2) 1131 (4.6) 0.598 Values in parentheses describe average number of unwanted events per subject. P-value is based on comparison of number of event per subject using non-parametric Wilcoxon test.

The cumulative percentage of patients who were not fully awake after arrival at the post anesthesia care unit (PACU) was higher in the control group (Figure 9. By contrast, there were no significant differences between the groups in level of consciousness, oxygen saturation, post-operative nausea and vomiting (PONV), or VAS at time points of arrival at the PACU, 30 minutes after arrival or 1 hour after arrival (p-values between 0.107 to1.00, for the complete results see Table 3 in Study I). The measured blood pressures were mostly within 20% of the preanesthetic level. On arrival at the PACU a slightly higher percentage of patients in the control group had values > 20% of the level measured before anesthesia (17% vs 9%, p-value 0.025). The blood pressures were normalized after 30 minutes after arrival to PACU (BP <20%: 88% vs. 87%, p-value 0.944). In the post-operative survey, the mean satisfaction with overall anesthesia was 95.8 on a scale of 0 to 100. Most of the patients (74%) reported the induction of anesthesia (mask on face) as the last thing they remembered prior to anesthesia. One patient in the control group reported events during surgery and one in the test group reported pain during surgery. These did not necessitate any intervention. After the procedure, most patients reported remembering events from the post-anesthesia care unit (87%) while very few patients recalled the way to the post-anesthesia care unit (13%) or removal of the endotracheal tube (2.3%).

Figure 9. Cumulative percentage of patients fully awake after arrival to the post anesthesia care unit (PACU). The log rank p-value comparing the test group and control group is 0.0427

5.2 Reactivity of SPI in small children (II)

The groups were compared using the mean number of intubation attempts (Block group 1.6 vs. Saline group 1.5 attempts), time from block to incision (23:33 vs. 18:57 minutes), total amount of fentanyl administered (3.0 vs. 2.4 µg/kg) and concentration of sevoflurane at incision (2.9% vs. 2.9% end tidal). There were no significant differences in the demographic data between the groups

SPI was found to be reactive with a median change at the times of intubation from 52.5 to 58.5 in block group and in the saline group from 49.0 to 62.8. By contrast, in the block group the change was non-significant at incision from 58.0 to 57.3. The changes in SPI, RRInorm, and PPGAnorm are described in Table 11. and in Figure 10. for SPI. The correlation between patients’ gestational age and the change in the SPI was weak (Spearman correlation 0.34, p=0.075).

Figure 10. Reactions in SPI at the point of incision in the block group (A) and saline group (B). Grey lines represent each individual and the black line the median of all measurements.

Table 11. SPI, PPGAnorm, and RRInorm values at different time points. For intubation and incision the median value of 30 seconds before (A) and after (B) each time point is presented. For signs of inadequate anesthesia the median value at the time point (B) and 5 minutes before (A) is presented. P-value is calculated with Wilcoxon signed rank test. Described as median (IQR). Modified and reproduced with the kind permission of Oxford University Press.

n A B p-value SPI Intubation 30 52.5 58.5 0.019* (40.5 ‒ 69.5) (50.5 ‒ 58.5) Incision SG 15 49.00 62.8 0.048* (37.6 ‒ 67.2) (53.0 ‒ 69.9) Incision BG 15 58.0 57.3 0.177 (44.9 ‒ 60.6) (51.8 ‒ 68.8) Signs of 28 56.0 60.0 0.148 inadequate (47.6 ‒ 68.9) (52.8 ‒ 76.0) anesthesia PPGAnorm Intubation 30 522.0 426.6 0.003* (273.0 ‒ 713.2) (157.1 ‒ 595.9) Incision SG 15 521.0 349.4 0.397 (252.5 ‒ 635.5) (214.8 ‒ 520.6) Incision BG 15 396.0 323.9 0.009* (274.0 ‒ 493.0) (178.4 ‒ 440.7) RRInorm Intubation 30 263.3 293.3 0.387 (97.0 ‒ 518.0) (129.0 ‒ 405.5) Incision SG 15 587.3 481.5 0.007* (480.3 ‒ 690.5) (186.4 ‒ 543.3) Incision BG 15 667.0 631.3 0.087 (564.8 ‒ 690.0) (243.4 ‒ 715.3)

5.3 Comparison of modified arterial applanation tonometry with intra- arterial pressure measurement. (III)

The comparison between invasive measurement and tonometry measurement in Bland–Altman plot revealed a bias in systolic pressure of 19.8 mmHg Limits of agreement (LoA): [LoAupper 59.6 mmHg (95% Confidence interval(CI) 44.4 to 67.8), LoAlower -20.1 (CI -4.9 to 28.3)], Spearman’s correlation was r = 0.61.

For diastolic pressure the difference was 4.8 mmHg [LoAupper 23.6 (CI 16.9 to

27.4), LoAlower -14.0 (CI -17.8 to -7.3)], r = 0.72, and for mean pressure it was

11.1 mmHg [LoAupper 34.3(CI 21.6 to 40.2), LoAlower -12.1 (CI -18.0 to 0.6)], r = 0.642. The Bland–Altman plot is illustrated in Figure 11. The measurement of blood pressure was unsuccessful in 22% of all attempted measurements. (n=2449). Even slight movement (35% vs. 20%, p<0.001) and peripheral arterial disease (27% vs. 8.6%, p=0.042) were found to increase the number of failed measurements. An increase in the RMSE describing the mean error was associated with increasing body weight (for diastolic pressure RMSE was 10.8, 8.0 and 5.5 for three equal quartiles, p=0.003) and wrist circumference (RMSE 11.3, 7.5 and 5.5 for three equal quartiles, p=0.004). Interestingly, movement decreased the error [for diastolic pressure 5.5(5.2 to7.7) to 7.5(5.2 to10.2) p<0.001]. The trending ability to show changes of similar direction is described in Figure 12.

Figure 11. Bland–Altman plot and Spearman’s correlation coefficient for systolic pressure (mmHg) (n = 28 patients). The black lines and numbers describe the mean and limits of agreement. The gray areas indicate corresponding 95% confidence intervals. Modified and reproduced with the kind permission of Springer Science + Business Media Dordrecht.

Figure 12. Four-quadrant plot visualizing the trending ability in mean arterial pressures between invasive blood pressure ((DB-IVMAP) measurements and BPro® sensor (DBP-BProMAP). The horizontal line describes the change at 5-minute intervals in invasive pressure while the vertical axis describes similar changes in BPro® sensor measurements. The exclusion zone of 3 mmHg is shown in gray. Each asterisk represents a data point. Data points at the lower left or upper right corner indicate a change in the same direction. Concordance rate indicates the percentage of data points with in the same direction. Included data points n= 1128, excluded data points n= 348. Reproduced with kind permission from Springer Science + Business Media Dordrecht

5.4 Detection of vital signs using plethysmography (IV)

5.4.1 Comparison of reflection type saturation measurement at the distal forearm with finger measurement using transmission mode. (IV)

The measurements recorded with a finger sensor were compared to those from the distal forearm recorded with the study device. There were 28 patients and 10,767 measurements left for comparison. In the Bland–Altman plot, we found a low mean difference in SpO2 [-0.3 % points (95 % conﬁdence interval -3.9 to

5.1), Limits of Agreement (LoAupper): 7.2 (CI 6.1 to 8.1), LoAlower -7.9 (CI -8.8 to -6.8)] with a low patient weighted Spearman’s correlation between devices (r = 0.142) and the RMSE 4.2 points. The accuracy of the sensor was associated with movement (RMSE 4.1 vs. 3.8, p=0.005), but not with increasing wrist circumference (RMSE 4.1, 4.7 and 2.9 in three equally sized quartiles, p=1.00).

5.4.2 Comparison of plethysmographic heart rate measurements (IV)

The heart rate measurements from the finger saturation sensor were compared to readings from the plethysmography sensor at the distal forearm. The analysis was composed of 28 patients and 14,832 measurements. In the Bland–Altman plot the HR showed a low mean difference [0.6 bpm (CI -0.85 to 2.05), LoAupper 5.6 (CI 5.2 to 5.9). LoAlower -4.4 (CI -4.8 to -4.0)] with good Spearman’s correlation (r = 0.997) and low RMSE (1.8 bpm). The RMSE was slightly higher for movement (2.4 bpm vs 1.9 bpm, Bonferroni corrected p=0.005), but not with peripheral arterial disease (RMSE 1.8 vs 1.9, p-Bonf =1) or increasing wrist circumference (RMSE 2.4, 1.7 and 1.7 in three equal quartiles, p-Bonf=1).

5.4.3 Comparison of tonometric heart rate with intra-arterial measurement (IV)

The heart rate measurement from the tonometric sensor was compared against the measurement from the invasive pressure measurements. There was a mean bias of -1.2 beats/min with a 95% Confidence Interval of -5.8 to 3.3. The results are presented in Figure 13. with a Bland–Altman plot.

Figure 13. Bland–Altman plot for arterial heart rate measurement and tonometric heart rate measurement. Described as patient weighted mean. Black lines and numbers represent mean and 95% Confidence Intervals. n=29.

5.4.4 Comparison plethysmographic respiratory rate measurement to impedance pneumography (IV)

The respiratory rate measurements from ECG impedance pneumography were compared to the plethysmography sensor readings from the distal forearm using the study data. A total of 28 patients and 18,857 measurements were compared. The respiratory rate algorithm was also tested on validation data comparing impedance pneumography to readings obtained from finger plethysmography (Charlton) (38 patients and 25,154 measurements). The results are presented in Table 12. The RMSE was associated with movement statistically (p-Bonf=0.005), but the numerical change in the value was low (3.8 1/min vs. 4.1 1/min).

Table 12. Comparison between respiratory rates from pneumography impedance measurement (ECG) and the values obtained with the study device from the distal forearm (study data) or from the finger (validation data) respiratory algorithm using study data and validation data. The data are presented as patient weighted mean (95% Confidence Intervals) associated with corresponding lower (LoAlower) and upper (LoAupper) limits of agreement (95% Confidence Interval).

Site n Bias LoAlower LoAupper (95% CI) (95% CI) (95% CI) Study data ECG- 28 0.6 -6.8 8.0 forearm (-3.9 – 5.1) (-7.4 – -6.0) (7.2 – 8.6) Validation data ECG – 39 4.1 -3.8 11.9 finger (-2.7 – 10.8) (-3.0 – -4.7) (11.1 – 12.7)

6 Discussion

Our findings describe different ways to utilize the peripheral pulse wave. The measurement at a peripheral location is known to be highly sensitive to errors and sensors located on the fingers are typically displaced or even easily detached. However, as seen in Studies III-IV, the distal forearm region was sensitive to movement when either plethysmographic or tonometric technology was used. As described in the previous sections, PPG waveform is influenced by several physiological factors, which can also be quantified using the waveform. For example, plethysmographic technology has been used to measure fluid responsiveness (Biais et al. 2011; Broch et al. 2011; Cannesson et al. 2008) in which the measurement is based on changes in plethysmographic amplitude (Delerme et al. 2007). Changes in the same waveform are similarly used in the calculation of SPI (Huiku et al. 2007) and respiratory rate (Addison et al. 2015). It is therefore highly likely that changes in volume may affect the measurements derived from peripheral pulse wave. This was likewise seen in SPI when the Trendelenburg position elevated the values significantly (Ilies et al. 2012). The patients in our studies were relatively low-risk and in stable condition, thus conclusions on how the measurement would work on unstable patients cannot be drawn on the basis of our present findings. In Study I several patients were undergoing gynecological surgery and they were placed in the Trendelenburg position, which did not prevent SPI measurement. Moreover, the patients were not studied under sympathomimetic medication, which might have drastically affected all the measurements taken.

6.1 Adequacy of anesthesia (I) Interestingly, in Study I we found hardly any differences in the primary or secondary endpoints in the test group treated with depth of nociception and anesthesia monitoring when compared to standard monitoring alone. Only the consumption of propofol and time to eye-opening were slightly different

between groups in favor of the test group. This is in contrast with earlier findings reporting a favorable, although sometimes minor and partly controversial effect of SPI and depth of anesthesia monitoring when compared to standard monitoring (Bergmann et al. 2013; Chen et al. 2010; Colombo et al. 2015; Park et al. 2015; Rogobete et al. 2017, Won et al. 2016). A recent meta- analysis combining these studies failed to show benefits for nociception monitoring (Gruenewald & Dempfle 2017). As discussed in the review of the literature it is generally accepted that SPI reflects nociceptive reaction. The ability of the index to appropriately guide anesthesia is more controversial. In earlier research only one study has reported a major decrease in the number of unwanted effects and the number of movements was highly reduced (Chen et al. 2010). In others, this has not been reported (Colombo et al. 2015) nor was any major benefit seen (Bergmann et al. 2013; Gruenewald et al. 2014; Park et al. 2015). One possible explanation for this could be that in the study by Chen et al (2010) mean consumption of propofol was very low in both groups (5.3 vs 5.6 mg/kg/h). In comparison, Bergman et al. (2013) reported a much higher consumption (6.0 vs 7.5 mg/kg/h), which is closer, but still much less than that used in our study. As seen in our study, time to eye-opening similarly favored the test group when described by Kaplan-Meier curve (Bergmann et al. 2013; Chen et al. 2010; Colombo et al. 2014). This shows that there may be some advantage in using adequacy of anesthesia monitoring but confounding factors may hinder this minor effect seen in other studies. Interestingly, in our study the consumption of remifentanil was slightly increased at all but one site, where in the test group it actually decreased. The study protocol did not limit the interrelationship of propofol and remifentanil. In other words, the concentration of propofol was allowed to be very low even if remifentanil was at a very high level. Indeed, there seemed to be differences between sites as to how the guidelines were interpreted and the consumption of anesthetics also varied slightly between hospitals. This may explain why there was no change in the consumption of anesthetics as has been reported in some earlier studies (Bergmann et al. 2013; Chen et al. 2010; Park et al. 2015). It must nevertheless be noted that depth of anesthesia measurements is optimized to GABAergic drugs and the measurement may not be reliable if very low concentrations of anesthetics are used (Särkelä et al. 2002; Viertiö-Oja et al. 2004) which was probably linked to the fact that some of the study sites first

76 attempted to lower the concentration of remifentanil before lowering the concentration of propofol to the lowest accepted level of 2 µg/ml. The definition of signs of inadequate anesthesia on the basis of blood pressure and heart rate values may be problematic. One problem may be that the level of anxiety raises the blood pressure if they are measured immediately before surgery and the values might conversely be too low if they are taken immediately after induction of anesthesia. In our study the baseline values were recorded with patients awake prior to anesthesia, as also in earlier study settings (Chen et al. 2010; Gruenewald et al. 2014). The blood pressure values measured at the post-operative care unit were mostly within 20% of the preanesthetic level indicating acceptable baseline values. Hence it is likely that this did not affect the absence of differences between the groups. The conclusion of our study was disappointing. Monitoring SPI and Entropy yielded no significant benefit. It may be that if we had used Entropy in both groups it would have reduced the number of confounding factors, thus making it more likely to see the small benefit of monitoring SPI seen in other studies. On the other hand, it is unlikely that the index would show any benefit when more slowly acting opioids are used. One sign of such an effect was a study with sevoflurane and sufentanil that also failed to show any benefit in the SPI guided group (Gruenewald et al. 2014). The same study proposed that using a rapid change in SPI as an indicator for nociceptive reaction might be superior to the strict target range used in previous studies.

6.2 SPI in small children (II)

In Study II SPI was shown to be reactive at intubation and incision in small children. The reaction was also blunted with the use of neuraxial block, as one might anticipate. However, the magnitude of the reactions was rather small, nor was the reaction in the components of SPI, PPGAnorm, and RRInorm completely logical. The changes in PPGAnorm were consistent. In all groups the numerical value of PPGAnorm decreased indicating sympathetic vasoconstriction in the fingers as a reaction to nociceptive stimulus. This is in line with findings in adults and older children with a similar reaction after incision and intubation (Huiku et al. 2007; Kallio et al. 2008). The change was significant only at

intubation and incision in the block group, probably because of a wide range in values described by inter-quartile range. In the RRI there was a decrease in the values at incision (=increase in heart rate) in both groups. This change was greater and significant only in the saline group, indicating that the reaction was blunted with neuraxial block. At intubation the reaction was paradoxical with increasing RRI (=decreasing heart rate). This may be due to a strong vagal stimulus caused by intubation. A similar reaction could likewise be seen after intubation in some patients, including adults (Luginbuhl et al. 2007) and older children (Kallio et al. 2008). Our study is the first SPI study conducted on very small children. Although SPI was reactive, we found a marked inter-individual variability as can be seen in Figure 10. The nervous system of small children is still developing (Davis 2011), which may explain why some patients reacted strongly while others had almost no reaction at all. Furthermore, the reaction was rather small and the pre-stimulus level high when compared to that of adults (Huiku et al. 2007; Struys et al. 2007). As described by Huiku et al. (2007), the heart rate constitutes about one third of the calculation. The heart rate is typically much higher in small children than in adults (Daymont et al. 2015). As a result, the proportion of heart rate in the calculation of SPI increases and the baseline level increases. For the same reason, a much greater increase in heart rate is also needed to cause a similar change in SPI. In other words, one might expect a higher baseline value and smaller percentage increase in small children than in older children (Kallio et al. 2008) or adults (Huiku et al. 2007). Our study shows that SPI is potentially useful in small children, but more research and probably modification of the algorithm is needed before it can be adopted for clinical use.

6.3 Measurement of blood pressure using modified applanation tonometry (III)

In Study III we compared the readings from BPro® applanation tonometry sensor to invasive blood pressure measurement. The recordings of systolic and mean blood pressure were inaccurate, whereas diastolic pressures were just within the recommendations by Association for the Advancement of Medical Instrumentation (Association for the Advancement of Medical Instrumentation 2013). In all pressures the limits of agreement were also unacceptably high and

78 the trending ability very poor with a probability to predict the direction of change in pressures similarly in only some 50% of cases. Therefore the tonometry sensor cannot successfully replace the gold standard of invasive blood pressure measurement. In contrast to our findings, earlier studies have reported good performance for the sensor (Komori et al. 2013; Nair et al. 2008). This was seen especially in the study by Nair et al. (2008) showing a mean difference of 1.3 mmHg in systolic pressure compared to non-invasive sphygmanometer readings. In the study by Komori et al. (2013) the mean difference was slightly greater at 4.5 mmHg when awake. In both these studies, the patients were young and healthy, which is not the case with our patients. In patients with more comorbidity the number of confounding factors increases, impairing the performance of the device. There is very little evidence on the performance of the BPro sensor compared to invasive blood pressure. One study measured central blood pressure and found a low mean bias of 0.87mmHg (SD 13 mmHg) (Ott et al. 2012), which is better than that found in our study. In that study, an average of the invasive measurements was used which may have improved the accuracy. Yet the accuracy was still rather low. In our study the increasing inaccuracy was seen with increasing wrist circumference and increasing body mass index. These result in longer distances between the tonometer and the artery. Komori et al. also reported a high proportion of failed measurements with an overall 24-h failure rate 49% (Komori et al. 2013), which is even higher than the value of 22% reported in our study. In ambulatory monitoring the patients were engaged in activities of daily living, and thus probably moving much more than our patients, who were under post-operative surveillance. As we demonstrated, even slight movement was associated with a higher failure rate, whereas the inaccuracy was not increased. We set the device to measure for 10 seconds at one-minute intervals. It seems that if the device is able to catch even some of the pulse waves correctly within the 10 seconds measuring period, the algorithm still produces accurate measurements. However, if the algorithm fails to measure the pulse waves, it discards the measurement. The BPro® sensor allows fairly free movement of the wrist. The sensor is placed in the best position by manual palpation of the radial pulse, but thereafter no further procedures should be needed. In our study it was evident that this was not enough to produce adequate measurements. It may be a failure of the algorithm to return an error when inadequate measurements were

present. Thus more measurements should have been found inadequate and rejected as a measurement failure. Another approach is used in the Tensys T- Line device. The best signal is constantly located with a servo motor that is able to move the sensor. The manufacturer also recommends fixation of the wrist, thereby reducing the movement between the radial pulse and the sensor. This approach has resulted in very good pressure readings (Langwieser et al. 2015; Meidert et al. 2013; Saugel et al. 2014), but at the cost of less mobility.

6.4 Indications for photoplethysmography

6.4.1 Measurement of saturation at distal forearm (IV)

In Study IV we compared recordings from a saturation sensor placed in the distal forearm region to readings from the fingers. We found a low mean difference and relatively large limits of agreement. In earlier studies the reflectance sensor has shown good results when used at its intended location, the forehead compared to finger sensors (Casati et al. 2007; Nesseler et al. 2012; Sugino et al. 2004; Wax et al. 2009). The study by Nesseler et al. (2012) compared saturation readings from the forehead with arterial oxygen saturation readings and the results correlated even better than finger readings with a mean bias of +1.0 (LoA -4.0 – 6.0). It therefore seems evident that the high RMSE and low correlation between devices in our study is more related to the measurement site than to the technology. Low mean difference indicates that with better tolerance to artifacts the performance of the device might be enhanced. In studies comparing transmission type sensors to arterial oxygen saturation the bias has been within 2% (1SD) (Jensen et al. 1998). Most manufacturers claim an accuracy of 2% for pulse oximetry devices, which means that 95% of the values should be within 4% of the true value (Nitzan et al. 2014). Thus finger measurement seems to be more reliable than distal forearm measurements as was also the case in our study. Theoretically, the measurement at the fingers suffers from hypothermia and vasoconstriction (Hynson et al. 1992; Talke & Stapelfeldt 2006), which might be better tolerated at the distal forearm but the assumption still lacks sufficient evidence.

6.4.2 Measurement of heart rate (IV)

In Study IV the measurement of heart rate was found to be very accurate. Both bias and precision were very low and the measurement seemed to be very resistant to movement. This is in line with earlier findings suggesting an only slightly lower relative power of frequencies corresponding to heart rate at the wrist over the radial artery when compared to the fingers and lower density when placed more proximally (Nilsson et al. 2007). In our study placement of the sensor was over the radial bone more dorsally and not directly over the radial artery due to technical reasons. Photoplethysmographic sensors have also been tested on athletes and healthy volunteers, with mixed results (El-Amrawy & Nounou 2015; Fukushima et al. 2012; Y. Wang et al. 2014). The commercially available sensors have not so far been adequately validated for medical use (Patel et al. 2015), but our findings suggest that these do indeed have potential for heart rate monitoring at the wrist.

6.4.3 Measurement of respiratory rate (IV)

The recording of respiratory rate using a novel algorithm at the distal forearm had a low bias and relatively high precision when compared to readings at the fingers. In contrast, the accuracy of the device was also tested using readily available validation data (Charlton), which revealed a much higher bias and lower precision. The validation data consisted of raw plethysmographic waveform obtained gained during exercise and from healthy volunteers using a finger transmittance sensor (Charlton et al. 2016). This finding supports our suggestion that the distal forearm region is a more suitable place for measuring respiratory rate than the finger. Our findings are in line with those of an earlier study claiming much higher relative power of frequencies corresponding to respiratory rate at the forearm than at the fingers (Nilsson et al. 2007). Unlike in the study by Nilsson et al. our sensor was placed more laterally and distally on the forearm. Our sensor was fitted in the same wristband with a tonometry sensor limiting the placement of plethysmography sensor. The results might have been even better if the sensor had been more proximally on the forearm. The movement had only a slight effect on the measurement, although it was statistically significant (p=0.005). Similar findings were also reported by Lee et al (Lee et al. 2011). Other studies investigating accuracy have mainly confirmed the high coherence between the signals, indicating a stable relationship between

the signals at the respiratory rate (Leonard et al. 2006; Nilsson et al. 2003a; Nilsson et al. 2003b; Nilsson et al. 2005; Vegfors et al. 1993). Nilsson studied the signal on the medial forearm, while Leonard had the sensor on the finger. As a conclusion the distal forearm region seems to have potential for respiratory rate measurement. In our study the agreement between the devices remained constant even with a single subject breathing at very high frequency. None of the technologies presented have undergone proper validation and more studies are needed before the technology can be introduced into clinical practice.

6.4.4 Measurement of vital signs at the distal forearm

In this thesis we described the performance of a device with potential for the measurement of basic vital signs at one single location. While the placement of the tonometry sensor was the most limiting, it was also the most disappointing. The accuracy was far from acceptable. Also, the choice of technology necessitated the use of two different sensors, while, for example, pulse wave velocity could have been measured using photoplethysmography. Furthermore, the measurement of saturation was associated with rather low precision. A display showing the signal might have enabled better placement of the sensor.

6.5 Strengths and weaknesses of the studies

The aim of this thesis was to prospectively study clinically important issues associated with the peripheral pulse wave. The peripheral pulse is easily accessible and contains a great deal of important information that is so far only partially utilized. The first purpose of these studies was to explore the measurement of nociception. Improvement of delivery of anesthetics aims to reduce intraoperative awareness and improve operative conditions by simultaneously decreasing the side effects of anesthesia during postoperative care. Small children have also a marked risk for postoperative apnea, which could potentially be reduced by optimization of the delivery of anesthetics. Studies I and II were randomized case-control studies. Both were at least partially blinded decreasing the study bias. Study I was a multicenter trial

82 including a considerable number of patients. This increases the credibility of the study. All the sites followed a protocol in the conduct of anesthesia. However, the range of infusions was relatively wide, resulting in different balances between anesthetics at different sites. While the selection of patients was not restricted according to the type of surgery, each of the sites had a different profile of surgical operations performed, making comparison between sites difficult. Also, a relatively small number of anesthesiologists participated in the study, thus limiting the generalizability of the results and possibly increasing bias. Although there was a strict protocol, it was not followed very well. The time fractions of SE and SPI were outside the limits for quite a long time during each anesthesia. Many of the operations were relatively short, leaving rather little time to reach the targets after surgery commenced. In Studies II-IV all study measurements were recorded electronically. This reduces the likelihood of typing errors. Study II examined the reactivity of SPI in small children. The study was randomized, double-blinded, and placebo controlled. The conduct of the anesthesia was moreover standardized. Study II had some limitations. Firstly, in order to ensure a stable and secure anesthesia, all patients were intubated. The large amount of fentanyl needed to facilitate endotracheal intubation may have blunted the relatively minor trauma caused by the incision. The high concentration of sevoflurane may also have blunted the reaction. Secondly, we chose to use ilioinguinal/ iliohypogastric nerve block to prevent reactions in block group. The block is known to have a limited effect on the abdominal wall and testicles, which made the groups comparable only at the time of incision and shortly after. Thirdly, the anticholinergic effect of glycopyrrolate may blunt the reactions in the SPI. Glycopyrrolate was used as a standard regimen in order to reduce the known adverse effects of succinylcholine that was used to facilitate endotracheal intubation. The data for Studies III and IV were collected simultaneously. The data was collected during normal treatment in the postoperative care unit and no procedures to artificially lower blood pressure, saturation or other measured parameters were used. The conditions of all patients were also relatively stable and they were mostly lying in the supine position, which limits generalizability of the values to all patients. The study was originally designed only for the measurement of blood pressure and the idea also to compare plethysmography measurements came after the initiation of patient collection. Therefore we did

not compare the saturation readings to gold standard arterial oxygen saturation or respiratory rate to carbon dioxide measurements. The use of a study platform in Studies III-IV caused several difficulties in the measurement. The device was planned to send and display all the data on a web page instantaneously throughout the study period. Therefore, it only had led light indicating proper Wi-Fi-signal and proper placement of the device. However, at the post anesthesia care unit the strength of the Wi-Fi signal was occasionally too low for adequate verification of data simultaneously with the measurement. This also caused the partial breakdown of the sensor cable to go unnoticed during the recording of two patients’ data and caused a partial loss of data. Additionally, at the time the device had no independent software and data bank for blood pressure analysis. The device sent the raw pressure signal to a remote server, which returned the blood pressure readings. An error in the server settings caused false initial blood pressure readings and necessitated the recalculation of all study device blood pressure readings during the post hoc analysis. At this point a strategy similar to what should have happened automatically was chosen. We manually selected first a data package containing a good quality signal from each patient and corresponding blood pressure from Carescape B650 monitor readings was set as a calibration pressure. The data package was then reanalyzed at the remote server, which returned the blood pressure readings used in the final analysis. In two patients there were no data packages with adequate plethysmography signal and those were discarded from the analysis. The first data package did not contain a good quality signal for any of the patients. This caused significant data loss, but enabled data comparison similar to what takes place in a commercially available BPro® device.

6.6 Clinical aspects and future perspectives

In Study I we found no benefit of using SPI and Entropy in addition to standard monitoring although other smaller studies have suggested such benefit. This may be due to bias in smaller studies, or then the target range and way to use SPI may have been suboptimal. (Gruenewald et al. 2009; Gruenewald et al. 2015) As described in the review of literature, a rapid change in the value, or another baseline value, might be prefereable for SPI. The studies on Entropy have also shown a varying effect with relatively minor

84 benefit. Recent studies on the potential side effects of too deep anesthesia warrant further research to show all benefits for depth of anesthesia monitoring. In Study II we were able to show that the reactivity seen in adults and older children is also present in small children. Additionally, at intubation and incision the PPGAnorm decreased in all groups, as expected. This suggests that with a proper modification for the algorithm for small children SPI could be improved for use in this age group. High heart rate currently causes impractically high SPI values and decreases the numeric change in the value. In Studies III and IV the performance of the study device proved disappointing. Of the values measured only heart rate was measured very accurately while the other measurements had varying degrees of bias and accuracy. Simultaneous measurement with an accelerometer would enable better artifact rejection. Thus the mean bias in all devices was low. A further problem was the absence of a display on the device. Although we were able to see the values on the web page, it was too slow and impractical. The comparison of respiratory rate measurement to validation data shows that the distal forearm region is a good place for measurements while measurement at the fingers has much higher bias and lower accuracy. In our measurements movement was associated with poorer performance. Saturation measurement was especially sensitive to movement artifacts. The forehead has been used for saturation measurements. Other places where there is bone close to the skin, such as the sternum or middle ear, might also be feasible for measurements. Currently there is not enough data for respiratory rate measurements on these locations. The blood pressure measurements were very inaccurate. It was evident that movement tolerance needs to be improved. The accelerometer might also help in that development. This will not help to reduce the bias associated with increasing wrist diameter and body mass index. These limitations might be overcome with the development of pulse wave velocity technology.

7 Conclusions

Based on these studies the following conclusions can be drawn: 1. According to our findings, monitoring adequacy of anesthesia was not associated with a reduced number of unwanted events. The amount of propofol was slightly reduced. There was a slight trend favoring shorter time to eye-opening during post-operative care. 2. The Surgical Pleth Index is reactive in children under two years of age after intubation and after surgical stimuli. The reactivity is rather small and a marked inter-individual variability in reactions limits the usefulness of the index in this age group. 3. The blood pressure readings obtained using arterial tonometric measurement and BPro® data analyses are not sufficient to replace invasive arterial blood pressure measurement in post-operative care. 4. The saturation measurements measured with Nellcor™ forehead SpO2- saturation sensor at the distal forearm have small bias but decidedly low precision when compared to finger plethysmography. 5. The heart rate can be measured from the distal forearm using Nellcor™ forehead SpO2 saturation sensor in stable conditions with high accuracy and small bias. 6. The novel respiratory rate algorithm can be used to measure respiratory rate in stable conditions with small bias and moderate accuracy. The algorithm is more reliable at the distal forearm while measurement at the fingers yields very inaccurate readings.

8 Acknowledgements

This work was carried out over a period of six years. It has been a very long, interesting, instructive and sometimes very frustrating journey. Many people were involved, some for a short period time and others for the whole process. I would like to express my deepest gratitude to the following people: First and foremost, I wish to express my deepest gratitude to Professor Arvi Yli-Hankala, first supervisor of this thesis, for guiding me through the whole process. Without your guidance, relationships and inspiration at several critical moments, this whole process would probably not even have started. Special thanks also go to Professor Niku Oksala, the second supervisor of this thesis, who joined the process a little later, but managed to save the whole process with his expertise in Studies III and IV. My sincere thanks go to my opponent Docent Anne Vakkuri as well as to the reviewers of the thesis, Professor Tarmo Lipping and Docent Vesa Kontinen for giving their precious time, suggestions and remarks. The co-authors of the publication of this thesis: Docent Maija Kalliomäki, who helped me through the first publication, Docent Antti Vehkaoja for being the main technical consultant, Heli Leppikangas PhD, Dr. Marja Kiviharju, Pekka Kumpulainen PhD, Stefano Campadello M.Sc., Dr. Ville Lindroos, Dr. Sasu Liuhanen, Prof. Matthias Grünewald, Prof. Berthold Bein as well as the whole AoA study group. I’m very grateful for your support. Language editor Virginia Mattila MA for finding and correcting grammatical errors in the manuscript of this thesis. Statistician Heini Huhtala M.Sc., for being the statistician on call. You have helpfully guided me through the numbers when I was totally lost with the statistics. My special thanks go to study nurses Simo Varila and Atte Kukkurainen for your invaluable help and support. GE Healthcare and its workers. Especially Petra Peltola and Matti Huiku PhD, who gave a major contribution to this work during the whole process. Drs. Anne Mäyrä, Susanna Mennander and Kati Rautaneva at LE4, who probably did not realize what they were promising when they agreed to treat the study patients. You have done enormous work in helping me.

Docent Sanna Hoppu, who pushed me forward at the beginning of the whole process. Sorry for not being good enough to finish the first direction of my study intentions. The nursing staff in our operating theatres and wards for their invaluable help in my studies. Without your assistance and patience in screening, recruiting, and treating the patients included in the studies, this thesis and these publications would not have been possible. There has always been a friendly smile and patience although conducting the study causedextra work. All my fellow anesthesiologists and friends at Tampere University Hospital. Especially the pediatric anesthesiologists Drs. Miia Kokkonen, Jenni Vieri and Hilkka Sivula, who all helped me in recruiting the study patients and conducting the studies. All you friends and colleagues make the hospital a great place to work. This thesis was financially supported by the Finnish Cultural Foundation: Pirkanmaa Regional Fund, the Finnish Society of Anaesthesiologists, The Paolo Foundation and City of Tampere (Tiederahasto) which I would like to thank sincerely. Also, thanks to my friends and relatives for giving something else to think about. This process has also taken time away from you. To my family, my parents, Simo and Pirkko, and my siblings Kirsi, Tiina, Paula and Tommi. I can only say – thank you. For everything. Finally, Eeva, my best friend, my beloved wife and the wonderful mother of our children. As we have gone through this same process at the same time you have been a great support, motivator and understander all the way. Our children, Anna and Olli, you are wonderful children and the light of my life. Youmake me proud every day.

Kangasala, May 2018

Jarkko Harju

9 References

Abad Torrent, A., Rodríguez Bustamante, V., Carrasco Fons, N., Roca Tutusaus, F. J., Blanco Vargas, D., & González García, C. (2015). The use of pupillometry as monitoring of intraoperative analgesia in the consumption of analgesics during the first 12 hours after surgery. Revista Espanola De Anestesiologia Y Reanimacion, doi:10.1016/j.redar.2015.07.006 Abay, T. Y., & Kyriacou, P. A. (2015). Reflectance photoplethysmography as noninvasive monitoring of tissue blood perfusion. IEEE Transactions on Bio-Medical Engineering, 62(9), 2187-2195. Addison, P. S., Wang, R., Uribe, A. A., & Bergese, S. D. (2015). On better estimating and normalizing the relationship between clinical parameters: Comparing respiratory modulations in the photoplethysmogram and blood pressure signal (DPOP versus PPV). Computational and Mathematical Methods in Medicine, 2015, 576340. Addison, P. S., Watson, J. N., Mestek, M. L., Ochs, J. P., Uribe, A. A., & Bergese, S. D. (2015). Pulse oximetry-derived respiratory rate in general care floor patients. Journal of Clinical Monitoring and Computing, 29(1), 113-120. Agashe, G. S., Coakley, J., & Mannheimer, P. D. (2006). Forehead pulse oximetry: Headband use helps alleviate false low readings likely related to venous pulsation artifact. Anesthesiology, 105(6), 1111-1116. Aho, A. J., Lyytikainen, L. P., Yli-Hankala, A., Kamata, K., & Jantti, V. (2011). Explaining entropy responses after a noxious stimulus, with or without neuromuscular blocking agents, by means of the raw electroencephalographic and electromyographic characteristics. British Journal of Anaesthesia, 106(1), 69-76. Aho, A. J., Yli-Hankala, A., Lyytikäinen, L. -., & Jäntti, V. (2009). Facial muscle activity, response entropy, and state entropy indices during noxious stimuli in propofol– nitrous oxide or propofol–nitrous oxide–remifentanil anaesthesia without neuromuscular block. British Journal of Anaesthesia, 102(2), 227-233. doi:10.1093/bja/aen356 Ahonen, J., Jokela, R., Uutela, K., & Huiku, M. (2007). Surgical stress index reflects surgical stress in gynaecological laparoscopic day-case surgery. British Journal of Anaesthesia, 98(4), 456-461. Aime, I., Verroust, N., Masson-Lefoll, C., Taylor, G., Laloe, P. A., Liu, N., et al. (2006). Does monitoring bispectral index or spectral entropy reduce sevoflurane use? Anesthesia and Analgesia, 103(6), 1469-1477. Aissou, M., Snauwaert, A., Dupuis, C., Atchabahian, A., Aubrun, F., & Beaussier, M. (2012). Objective assessment of the immediate postoperative analgesia using pupillary reflex measurement: A prospective and observational study. Anesthesiology, 116(5), 1006-1012. Alian, A. A., & Shelley, K. H. (2014). Photoplethysmography. Best Practice & Research.Clinical Anaesthesiology, 28(4), 395-406.

Allen, J. (2007). Photoplethysmography and its application in clinical physiological measurement. Physiological Measurement, 28(3), R1. Ameloot, K., Palmers, P. J., & Malbrain, M. L. (2015). The accuracy of noninvasive cardiac output and pressure measurements with finger cuff: A concise review. Current Opinion in Critical Care, 21(3), 232-239. Anderson, R. E., & Jakobsson, J. G. (2004). Entropy of EEG during anaesthetic induction: A comparative study with propofol or nitrous oxide as sole agent. British Journal of Anaesthesia, 92(2), 167-170. doi:10.1093/bja/aeh036 Andrews, P. J. D., Ackerman, W. E., Juneja, M. J., & Vogel, R. (1995). Facial sensory evoked muscle potentials during labour. A continuous objective monitor of adequacy of analgesia? Anaesthesia, 50(1), 9-13. doi:10.1111/j.1365- 2044.1995.tb04504.x Annabi, E. H., & Barker, S. J. (2009). Severe methemoglobinemia detected by pulse oximetry. Anesthesia and Analgesia, 108(3), 898-899. doi:10.1213/ane.0b013e318172af73 Antognini, J. F., & Carstens, E. (2002). In vivo characterization of clinical anaesthesia and its components. British Journal of Anaesthesia, 89(1), 156-166. Antognini, J. F., & Schwartz, K. (1993). Exaggerated anesthetic requirements in the preferentially anesthetized brain. Anesthesiology, 79(6), 1244-1249. Aoyagi, T., Fuse, M., Kobayashi, N., Machida, K., & Miyasaka, K. (2007). Multiwavelength pulse oximetry: Theory for the future. Anesthesia and Analgesia, 105(6 Suppl), S53-8, tables of contents. Applegate, I.,Richard L., Dorotta, I. L., Wells, B., Juma, D., & Applegate, P. M. (2016). The relationship between oxygen reserve index and arterial partial pressure of oxygen during surgery. Anesthesia & Analgesia, 123(3), 626-633. doi:10.1213/ANE.0000000000001262 Arakaki, L. S. L., Ciesielski, W. A., Thackray, B. D., Feigl, E. O., & Schenkman, K. A. (2010). Simultaneous optical spectroscopic measurement of hemoglobin and myoglobin saturations and cytochrome aa3 oxidation in vivo. Applied Spectroscopy, 64(9), 973-979. doi:10.1366/000370210792434387 ASA House of Delegates. (2015). Standards for basic anesthetic monitoring; ASA committee of origin: Standards and practice parameters. Retrieved March/8, 2016, from http://www.asahq.org/quality-and-practice-management/standards-and-guidelines Association for the Advancement of Medical instrumentation. (2013). Non-invasive sphygmomanometers—Part 2: Clinical investigation of automated measurement type, ANSI/ AAMI/ ISO 81060–2:2013. (pp. 1-22) Atkins, J. H., & Mandel, J. E. (2014). Performance of masimo rainbow acoustic monitoring for tracking changing respiratory rates under laryngeal mask airway general anesthesia for surgical procedures in the operating room: A prospective observational study. Anesthesia and Analgesia, 119(6), 1307-1314. Avidan, M. S., Jacobsohn, E., Glick, D., Burnside, B. A., Zhang, L., Villafranca, A., et al. (2011). Prevention of intraoperative awareness in a high-risk surgical population. The New England Journal of Medicine, 365(7), 591-600. Avidan, M. S., Zhang, L., Burnside, B. A., Finkel, K. J., Searleman, A. C., Selvidge, J. A., et al. (2008). Anesthesia awareness and the bispectral index. N Engl J Med, 358(11), 1097-1108.

Barker, S. J., Curry, J., Redford, D., & Morgan, S. (2006). Measurement of carboxyhemoglobin and methemoglobin by pulse oximetry: A human volunteer study. Anesthesiology, 105(5), 892-897. doi:10.1097/00000542-200611000-00008 Barker, S. J., Shander, A., & Ramsay, M. A. (2016). Continuous noninvasive hemoglobin monitoring: A measured response to a critical review. Anesthesia & Analgesia, 122(2), 565-572. doi:10.1213/ANE.0000000000000605 Bartels, K., & Thiele, R. H. (2015). Advances in photoplethysmography: Beyond arterial oxygen saturation. Canadian Journal of Anaesthesia = Journal Canadien d'Anesthesie, 62(12), 1313-1328. Bause, G. S., Weintraub, A. C., & Tanner, G. E. (1986). Skin avulsion during oscillometry. Journal of Clinical Monitoring, 2(4), 262-263. Ben-Israel, N., Kliger, M., Zuckerman, G., Katz, Y., & Edry, R. (2013). Monitoring the nociception level: A multi-parameter approach. Journal of Clinical Monitoring and Computing, 27(6), 659-668. Bergmann, I., Gohner, A., Crozier, T. A., Hesjedal, B., Wiese, C. H., Popov, A. F., et al. (2013). Surgical pleth index-guided remifentanil administration reduces remifentanil and propofol consumption and shortens recovery times in outpatient anaesthesia. British Journal of Anaesthesia, 110(4), 622-628. Biais, M., Cottenceau, V., Petit, L., Masson, F., Cochard, J., & Sztark, F. (2011). Impact of norepinephrine on the relationship between pleth variability index and pulse pressure variations in ICU adult patients. Critical Care, 15(4), R168-R168. doi:10.1186/cc10310 Bischoff, P., Kochs, E., Droese, D., Meyer-Moldenhauer, W. H., & Schulte am Esch, J. (1993). Topographic-quantitative EEG-analysis of the paradoxical arousal reaction. EEG changes during urologic surgery using isoflurane/ N2O anesthesia. Der Anaesthesist, 42(3), 142. Bohnhorst, B., Peter, C. S., & Poets, C. F. (2000). Pulse oximeters' reliability in detecting hypoxemia and bradycardia: Comparison between a conventional and two new generation oximeters. Critical Care Medicine, 28(5), 1565-1568. Boselli, E., Bouvet, L., Begou, G., Dabouz, R., Davidson, J., Deloste, J. Y., et al. (2014). Prediction of immediate postoperative pain using the analgesia/nociception index: A prospective observational study. British Journal of Anaesthesia, 112(4), 715-721. Boselli, E., Bouvet, L., Begou, G., Torkmani, S., & Allaouchiche, B. (2015). Prediction of hemodynamic reactivity during total intravenous anesthesia for suspension laryngoscopy using analgesia/nociception index (ANI): A prospective observational study. Minerva Anestesiologica, 81(3), 288-297. Boselli, E., Daniela-Ionescu, M., Begou, G., Bouvet, L., Dabouz, R., Magnin, C., et al. (2013). Prospective observational study of the non-invasive assessment of immediate postoperative pain using the analgesia/nociception index (ANI). British Journal of Anaesthesia, 111(3), 453-459. Bouillon, T. W., Bruhn, J., Radulescu, L., Andresen, C., Shafer, T. J., Cohane, C., et al. (2004). Pharmacodynamic interaction between propofol and remifentanil regarding hypnosis, tolerance of laryngoscopy, bispectral index, and electroencephalographic approximate entropy. Anesthesiology, 100(6), 1353-1372. Briggs, M., & Closs, J. S. (1999). A descriptive study of the use of visual analogue scales and verbal rating scales for the assessment of postoperative pain in orthopedic patients. Journal of Pain and Symptom Management, 18(6), 438-446.

Broch, O., Bein, B., Gruenewald, M., Höcker, J., Schöttler, J., Meybohm, P., et al. (2011). Accuracy of the pleth variability index to predict fluid responsiveness depends on the perfusion index: Plethysmography and fluid responsiveness. Acta Anaesthesiologica Scandinavica, 55(6), 686-693. doi:10.1111/j.1399-6576.2011.02435.x Broucqsault-Dedrie, C., De Jonckheere, J., Jeanne, M., & Nseir, S. (2016). Measurement of heart rate variability to assess pain in sedated critically ill patients: A prospective observational study. PloS One, 11(1), e0147720. Bruhn, J., Myles, P. S., Sneyd, R., & Struys, M. M. (2006). Depth of anaesthesia monitoring: What's available, what's validated and what's next? British Journal of Anaesthesia, 97(1), 85-94. Budidha, K., & Kyriacou, P. A. (2015). Investigation of photoplethysmography and arterial blood oxygen saturation from the ear-canal and the finger under conditions of artificially induced hypothermia. Paper presented at the pp. 7954-7957. doi:10.1109/EMBC.2015.7320237 Budidha, K., & Kyriacou, P. A. (2014). The human ear canal: Investigation of its suitability for monitoring photoplethysmographs and arterial oxygen saturation. Physiological Measurement, 35(2), 111-128. Buist, M., Bernard, S., Nguyen, T. V., Moore, G., & Anderson, J. (2004). Association between clinically abnormal observations and subsequent in-hospital mortality: A prospective study. Resuscitation, 62(2), 137-141. Byon, H.., Lim, C.., Lee, J.., Park, Y. ., Kim, H.., Kim, C.., et al. (2013). Prediction of fluid responsiveness in mechanically ventilated children undergoing neurosurgery. British Journal of Anaesthesia, 110(4), 586-591. doi:10.1093/bja/aes467 Callaghan, F. J., Babbs, C. F., Bourland, J. D., & Geddes, L. A. (1984). The relationship between arterial pulse-wave velocity and pulse frequency at different pressures. Journal of Medical Engineering & Technology, 8(1), 15-18. Cannesson, M. (2007). Respiratory variations in pulse oximetry plethysmographic waveform amplitude to predict fluid responsiveness in the operating room. Anesthesiology, 106(6), 1105-1111. doi:10.1097/01.anes.0000267593.72744.20 Cannesson, M., Desebbe, O., Rosamel, P., Delannoy, B., Robin, J., Bastien, O., et al. (2008). Pleth variability index to monitor the respiratory variations in the pulse oximeter plethysmographic waveform amplitude and predict fluid responsiveness in the operating theatre. British Journal of Anaesthesia, 101(2), 200-206. doi:10.1093/bja/aen133 Casati, A., Squicciarini, G., Baciarello, M., Putzu, M., Salvadori, A., & Fanelli, G. (2007). Forehead reflectance oximetry: A clinical comparison with conventional digit sensors during laparotomic and laparoscopic abdominal surgery. Journal of Clinical Monitoring and Computing, 21(5), 271-276. Chan, M. T. V., Cheng, B. C. P., Lee, T. M. C., Gin, T., CODA Trial Grp, & CODA Trial Group. (2013). BIS-guided anesthesia decreases postoperative delirium and cognitive decline. Journal of Neurosurgical Anesthesiology, 25(1), 33-42. doi:10.1097/ANA.0b013e3182712fba Charlton, P. H. . Retrieved August/10, 2016, from http://peterhcharlton.github.io/RRest/synthetic_dataset.html Charlton, P. H., Bonnici, T., Tarassenko, L., Clifton, D. A., Beale, R., & Watkinson, P. J. (2016). An assessment of algorithms to estimate respiratory rate from the electrocardiogram and photoplethysmogram. Physiological Measurement, 37(4), 610.

Chen, X., Thee, C., Gruenewald, M., Wnent, J., Illies, C., Hoecker, J., et al. (2010). Comparison of surgical stress index-guided analgesia with standard clinical practice during routine general anesthesia: A pilot study. Anesthesiology, 112(5), 1175-1183. Chen, X., Thee, C., Gruenewald, M., Ilies, C., Höcker, J., Hanss, R., et al. (2012). Correlation of surgical pleth index with stress hormones during propofol- remifentanil anaesthesia. The Scientific World Journal, 2012, 879158-8. doi:10.1100/2012/879158 Chernik, D. A., Chernik, D. A., Gillings, D., Gillings, D., Laine, H., Laine, H., et al. (1990). Validity and reliability of the observer's assessment of alertness/sedation scale: Study with intravenous midazolam. Journal of Clinical Psychopharmacology, 10(4), 244- 251. Chhabra, A., Subramaniam, R., Srivastava, A., Prabhakar, H., Kalaivani, M., & Paranjape, S. (2016). Spectral entropy monitoring for adults and children undergoing general anaesthesia. The Cochrane Database of Systematic Reviews, 3, CD010135. Chim, H., Bakri, K., & Moran, S. L. (2015). Complications related to radial artery occlusion, radial artery harvest, and arterial lines. Hand Clinics, 31(1), 93-100. Cividjian, A., Martinez, J. Y., Combourieu, E., Precloux, P., Beraud, A. M., Rochette, Y., et al. (2007). Beat-by-beat cardiovascular index to predict unexpected intraoperative movement in anesthetized unparalyzed patients: A retrospective analysis. Journal of Clinical Monitoring and Computing, 21(2), 91-101. Clayton, D. G., Webb, R. K., Ralston, A. C., Duthie, D., & Runciman, W. B. (1991). Pulse oximeter probes. A comparison between finger, nose, ear and forehead probes under conditions of poor perfusion. Anaesthesia, 46(4), 260-265. Cockings, J. G., Webb, R. K., Klepper, I. D., Currie, M., & Morgan, C. (1993). The Australian incident monitoring study. Blood pressure monitoring--applications and limitations: An analysis of 2000 incident reports. Anaesthesia and Intensive Care, 21(5), 565-569. Collins, J., Rudenski, A., Gibson, J., Howard, L., & O'Driscoll, R. (2015). Relating oxygen partial pressure, saturation and content: The haemoglobin-oxygen dissociation curve. Breathe (Sheffield, England), 11(3), 194-201. doi:10.1183/20734735.001415 Colombo, R., Raimondi, F., Rech, R., Castelli, A., Fossali, T., Marchi, A., et al. (2015). Surgical pleth index guided analgesia blunts the intraoperative sympathetic response to laparoscopic cholecystectomy. Minerva Anestesiologica, 81(8), 837-845. Colombo, R., Raimondi, F., Corona, A., Rivetti, I., Pagani, F., Porta, V. D., et al. (2014). Comparison of the surgical pleth index with autonomic nervous system modulation on cardiac activity during general anaesthesia: A randomised cross-over study. European Journal of Anaesthesiology, 31(2), 76-84. doi:10.1097/01.EJA.0000436116.06728.b3 Connelly, M. A., Brown, J. T., Kearns, G. L., Anderson, R. A., St Peter, S. D., & Neville, K. A. (2014). Pupillometry: A non-invasive technique for pain assessment in paediatric patients. Archives of Disease in Childhood, 99(12), 1125-1131. Constant, I., Nghe, M. C., Boudet, L., Berniere, J., Schrayer, S., Seeman, R., et al. (2006). Reflex pupillary dilatation in response to skin incision and alfentanil in children anaesthetized with sevoflurane: A more sensitive measure of noxious stimulation than the commonly used variables. British Journal of Anaesthesia, 96(5), 614-619. Davies, D. J., Clancy, M., Lighter, D., Balanos, G. M., Lucas, S. J. E., Dehghani, H., et al. (2016). Frequency-domain vs continuous-wave near-infrared spectroscopy devices:

A comparison of clinically viable monitors in controlled hypoxia. Journal of Clinical Monitoring and Computing, doi:10.1007/s10877-016-9942-5 Davis, C., Motoyama. (2011). 4. cardiovascular physiology. Smith's anesthesia for infants and children (8th ed., pp. 100-107). Philadephia, USA: Elsevier inc. Daymont, C., Bonafide, C. P., & Brady, P. W. (2015). Heart rates in hospitalized children by age and body temperature. Pediatrics, 135(5), e1173-81. De Jonckheere, J., Rommel, D., Nandrino, J. L., Jeanne, M., & Logier, R. (2012). Heart rate variability analysis as an index of emotion regulation processes: Interest of the analgesia nociception index (ANI). Conference Proceedings: .Annual International Conference of the IEEE Engineering in Medicine and Biology Society.IEEE Engineering in Medicine and Biology Society. Annual Conference, 2012, 3432-3435. DeBarros, M., Causey, M. W., Chesley, P., & Martin, M. (2015). Reliability of continuous non-invasive assessment of hemoglobin and fluid responsiveness: Impact of obesity and abdominal insufflation pressures. Obesity Surgery, 25(7), 1142-1148. doi:10.1007/s11695-014-1505-6 Deitch, K., Miner, J., Chudnofsky, C. R., Dominici, P., & Latta, D. (2010). Does end tidal CO2 monitoring during emergency department procedural sedation and analgesia with propofol decrease the incidence of hypoxic events? A randomized, controlled trial. Annals of Emergency Medicine, 55(3), 258-264. Delerme, S., Renault, R., Le Manach, Y., Lvovschi, V., Bendahou, M., Riou, B., et al. (2007). Variations in pulse oximetry plethysmographic waveform amplitude induced by passive leg raising in spontaneously breathing volunteers. The American Journal of Emergency Medicine, 25(6), 637-642. Delgado-Gonzalo, R., Parak, J., Tarniceriu, A., Renevey, P., Bertschi, M., & Korhonen, I. (2015). Evaluation of accuracy and reliability of PulseOn optical heart rate monitoring device. Conference Proceedings: Annual International Conference of the IEEE Engineering in Medicine and Biology Society.IEEE Engineering in Medicine and Biology Society.Annual Conference, 2015, 430-433. Desgranges, F. ., Desebbe, O., Ghazouani, A., Gilbert, K., Keller, G., Chiari, P., et al. (2011). Influence of the site of measurement on the ability of plethysmographic variability index to predict fluid responsiveness. British Journal of Anaesthesia, 107(3), 329-335. doi:10.1093/bja/aer165 Ding, F., Fan, W., Zhang, R., Zhang, Q., Li, Y., & Wang, J. (2011). Validation of the noninvasive assessment of central blood pressure by the SphygmoCor and omron devices against the invasive catheter measurement. American Journal of Hypertension, 24(12), 1306-1311. doi:10.1038/ajh.2011.145 Domino, K. B., Posner, K. L., Caplan, R. A., & Cheney, F. W. (1999). Awareness during anesthesia: A closed claims analysis. Anesthesiology, 90(4), 1053-1061. doi:10.1097/00000542-199904000-00019 Donati, A., Damiani, E., Domizi, R., Scorcella, C., Carsetti, A., Tondi, S., et al. (2016). Near-infrared spectroscopy for assessing tissue oxygenation and microvascular reactivity in critically ill patients: A prospective observational study. Critical Care, 20(1) doi:10.1186/s13054-016-1500-5 Douniama, C., Sauter, C., & Couronne, R. (2009). Blood pressure tracking capabilities of pulse transit times in different arterial segments: A clinical evaluation. Computers in Cardiology 2009; 36:201−204.

Dubois, P. E., Putz, L., Jamart, J., Marotta, M. L., Gourdin, M., & Donnez, O. (2014). Deep neuromuscular block improves surgical conditions during laparoscopic hysterectomy: A randomised controlled trial. European Journal of Anaesthesiology, 31(8), 430-436. Edmonds Jr, H. L., Edmonds Jr, H. L., Couture, L. J., Couture, L. J., Stolzy, S. L., Stolzy, S. L., et al. (1986). Quantitative surface electromyography in anesthesia and critical care. International Journal of Clinical Monitoring and Computing, 3(2), 135-145. doi:10.1007/BF01880767 Edmonds Jr, H. L., Edmonds Jr, H. L., Paloheimo, M., & Paloheimo, M. (1985). Computerized monitoring of the EMG and EEG during anesthesia. an evaluation of the anesthesia and brain activity monitor (ABM R). International Journal of Clinical Monitoring and Computing, 1(4), 201-210. doi:10.1007/BF01720184 Edmonds, Z. V., Mower, W. R., Lovato, L. M., & Lomeli, R. (2002). The reliability of vital sign measurements. Annals of Emergency Medicine, 39(3), 233-237. Edry, R., Recea, V., Dikust, Y., & Sessler, D. I. (2016). Preliminary intraoperative validation of the nociception level index: A noninvasive nociception monitor. Anesthesiology, 125(1), 193-203. doi:10.1097/ALN.0000000000001130 Egan, T. D., Minto, C. F., Hermann, D. J., Barr, J., Muir, K. T., & Shafer, S. L. (1996). Remifentanil versus alfentanil: Comparative pharmacokinetics and pharmacodynamics in healthy adult male volunteers. Anesthesiology, 84(4), 821-833. doi:10.1097/00000542-199604000-00009 Eger, E. I., Saidman, L. J., & Brandstater, B. (1965). Minimum alveolar anesthetic concentration: A standard of anesthetic potency. Anesthesiology, 26(6), 756-763. Eisen, L., Fine, I., & Goldinov, L. (2013). Wearable pulse oximetry device Google Patents. El Hor, T., Van Der Linden, P., De Hert, S., Melot, C., & Bidgoli, J. (2013). Impact of entropy monitoring on volatile anesthetic uptake. Anesthesiology, 118(4), 868-873. El-Amrawy, F., & Nounou, M. I. (2015). Are currently available wearable devices for activity tracking and heart rate monitoring accurate, precise, and medically beneficial? Healthcare Informatics Research, 21(4), 315-320. Ellerkmann, R. K., Grass, A., Hoeft, A., & Soehle, M. (2013). The response of the composite variability index to a standardized noxious stimulus during propofol- remifentanil anesthesia. Anesthesia and Analgesia, 116(3), 580-588. Epstein, C. D., & Haghenbeck, K. T. (2014). Bedside assessment of tissue oxygen saturation monitoring in critically ill adults: An integrative review of the literature. Critical Care Research and Practice, 2014, 709683. Fahlenkamp, A. V., Krebber, F., Rex, S., Grottke, O., Fries, M., Rossaint, R., et al. (2010). Bispectral index monitoring during balanced xenon or sevoflurane anaesthesia in elderly patients. European Journal of Anaesthesiology, 27(10), 906-911. doi:10.1097/EJA.0b013e32833d1289 Fanconi, S., & Tschupp, A. (1994). Accuracy of a new transmittance-reflectance pulse oximetry sensor in critically ill neonates. Critical Care Medicine, 22(7), 1142-1146. Feiner, J. R., Rollins, M. D., Sall, J. W., Eilers, H., Au, P., & Bickler, P. E. (2013). Accuracy of carboxyhemoglobin detection by pulse CO-oximetry during hypoxemia. Anesthesia & Analgesia, 117(4), 847-858. doi:10.1213/ANE.0b013e31828610a0 Feissel, M., Kalakhy, R., Banwarth, P., Badie, J., Pavon, A., Faller, J., et al. (2013). Plethysmographic variation index predicts fluid responsiveness in ventilated patients in the early phase of septic shock in the emergency department: A pilot study. Journal of Critical Care, 28(5), 634-639. doi:10.1016/j.jcrc.2013.03.011

Feissel, M., Teboul, J., Merlani, P., Badie, J., Faller, J., & Bendjelid, K. (2007). Plethysmographic dynamic indices predict fluid responsiveness in septic ventilated patients. Intensive Care Medicine, 33(6), 993-999. doi:10.1007/s00134-007-0602-6 Ferenets, R. (2007). Behavior of entropy/complexity measures of the electroencephalogram during propofol-induced sedation: Dose-dependent effects of remifentanil. Anesthesiology, 106(4), 696-706. doi:10.1097/01.anes.0000264790.07231.2d Fischer, M. O., Fornier, W., Hanouz, J. L., & Fellahi, J. L. (2015). Cephalic and digital pulse oximetry in cardiac surgery: A comparative pilot study with arterial oximetry. European Journal of Anaesthesiology, 32(1), 60-61. Folke, M., Cernerud, L., Ekstrom, M., & Hok, B. (2003). Critical review of non-invasive respiratory monitoring in medical care. Medical & Biological Engineering & Computing, 41(4), 377-383. Forget, P., Lois, F., & De Kock, M. (2010). Goal-directed fluid management based on the pulse oximeter-derived pleth variability index reduces lactate levels and improves fluid management. Anesthesia and Analgesia, 111(4), 910-914. doi:10.1213/ANE.0b013e3181eb624f Fortin, J., Wellisch, A., & Maier, K. (2013). CNAP - evolution of continuous non-invasive arterial blood pressure monitoring. Biomedizinische Technik.Biomedical Engineering, Frasca, D., Mounios, H., Giraud, B., Boisson, M., Debaene, B., & Mimoz, O. (2015). Continuous monitoring of haemoglobin concentration after in‐vivo adjustment in patients undergoing surgery with blood loss. Anaesthesia, 70(7), 803-809. doi:10.1111/anae.13028 Frasca, D., Dahyot-Fizelier, C., Catherine, K., Levrat, Q., Debaene, B., & Mimoz, O. (2011). Accuracy of a continuous noninvasive hemoglobin monitor in intensive care unit patients. Critical Care Medicine, 39(10), 2277-2282. doi:10.1097/CCM.0b013e3182227e2d Fukushima, H., Kawanaka, H., Bhuiyan, M. S., & Oguri, K. (2012). Estimating heart rate using wrist-type photoplethysmography and acceleration sensor while running. Conference Proceedings : Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society.Annual Conference, 2012, 2901-2904. Gall, O., Champigneulle, B., Schweitzer, B., Deram, T., Maupain, O., Montmayeur Verchere, J., et al. (2015). Postoperative pain assessment in children: A pilot study of the usefulness of the analgesia nociception index. British Journal of Anaesthesia, 115(6), 890-895. Gallagher, J. D. (1999). Pacer-induced artifact in the bispectral index during cardiac surgery [6]. Anesthesiology, 90(2), 636-636. doi:10.1097/00000542-199902000-00050 Galvagno Jr, S. M., Hu, P., Yang, S., Gao, C., Hanna, D., Shackelford, S., et al. (2015). Accuracy of continuous noninvasive hemoglobin monitoring for the prediction of blood transfusions in trauma patients. Journal of Clinical Monitoring and Computing, 29(6), 815-821. doi:10.1007/s10877-015-9671-1 Garde, A., Karlen, W., Ansermino, J. M., & Dumont, G. A. (2014). Estimating respiratory and heart rates from the correntropy spectral density of the photoplethysmogram. PloS One, 9(1), e86427.

Gaucher, A., Frasca, D., Mimoz, O., & Debaene, B. (2012). Accuracy of respiratory rate monitoring by capnometry using the capnomask(R) in extubated patients receiving supplemental oxygen after surgery. British Journal of Anaesthesia, 108(2), 316-320. Gesche, H., Grosskurth, D., Kuchler, G., & Patzak, A. (2012). Continuous blood pressure measurement by using the pulse transit time: Comparison to a cuff-based method. European Journal of Applied Physiology, 112(1), 309-315. Gibbs, F. A., Gibbs, E. L., & Lennox, W. G. (1937). Effect on the electro-encephalogram of certain drugs which influence nervous activity. Archives of Internal Medicine, 60(1), 154-166. doi:10.1001/archinte.1937.00180010159012 Gomez-Cambronero, J. (2001). The oxygen dissociation curve of hemoglobin: Bridging the gap between biochemistry and physiology. Journal of Chemical Education, 78(6), 757- 759. Goodman, P. (2009). Predicting ischemic brain injury after intraoperative cardiac arrest during cardiac surgery using the BIS monitor. J Clin Anesth, 21(8), 609-612. doi:10.1016/j.jclinane.2009.01.014 Goto, T., Nakata, Y., Saito, H., Ishiguro, Y., Niimi, Y., Suwa, K., et al. (2000). Bispectral analysis of the electroencephalogram does not predict responsiveness to verbal command in patients emerging from xenon anaesthesia. British Journal of Anaesthesia, 85(3), 359-363. doi:10.1093/bja/85.3.359 Grosse-Sundrup, M., Henneman, J. P., Sandberg, W. S., Bateman, B. T., Uribe, J. V., Nguyen, N. T., et al. (2012). Intermediate acting non-depolarizing neuromuscular blocking agents and risk of postoperative respiratory complications: Prospective propensity score matched cohort study. BMJ (Clinical Research Ed.), 345, e6329. Gruenewald, M., & Dempfle, A. (2017). Analgesia/nociception monitoring for opioid guidance: Meta-analysis of randomized clinical trials. Minerva Anestesiologica, 83(2), 200-213. Gruenewald, M., Herz, J., Schoenherr, T., Thee, C., Steinfath, M., & Bein, B. (2014). Measurement of the nociceptive balance by analgesia nociception index (ANI) and surgical pleth index (SPI) during sevoflurane - remifentanil anaesthesia. Minerva Anestesiologica, Gruenewald, M., Herz, J., Schoenherr, T., Thee, C., Steinfath, M., & Bein, B. (2015). Measurement of the nociceptive balance by analgesia nociception index and surgical pleth index during sevoflurane-remifentanil anesthesia. Minerva Anestesiologica, 81(5), 480-489. Gruenewald, M., Meybohm, P., Ilies, C., Hocker, J., Hanss, R., Scholz, J., et al. (2009). Influence of different remifentanil concentrations on the performance of the surgical stress index to detect a standardized painful stimulus during sevoflurane anaesthesia. British Journal of Anaesthesia, 103(4), 586-593. Gruenewald, M., Willms, S., Broch, O., Kott, M., Steinfath, M., & Bein, B. (2014). Sufentanil administration guided by surgical pleth index vs standard practice during sevoflurane anaesthesia: A randomized controlled pilot study. British Journal of Anaesthesia, 112(5), 898-905. Gruenewald, M., Zhou, J., Schloemerkemper, N., Meybohm, P., Weiler, N., Tonner, P. H., et al. (2007). M-entropy guidance vs standard practice during propofol-remifentanil anaesthesia: A randomised controlled trial. Anaesthesia, 62(12), 1224-1229. Guglielminotti, J., Grillot, N., Paule, M., Mentre, F., Servin, F., Montravers, P., et al. (2015). Prediction of movement to surgical stimulation by the pupillary dilatation reflex amplitude evoked by a standardized noxious test. Anesthesiology, 122(5), 985-993.

Guglielminotti, J., Mentre, F., Gaillard, J., Ghalayini, M., Montravers, P., & Longrois, D. (2013). Assessment of pain during labor with pupillometry: A prospective observational study. Anesthesia and Analgesia, 116(5), 1057-1062. Guignard, B. (2006). Monitoring analgesia. Best Practice & Research.Clinical Anaesthesiology, 20(1), 161-180. Haas, S., Trepte, C., Hinteregger, M., Fahje, R., Sill, B., Herich, L., et al. (2012). Prediction of volume responsiveness using pleth variability index in patients undergoing cardiac surgery after cardiopulmonary bypass. Journal of Anesthesia, 26(5), 696-701. doi:10.1007/s00540-012-1410-x Haenggi, M., Ypparila-Wolters, H., Buerki, S., Schlauri, R., Korhonen, I., Takala, J., et al. (2009). Auditory event-related potentials, bispectral index, and entropy for the discrimination of different levels of sedation in intensive care unit patients. Anesthesia and Analgesia, 109(3), 807-816. doi:10.1213/ane.0b013e3181acc85d Hampson, N. B. (2012). Noninvasive pulse CO-oximetry expedites evaluation and management of patients with carbon monoxide poisoning. American Journal of Emergency Medicine, 30(9), 2021-2024. doi:10.1016/j.ajem.2012.03.026 Hamunen, K., Kontinen, V., Hakala, E., Talke, P., Paloheimo, M., & Kalso, E. (2012). Effect of pain on autonomic nervous system indices derived from photoplethysmography in healthy volunteers. British Journal of Anaesthesia, 108(5), 838-844. doi:10.1093/bja/aes001 Hans, P., Bonhomme, V., Benmansour, H., Dewandre, P. Y., Brichant, J. F., & Lamy, M. (2001). Effect of nitrous oxide on the bispectral index and the 95% spectral edge frequency of the electroencephalogram during surgery. Anaesthesia, 56(10), 999-1002. doi:10.1046/j.1365-2044.2001.01974-4.x Hans, P., Verscheure, S., Uutela, K., Hans, G., & Bonhomme, V. (2012). Effect of a fluid challenge on the surgical pleth index during stable propofol-remifentanil anaesthesia: Fluid challenge and SPI. Acta Anaesthesiologica Scandinavica, 56(6), 787- 796. doi:10.1111/j.1399-6576.2011.02639.x Hans, P., Dewandre, P., Brichant, J. F., & Bonhomme, V. (2005). Comparative effects of ketamine on bispectral index and spectral entropy of the electroencephalogram under sevoflurane anaesthesia. British Journal of Anaesthesia, 94(3), 336-340. doi:10.1093/bja/aei047 Hart, S. M., Buchannan, C. R., & Sleigh, J. W. (2009). A failure of M-entropy™ to correctly detect burst suppression leading to sevoflurane overdosage. Anaesthesia and Intensive Care, 37(6), 1002-1004. Hemmerling, T., & Fortier, J. (2002). Falsely increased bispectral index values in a series of patients undergoing cardiac surgery using forced-air-warming therapy of the head. Anesthesia and Analgesia, 95(2), 322-323. doi:10.1213/01.ANE.0000022369.06971.7D Hennig,A., & Patzak,A. (2013). Continuous blood pressure measurement using pulse transit time. Somnologie, 17(2), 104-110. Hernandez-Silveira, M., Ahmed, K., Ang, S. S., Zandari, F., Mehta, T., Weir, R., et al. (2015). Assessment of the feasibility of an ultra-low power, wireless digital patch for the continuous ambulatory monitoring of vital signs. BMJ Open, 5(5), e006606-2014- 006606. Hers, V., Corbugy, D., Joslet, I., Hermant, P., Demarteau, J., Delhougne, B., et al. (2013). New concept using passive infrared (PIR) technology for a contactless detection of

breathing movement: A pilot study involving a cohort of 169 adult patients. Journal of Clinical Monitoring and Computing, 27(5), 521-529. Hirota, K., Kubota, T., Ishihara, H., & Matsuki, A. (1999). The effects of nitrous oxide and ketamine on the bispectral index and 95% spectral edge frequency during propofol- fentanyl anaesthesia. European Journal of Anaesthesiology, 16(11), 779-783. doi:10.1046/j.1365-2346.1999.00585.x Hocker, J., Broch, O., Grasner, J. T., Gruenewald, M., Ilies, C., Steinfath, M., et al. (2010). Surgical stress index in response to pacemaker stimulation or atropine. British Journal of Anaesthesia, 105(2), 150-154. Hoffman, W., Cunningham, F., James, M., Baughman, V., & Albrecht, R. (1993). Effects of remifentanil, a new short-acting opioid, on cerebral blood flow, brain electrical activity, and intracranial pressure in dogs anesthetized with isoflurane and nitrous oxide. Anesthesiology, 79(1), 107-113. Hohn, A., Defosse, J. M., Becker, S., Steffen, C., Wappler, F., & Sakka, S. G. (2013). Non- invasive continuous arterial pressure monitoring with nexfin does not sufficiently replace invasive measurements in critically ill patients. British Journal of Anaesthesia, 111(2), 178-184. Huiku, M., Uutela, K., van Gils, M., Korhonen, I., Kymalainen, M., Merilainen, P., et al. (2007). Assessment of surgical stress during general anaesthesia. British Journal of Anaesthesia, 98(4), 447-455. Hynson, J. M., Sessler, D. I., Belani, K., Washington, D., McGuire, J., Merrifield, B., et al. (1992). Thermoregulatory vasoconstriction during propofol/nitrous oxide anesthesia in humans: Threshold and oxyhemoglobin saturation. Anesthesia & Analgesia, 75(6), 947-952. doi:10.1213/00000539-199212000-00013 Hyttel-Sorensen, S., Hessel, T. W., & Greisen, G. (2014). Peripheral tissue oximetry: Comparing three commercial near-infrared spectroscopy oximeters on the forearm. Journal of Clinical Monitoring and Computing, 28(2), 149-155. doi:10.1007/s10877-013- 9507-9 Hyttel-Sorensen, S., Sorensen, L. C., Riera, J., & Greisen, G. (2011). Tissue oximetry: A comparison of mean values of regional tissue saturation, reproducibility and dynamic range of four NIRS-instruments on the human forearm. Biomedical Optics Express, 2(11), 3047-3057. doi:10.1364/BOE.2.003047 Ilies, C., Bauer, M., Berg, P., Rosenberg, J., Hedderich, J., Bein, B., et al. (2012). Investigation of the agreement of a continuous non-invasive arterial pressure device in comparison with invasive radial artery measurement. British Journal of Anaesthesia, 108(2), 202-210. Ilies, C., Grudev, G., Hedderich, J., Renner, J., Steinfath, M., Bein, B., et al. (2015). Comparison of a continuous noninvasive arterial pressure device with invasive measurements in cardiovascular postsurgical intensive care patients: A prospective observational study. European Journal of Anaesthesiology, 32(1), 20-28. Ilies, C., Gruenewald, M., Ludwigs, J., Thee, C., Hocker, J., Hanss, R., et al. (2010). Evaluation of the surgical stress index during spinal and general anaesthesia. British Journal of Anaesthesia, 105(4), 533-537. Ilies, C., Ludwigs, J., Gruenewald, M., Thee, C., Hanf, J., Hanss, R., et al. (2012). The effect of posture and anaesthetic technique on the surgical pleth index. Anaesthesia, 67(5), 508-513.

Isik, K., Unsal, A., Kalayci, A., & Durmus, E. (2011). Comparison of three pain scales after impacted third molar surgery. Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology, and Endodontics, 112(6), 715-718. Isnardon, S., Vinclair, M., Genty, C., Hebrard, A., Albaladejo, P., & Payen, J. F. (2013). Pupillometry to detect pain response during general anaesthesia following unilateral popliteal sciatic nerve block: A prospective, observational study. European Journal of Anaesthesiology, 30(7), 429-434. Janelle, G. M., & Gravenstein, N. (2006). An accuracy evaluation of the T-line tensymeter (continuous noninvasive blood pressure management device) versus conventional invasive radial artery monitoring in surgical patients. Anesthesia and Analgesia, 102(2), 484-490. Jeanne, M., Delecroix, M., De Jonckheere, J., Keribedj, A., Logier, R., & Tavernier, B. (2014). Variations of the analgesia nociception index during propofol anesthesia for total knee replacement. The Clinical Journal of Pain, 30(12), 1084-1088. Jensen, L. A., Onyskiw, J. E., & Prasad, N. G. (1998). Meta-analysis of arterial oxygen saturation monitoring by pulse oximetry in adults. Heart & Lung : The Journal of Critical Care, 27(6), 387-408. Jess, G., Pogatzki-Zahn, E. M., Zahn, P. K., & Meyer-Frieem, C. H. (2016). Monitoring heart rate variability to assess experimentally induced pain using the analgesia nociception index: A randomised volunteer study. European Journal of Anaesthesiology, 33(2), 118-125. Johansson, A., Ahlstrom, C., Lanne, T., & Ask, P. (2006). Pulse wave transit time for monitoring respiration rate. Medical & Biological Engineering & Computing, 44(6), 471- 478. Jones, D. A., DeVita, M. A., & Bellomo, R. (2011). Rapid-response teams. The New England Journal of Medicine, 365(2), 139-146. Jubran, A. (2015). Pulse oximetry. Critical Care (London, England), 19, 272-015-0984-8. Kallio, H., Lindberg, L. I., Majander, A. S., Uutela, K. H., Niskanen, M. L., & Paloheimo, M. P. (2008). Measurement of surgical stress in anaesthetized children. British Journal of Anaesthesia, 101(3), 383-389. Kantor, E., Montravers, P., Longrois, D., & Guglielminotti, J. (2014). Pain assessment in the postanaesthesia care unit using pupillometry: A cross-sectional study after standard anaesthetic care. European Journal of Anaesthesiology, 31(2), 91-97. Karlen, W., Garde, A., Myers, D., Scheffer, C., Ansermino, J. M., & Dumont, G. A. (2014). Respiratory rate assessment from photoplethysmographic imaging. Conference Proceedings : ..Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual Conference, 2014, 5397-5400. Keehn, L., Milne, L., McNeill, K., Chowienczyk, P., & Sinha, M. D. (2014). Measurement of pulse wave velocity in children: Comparison of volumetric and tonometric sensors, brachial-femoral and carotid-femoral pathways. Journal of Hypertension, Kertai, M., B-Unaware Study Grp, & B-Unaware Study Group. (2010). Association of perioperative risk factors and cumulative duration of low bispectral index with intermediate-term mortality after cardiac surgery in the B-unaware trial. Anesthesiology, 112(5), 1116-1127. doi:10.1097/ALN.0b013e3181d5e0a3 Kertai, M. D., Palanca, B. J. A., Pal, N., Burnside, B. A., Zhang, L., Sadiq, F., et al. (2011). Bispectral index monitoring, duration of bispectral index below 45, patient risk

100

factors, and intermediate-term mortality after noncardiac surgery in the B-unaware trial. Anesthesiology, 114(3), 545-556. doi:10.1097/ALN.0b013e31820c2b57 Khan, Y., Ostfeld, A. E., Lochner, C. M., Pierre, A., & Arias, A. C. (2016). Monitoring of vital signs with flexible and wearable medical devices. Advanced Materials (Deerfield Beach, Fla.), Kim, S. H., Lilot, M., Sidhu, K. S., Rinehart, J., Yu, Z., Canales, C., et al. (2014). Accuracy and precision of continuous noninvasive arterial pressure monitoring compared with invasive arterial pressure: A systematic review and meta-analysis. Anesthesiology, 120(5), 1080-1097. Komori, T., Eguchi, K., Hoshide, S., Williams, B., & Kario, K. (2013). Comparison of wrist-type and arm-type 24-h blood pressure monitoring devices for ambulatory use. Blood Pressure Monitoring, 18(1), 57-62. Korhonen, I., & Yli-Hankala, A. (2009). Photoplethysmography and nociception. Acta Anaesthesiologica Scandinavica, 53(8), 975-985. Kortelainen, J., Koskinen, M., Mustola, S., & Seppanen, T. (2009). Effect of remifentanil on the nonlinear electroencephalographic entropy parameters in propofol anesthesia. Conference Proceedings: .Annual International Conference of the IEEE Engineering in Medicine and Biology Society.IEEE Engineering in Medicine and Biology Society.Conference, 2009, 4994-4997. doi:10.1109/IEMBS.2009.5334612 Kortelainen, J., Koskinen, M., Mustola, S., & Seppänen, T. (2008). Remifentanil modifies the relation of electroencephalographic spectral changes and clinical endpoints in propofol anesthesia. Anesthesiology, 109(2), 198-205. doi:10.1097/ALN.0b013e31817f5bfc Krauss, B., & Hess, D. R. (2007). Capnography for procedural sedation and analgesia in the emergency department. Annals of Emergency Medicine, 50(2), 172-181. Kwon, Y., Hwang, S. M., Lee, J. J., & Kim, J. H. (2015). The effect of dexmedetomidine as an adjuvant to ropivacaine on the bispectral index for supraclavicular brachial plexus block. Korean Journal of Anesthesiology, 68(1), 32-36. doi:10.4097/kjae.2015.68.1.32 Laitio, R. (2008). Bispectral index, entropy, and quantitative electroencephalogram during single-agent xenon anesthesia. Anesthesiology, 108(1), 63-70. doi:10.1097/01.anes.0000296106.52472.a6 Landsverk, S. (2008). Poor agreement between respiratory variations in pulse oximetry photoplethysmographic waveform amplitude and pulse pressure in intensive care unit patients. Anesthesiology, 109(5), 849-855. doi:10.1097/ALN.0b013e3181895f9f Langwieser, N., Prechtl, L., Meidert, A. S., Hapfelmeier, A., Bradaric, C., Ibrahim, T., et al. (2015). Radial artery applanation tonometry for continuous noninvasive arterial blood pressure monitoring in the cardiac intensive care unit. Clinical Research in Cardiology: Official Journal of the German Cardiac Society, Larson, M. D., & Behrends, M. (2015). Portable infrared pupillometry: A review. Anesthesia and Analgesia, 120(6), 1242-1253. Le Guen, M., Jeanne, M., Sievert, K., Al Moubarik, M., Chazot, T., Laloe, P. A., et al. (2012). The analgesia nociception index: A pilot study to evaluation of a new pain parameter during labor. International Journal of Obstetric Anesthesia, 21(2), 146-151. Ledowski, T., Albus, S., Stein, J., & Macdonald, B. (2011). Skin conductance for monitoring of acute pain in adult postoperative patients: Influence of electrode surface area and sampling time. Journal of Clinical Monitoring and Computing, 25(6), 371- 376.

101

Ledowski, T., Averhoff, L., Tiong, W. S., & Lee, C. (2014). Analgesia nociception index (ANI) to predict intraoperative haemodynamic changes: Results of a pilot investigation. Acta Anaesthesiologica Scandinavica, 58(1), 74-79. Ledowski, T., Burke, J., & Hruby, J. (2016). Surgical pleth index: Prediction of postoperative pain and influence of arousal. British Journal of Anaesthesia, 117(3), 371- 374. doi:10.1093/bja/aew226 Ledowski, T., Pascoe, E., Ang, B., Schmarbeck, T., Clarke, M. W., Fuller, C., et al. (2010). Monitoring of intra-operative nociception: Skin conductance and surgical stress index versus stress hormone plasma levels. Anaesthesia, 65(10), 1001-1006. Lee, J., Matsumura, K., Yamakoshi, K., Rolfe, P., Tanaka, S., & Yamakoshi, T. (2013). Comparison between red, green and blue light reflection photoplethysmography for heart rate monitoring during motion. Conference Proceedings: .Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual Conference, 2013, 1724-1727. Lee, J., Florian, J. P., & Chon, K. H. (2011). Respiratory rate extraction from pulse oximeter and electrocardiographic recordings. Physiological Measurement, 32(11), 1763- 1773. doi:10.1088/0967-3334/32/11/S04 Leonard, P. A., Douglas, J. G., Grubb, N. R., Clifton, D., Addison, P. S., & Watson, J. N. (2006). A fully automated algorithm for the determination of respiratory rate from the photoplethysmogram. Journal of Clinical Monitoring and Computing, 20(1), 33-36. Leslie, K., Myles, P. S., Forbes, A., & Chan, M. T. V. (2010). The effect of bispectral index monitoring on long-term survival in the B-aware trial. Anesthesia and Analgesia, 110(3), 816-822. doi:10.1213/ANE.0b013e3181c3bfb2 Li, R. T., Kling, S. R., Salata, M. J., Cupp, S. A., Sheehan, J., & Voos, J. E. (2016). Wearable performance devices in sports medicine. Sports Health, 8(1), 74-78. Lien, C. A., & Kopman, A. F. (2014). Current recommendations for monitoring depth of neuromuscular blockade. Current Opinion in Anaesthesiology, 27(6), 616-622. Lima, A., van Bommel, J., Jansen, T. C., Ince, C., & Bakker, J. (2009). Low tissue oxygen saturation at the end of early goal-directed therapy is associated with worse outcome in critically ill patients. Critical Care (London, England), 13 Suppl 5, S13. Lin, C. C., Jawan, B., de Villa, M. V., Chen, F. C., & Liu, P. P. (2001). Blood pressure cuff compression injury of the radial nerve. Journal of Clinical Anesthesia, 13(4), 306-308. Lindberg, L. G., Ugnell, H., & Oberg, P. A. (1992). Monitoring of respiratory and heart rates using a fibre-optic sensor. Medical & Biological Engineering & Computing, 30(5), 533-537. Lindholm, M., Granath, F., Eriksson, L. I., & Sandin, R. (2011). Malignant disease within 5 years after surgery in relation to duration of sevoflurane anesthesia and time with bispectral index under 45. Anesthesia and Analgesia, 113(4), 778-783. doi:10.1213/ANE.0b013e31821f950e Lindholm, M., Träff, S., Granath, F., Greenwald, S. D., Ekbom, A., Lennmarken, C., et al. (2009). Mortality within 2 years after surgery in relation to low intraoperative bispectral index values and preexisting malignant disease. Anesthesia and Analgesia, 108(2), 508-512. doi:10.1213/ane.0b013e31818f603c Losa-Iglesias, M. E., Becerro-de-Bengoa-Vallejo, R., & Becerro-de-Bengoa-Losa, K. R. (2014). Reliability and concurrent validity of a peripheral pulse oximeter and health- app system for the quantification of heart rate in healthy adults. Health Informatics Journal,

102

Loupec, T., Nanadoumgar, H., Frasca, D., Petitpas, F., Laksiri, L., Baudouin, D., et al. (2011). Pleth variability index predicts fluid responsiveness in critically ill patients. Critical Care Medicine, 39(2), 294-299. doi:10.1097/CCM.0b013e3181ffde1c Lovett, P. B., Buchwald, J. M., Stürmann, K., & Bijur, P. (2005). The vexatious vital: Neither clinical measurements by nurses nor an electronic monitor provides accurate measurements of respiratory rate in triage. Annals of Emergency Medicine, 45(1), 68-76. Luginbuhl, M., Schumacher, P. M., Vuilleumier, P., Vereecke, H., Heyse, B., Bouillon, T. W., et al. (2010). Noxious stimulation response index: A novel anesthetic state index based on hypnotic-opioid interaction. Anesthesiology, 112(4), 872-880. Luginbuhl, M., Ypparila-Wolters, H., Rufenacht, M., Petersen-Felix, S., & Korhonen, I. (2007). Heart rate variability does not discriminate between different levels of haemodynamic responsiveness during surgical anaesthesia. British Journal of Anaesthesia, 98(6), 728-736. Macknet, M. R., Allard, M., Applegate, I., Richard L., & Rook, J. (2010). The accuracy of noninvasive and continuous total hemoglobin measurement by pulse CO-oximetry in human subjects undergoing hemodilution. Anesthesia & Analgesia, 111(6), 1424- 1426. doi:10.1213/ANE.0b013e3181fc74b9 Madsen, M. V., Istre, O., Staehr-Rye, A. K., Springborg, H. H., Rosenberg, J., Lund, J., et al. (2015). Postoperative shoulder pain after laparoscopic hysterectomy with deep neuromuscular blockade and low-pressure pneumoperitoneum: A randomised controlled trial. European Journal of Anaesthesiology, Maeda, Y., Sekine, M., & Tamura, T. (2011a). The advantages of wearable green reflected photoplethysmography. Journal of Medical Systems, 35(5), 829-834. doi:10.1007/s10916-010-9506-z Maeda, Y., Sekine, M., & Tamura, T. (2011b). Relationship between measurement site and motion artifacts in wearable reflected photoplethysmography. Journal of Medical Systems, 35(5), 969-976. doi:10.1007/s10916-010-9505-0 Mainzer Jr, J., & Mainzer, J. (1979). Awareness, muscle relaxants and balanced anaesthesia. Canadian Anaesthetists' Society Journal, 26(5), 386-393. doi:10.1007/BF03006453 Martinez, J. Y., Wey, P. F., Lions, C., Cividjian, A., Rabilloud, M., Bissery, A., et al. (2010). A beat-by-beat cardiovascular index, CARDEAN: A prospective randomized assessment of its utility for the reduction of movement during colonoscopy. Anesthesia and Analgesia, 110(3), 765-772. Martini, C. H., Boon, M., Broens, S. J., Hekkelman, E. F., Oudhoff, L. A., Buddeke, A. W., et al. (2015). Ability of the nociception level, a multiparameter composite of autonomic signals, to detect noxious stimuli during propofol-remifentanil anesthesia. Anesthesiology, 123(3), 524-534. Mashour, G. A., & Avidan, M. S. (2015). Intraoperative awareness: Controversies and non- controversies. British Journal of Anaesthesia, 115 Suppl 1, i20-i26. Mashour, G. A., Shanks, A., Tremper, K. K., Kheterpal, S., Turner, C. R., Ramachandran, S. K., et al. (2012). Prevention of intraoperative awareness with explicit recall in an unselected surgical population: A randomized comparative effectiveness trial. Anesthesiology, 117(4), 717-725. Mashour, G. A., Wang, L. Y. ., Turner, C. R., Vandervest, J. C., Shanks, A., & Tremper, K. K. (2009). A retrospective study of intraoperative awareness with methodological implications. Anesthesia and Analgesia, 108(2), 521-526. doi:10.1213/ane.0b013e3181732b0c

103

Mathews, D. M., Clark, L., Johansen, J., Matute, E., & Seshagiri, C. V. (2012). Increases in electroencephalogram and electromyogram variability are associated with an increased incidence of intraoperative somatic response. Anesthesia and Analgesia, 114(4), 759-770. Matsumura, K., Rolfe, P., Lee, J., & Yamakoshi, T. (2014). iPhone 4s photoplethysmography: Which light color yields the most accurate heart rate and normalized pulse volume using the iPhysioMeter application in the presence of motion artifact? Plos One, 9(3), e91205. doi:10.1371/journal.pone.0091205 McCulloch, T. J., & Thompson, C. L. (2010). Failure of M-entropy. Anaesthesia and Intensive Care, 38(3), 597-598. Meidert, A. S., Huber, W., Hapfelmeier, A., Schofthaler, M., Muller, J. N., Langwieser, N., et al. (2013). Evaluation of the radial artery applanation tonometry technology for continuous noninvasive blood pressure monitoring compared with central aortic blood pressure measurements in patients with multiple organ dysfunction syndrome. Journal of Critical Care, 28(6), 908-912. Meidert, A. S., Huber, W., Muller, J. N., Schofthaler, M., Hapfelmeier, A., Langwieser, N., et al. (2014). Radial artery applanation tonometry for continuous non-invasive arterial pressure monitoring in intensive care unit patients: Comparison with invasively assessed radial arterial pressure. British Journal of Anaesthesia, 112(3), 521- 528. Mendelson, Y., & McGinn, M. J. (1991). Skin reflectance pulse oximetry: In vivo measurements from the forearm and calf. Journal of Clinical Monitoring, 7(1), 7-12. Meng, X., Zang, G., Fan, L., Zheng, L., Dai, J., Wang, X., et al. (2013). Non-invasive monitoring of blood pressure using the Philips intellivue MP50 monitor cannot replace invasive blood pressure techniques in surgery patients under general anesthesia. Experimental and Therapeutic Medicine, 6(1), 9-14. Mesquida, J., Gruartmoner, G., & Espinal, C. (2013). Skeletal muscle oxygen saturation (StO2) measured by near-infrared spectroscopy in the critically ill patients. BioMed Research International, 2013, 1-8. doi:10.1155/2013/502194 Messina, A., Wang, M., Ward, M., Wilker, C., Smith, B., Vezina, D., et al. (2016). Anaesthetic interventions for prevention of awareness during surgery. Cochrane Database of Systematic Reviews, 2016(10) doi:10.1002/14651858.CD007272.pub2 Migeon, A., Desgranges, F. P., Chassard, D., Blaise, B. J., De Queiroz, M., Stewart, A., et al. (2013). Pupillary reflex dilatation and analgesia nociception index monitoring to assess the effectiveness of regional anesthesia in children anesthetised with sevoflurane. Paediatric Anaesthesia, 23(12), 1160-1165. Miller, R. D., Ward, T. A., McCulloch, C. E., & Cohen, N. H. (2012). Does a digital regional nerve block improve the accuracy of noninvasive hemoglobin monitoring? Journal of Anesthesia, 26(6), 845-850. doi:10.1007/s00540-012-1452-0 Monk, T. G., Saini, V., Weldon, B. C., & Sigl, J. C. (2005). Anesthetic management and one-year mortality after noncardiac surgery. Anesthesia & Analgesia, 100(1), 4-10. doi:10.1213/01.ANE.0000147519.82841.5E Monnet, X., Lamia, B., & Teboul, J. L. (2005). Pulse oximeter as a sensor of fluid responsiveness: Do we have our finger on the best solution? Critical Care (London, England), 9(5), 429-430.

104

Mourad, A., Carney, S. L., Gillies, A., Jones, B., Nanra, R., & Trevillian, P. (2003). Arm position and blood pressure: A risk factor for hypertension? Journal of Human Hypertension, 17(6), 389-395. doi:10.1038/sj.jhh.1001563 Mukkamala, R., Hahn, J., Inan, O. T., Mestha, L. K., Kim, C., Toreyin, H., et al. (2015). Toward ubiquitous blood pressure monitoring via pulse transit time: Theory and practice. IEEE Transactions on Biomedical Engineering, 62(8), 1879-1901. doi:10.1109/TBME.2015.2441951 Mustola, S., Parkkari, T., Uutela, K., Huiku, M., Kymalainen, M., & Toivonen, J. (2010). Performance of surgical stress index during sevoflurane-fentanyl and isoflurane- fentanyl anesthesia. Anesthesiology Research and Practice, 2010, 10.1155/2010/810721. Epub 2010 Apr 6. Myles, P. S., Leslie, K., McNeil, J., Forbes, A., & Chan, M. T. (2004). Bispectral index monitoring to prevent awareness during anaesthesia: The B-aware randomised controlled trial. Lancet (London, England), 363(9423), 1757-1763. Nair, D., Tan, S. Y., Gan, H. W., Lim, S. F., Tan, J., Zhu, M., et al. (2008). The use of ambulatory tonometric radial arterial wave capture to measure ambulatory blood pressure: The validation of a novel wrist-bound device in adults. Journal of Human Hypertension, 22(3), 220-222. Natalini, G., Rosano, A., Taranto, M., Faggian, B., Vittorielli, E., & Bernardini, A. (2006). Arterial versus plethysmographic dynamic indices to test responsiveness for testing fluid administration in hypotensive patients: A clinical trial. Anesthesia and Analgesia, 103(6), 1478-1484. doi:10.1213/01.ane.0000246811.88524.75 Nesseler, N., Frenel, J. V., Launey, Y., Morcet, J., Malledant, Y., & Seguin, P. (2012). Pulse oximetry and high-dose vasopressors: A comparison between forehead reflectance and finger transmission sensors. Intensive Care Medicine, 38(10), 1718-1722. Ng, K. G., Ting, C. M., Yeo, J. H., Sim, K. W., Peh, W. L., Chua, N. H., et al. (2004). Progress on the development of the MediWatch ambulatory blood pressure monitor and related devices. Blood Pressure Monitoring, 9(3), 149-165. Nilsson, L., Goscinski, T., Johansson, A., Lindberg, L. G., & Kalman, S. (2006). Age and gender do not influence the ability to detect respiration by photoplethysmography. Journal of Clinical Monitoring and Computing, 20(6), 431-436. Nilsson, L., Goscinski, T., Kalman, S., Lindberg, L. G., & Johansson, A. (2007). Combined photoplethysmographic monitoring of respiration rate and pulse: A comparison between different measurement sites in spontaneously breathing subjects. Acta Anaesthesiologica Scandinavica, 51(9), 1250-1257. Nilsson, L., Johansson, A., & Kalman, S. (2003a). Macrocirculation is not the sole determinant of respiratory induced variations in the reflection mode photoplethysmographic signal. Physiological Measurement, 24(4), 925-937. doi:10.1088/0967-3334/24/4/009 Nilsson, L., Johansson, A., & Kalman, S. (2003b). Respiratory variations in the reflection mode photoplethysmographic signal. Relationships to peripheral venous pressure. Medical & Biological Engineering & Computing, 41(3), 249-254. doi:10.1007/BF02348428 Nilsson, L., Johansson, A., & Kalman, S. (2005). Respiration can be monitored by photoplethysmography with high sensitivity and specificity regardless of anaesthesia and ventilatory mode. Acta Anaesthesiologica Scandinavica, 49(8), 1157-1162. Nitzan, M., Romem, A., & Koppel, R. (2014). Pulse oximetry: Fundamentals and technology update. Medical Devices (Auckland, N.Z.), 7, 231-239.

105

Oksala, N., Vehkaoja, A., & Melkoniemi, S. (2015). Method, device and arrangement for determining pulse transit time Oksala, N., & Liuhanen, S. (2015). Method and device for the detection of respiratory rate Google Patents. Orena, E. F., King, A. B., & Hughes, C. G. (2016). The role of anesthesia in the prevention of postoperative delirium: A systematic review. Minerva Anestesiologica, 82(6), 669- 683. Ott, C., Haetinger, S., Schneider, M. P., Pauschinger, M., & Schmieder, R. E. (2012). Comparison of two noninvasive devices for measurement of central systolic blood pressure with invasive measurement during cardiac catheterization. Journal of Clinical Hypertension (Greenwich, Conn.), 14(9), 575-579. Ozcan, M. S., Ozcan, M. D., Khan, Q. S., Thompson, D. M., & Chetty, P. K. (2010). Does nitrous oxide affect bispectral index and state entropy when added to a propofol versus sevoflurane anesthetic? Journal of Neurosurgical Anesthesiology, 22(4), 309-315. doi:10.1097/ANA.0b013e3181e4b7c8 Paloheimo, M. P., Sahanne, S., & Uutela, K. H. (2010). Autonomic nervous system state: The effect of general anaesthesia and bilateral tonsillectomy after unilateral infiltration of lidocaine. British Journal of Anaesthesia, 104(5), 587-595. Palve, H. (1992a). Comparison of reflection and transmission pulse oximetry after open- heart surgery. Critical Care Medicine, 20(1), 48-51. Palve, H. (1992b). Reflection and transmission pulse oximetry during compromised peripheral perfusion. Journal of Clinical Monitoring, 8(1), 12-15. Parak, J., Tarniceriu, A., Renevey, P., Bertschi, M., Delgado-Gonzalo, R., & Korhonen, I. (2015). Evaluation of the beat-to-beat detection accuracy of PulseOn wearable optical heart rate monitor. Conference Proceedings: Annual International Conference of the IEEE Engineering in Medicine and Biology Society.IEEE Engineering in Medicine and Biology Society.Annual Conference, 2015, 8099-8102. Park, J. H., Lim, B. G., Kim, H., Lee, I. O., Kong, M. H., & Kim, N. S. (2015). Comparison of surgical pleth index-guided analgesia with conventional analgesia practices in children: A randomized controlled trial. Anesthesiology, 122(6), 1280-7. Patel, M., Asch, D., & Volpp, K. (2015). Wearable devices as facilitators, not drivers, of health behavior change. JAMA, 313(5), 459-460. Patino, M., Schultz, L., Hossain, M., Moeller, J., Mahmoud, M., Gunter, J., et al. (2014). Trending and accuracy of noninvasive hemoglobin monitoring in pediatric perioperative patients. Anesthesia & Analgesia, 119(4), 920-925. doi:10.1213/ANE.0000000000000369 Patzak, A., Mendoza, Y., Gesche, H., & Konermann, M. (2015). Continuous blood pressure measurement using the pulse transit time: Comparison to intra-arterial measurement. Blood Pressure, 24(4), 217-221. Paulus, J., Roquilly, A., Beloeil, H., Theraud, J., Asehnoune, K., & Lejus, C. (2013). Pupillary reflex measurement predicts insufficient analgesia before endotracheal suctioning in critically ill patients. Critical Care (London, England), 17(4), R161. Pedersen, T., Nicholson, A., Hovhannisyan, K., Moller, A. M., Smith, A. F., & Lewis, S. R. (2014). Pulse oximetry for perioperative monitoring. The Cochrane Database of Systematic Reviews, 3, CD002013.

106

Pellicer, A., & Bravo, M. d. C. (2011). Near-infrared spectroscopy: A methodology-focused review. Seminars in Fetal and Neonatal Medicine, 16(1), 42-49. doi:10.1016/j.siny.2010.05.003 Perkins, G. D., McAuley, D. F., Giles, S., Routledge, H., & Gao, F. (2003). Do changes in pulse oximeter oxygen saturation predict equivalent changes in arterial oxygen saturation? Critical Care (London, England), 7(4), R67. Pesonen, A., Kauppila, T., Tarkkila, P., Sutela, A., Niinisto, L., & Rosenberg, P. H. (2009). Evaluation of easily applicable pain measurement tools for the assessment of pain in demented patients. Acta Anaesthesiologica Scandinavica, 53(5), 657-664. Phattraprayoon, N., Sardesai, S., Durand, M., & Ramanathan, R. (2012). Accuracy of pulse oximeter readings from probe placement on newborn wrist and ankle. Journal of Perinatology: Official Journal of the California Perinatal Association, 32(4), 276-280. Pruett, J. D., Bourland, J. D., & Geddes, L. A. (1988). Measurement of pulse-wave velocity using a beat-sampling technique. Annals of Biomedical Engineering, 16(4), 341-347. Punjasawadwong, Y., Phongchiewboon, A., & Bunchungmongkol, N. (2014). Bispectral index for improving anaesthetic delivery and postoperative recovery. The Cochrane Database of Systematic Reviews, 6, CD003843. Quasha, A. L., Eger II, E. I., & Tinker, J. H. (1980). Determination and applications of MAC. Anesthesiology, 53(4), 315-334. Radtke, F. M., Franck, M., Lendner, J., Krüger, S., Wernecke, K. D., & Spies, C. D. (2013). Monitoring depth of anaesthesia in a randomized trial decreases the rate of postoperative delirium but not postoperative cognitive dysfunction. British Journal of Anaesthesia, 110(1), i98-i105. doi:10.1093/bja/aet055 Ramesh, V., & Rao, G. (2006). Does nitrous oxide affect bispectral index in patients under isoflurane anesthesia? Journal of Neurosurgical Anesthesiology, 18(4), 267-268. doi:10.1097/00008506-200610000-00014 Rampil, I. J. (1994). Anesthetic potency is not altered after hypothermic spinal cord transection in rats. Anesthesiology, 80(3), 606-610. Rampil, I. J., Kim, J., Lenhardt, R., Negishi, C., & Sessler, D. I. (1998). Bispectral EEG index during nitrous oxide administration. Anesthesiology, 89(3), 671-677. doi:10.1097/00000542-199809000-00017 Ramsay, M. A., Usman, M., Lagow, E., Mendoza, M., Untalan, E., & De Vol, E. (2013). The accuracy, precision and reliability of measuring ventilatory rate and detecting ventilatory pause by rainbow acoustic monitoring and capnometry. Anesthesia and Analgesia, 117(1), 69-75. Rantanen, M., Yli-Hankala, A., van Gils, M., Ypparila-Wolters, H., Takala, P., Huiku, M., et al. (2006). Novel multiparameter approach for measurement of nociception at skin incision during general anaesthesia. British Journal of Anaesthesia, 96(3), 367-376. Reisner, A. (2008). Utility of the photoplethysmogram in circulatory monitoring. Anesthesiology, 108(5), 950-958. doi:10.1097/ALN.0b013e31816c89e1 Ribezzo, S., Spina, E., Di Bartolomeo, S., & Sanson, G. (2014). Noninvasive techniques for blood pressure measurement are not a reliable alternative to direct measurement: A randomized crossover trial in ICU. TheScientific World Journal, 2014, 353628. Robinson, D. H., & Toledo, A. H. (2012). Historical development of modern anesthesia. Journal of Investigative Surgery, 25(3), 141-149. doi:10.3109/08941939.2012.690328 Rogobete, A. F., Sandesc, D., Cradigati, C. A., Sarandan, M., Papurica, M., Popovici, S. E., et al. (2017). Implications of entropy and surgical pleth index-guided general anaesthesia on clinical outcomes in critically ill polytrauma patients. A prospective

107

observational non-randomized single centre study. Journal of Clinical Monitoring and Computing, doi:10.1007/s10877-017-0059-2 Roizen, M. F., Horrigan, R. W., & Frazer, B. M. (1981). Anesthetic doses blocking adrenergic (stress) and cardiovascular responses to incision - MAC BAR. Anesthesiology, 54(5), 390-398. Romagnoli, S., Ricci, Z., Quattrone, D., Tofani, L., Tujjar, O., Villa, G., et al. (2014). Accuracy of invasive arterial pressure monitoring in cardiovascular patients: An observational study. Critical Care (London, England), 18(6), 644-014-0644-4. Rosenberg, J., & Pedersen, M. H. (1990). Nasal pulse oximetry overestimates oxygen saturation. Anaesthesia, 45(12), 1070-1071. doi:10.1111/j.1365-2044.1990.tb14892.x Rossi, M., Cividjian, A., Fevre, M. C., Oddoux, M. E., Carcey, J., Halle, C., et al. (2012). A beat-by-beat, on-line, cardiovascular index, CARDEAN, to assess circulatory responses to surgery: A randomized clinical trial during spine surgery. Journal of Clinical Monitoring and Computing, 26(6), 441-449. Roth, D., Herkner, H., Schreiber, W., Hubmann, N., Gamper, G., Laggner, A. N., et al. (2011). Accuracy of noninvasive multiwave pulse oximetry compared with carboxyhemoglobin from blood gas analysis in unselected emergency department patients. Annals of Emergency Medicine, 58(1), 74-79. doi:10.1016/j.annemergmed.2010.12.024 Rouche, O., Wolak-Thierry, A., Destoop, Q., Milloncourt, L., Floch, T., Raclot, P., et al. (2013). Evaluation of the depth of sedation in an intensive care unit based on the photo motor reflex variations measured by video pupillometry. Annals of Intensive Care, 3(1), 5-5820-3-5. Ruiz-Rodriguez, J. C., Ruiz-Sanmartin, A., Ribas, V., Caballero, J., Garcia-Roche, A., Riera, J., et al. (2013). Innovative continuous non-invasive cuffless blood pressure monitoring based on photoplethysmography technology. Intensive Care Medicine, 39(9), 1618-1625. Sabourdin, N., Arnaout, M., Louvet, N., Guye, M. L., Piana, F., & Constant, I. (2013). Pain monitoring in anesthetized children: First assessment of skin conductance and analgesia-nociception index at different infusion rates of remifentanil. Paediatric Anaesthesia, 23(2), 149-155. Sabourdin, N., Barrois, J., Louvet, N., Rigouzzo, A., Guye, M. L., Dadure, C., et al. (2017). Pupillometry-guided intraoperative remifentanil administration versus standard practice influences opioid use: A randomized study. Anesthesiology, 127(2), 284-292. doi:10.1097/ALN.0000000000001705 [doi Safar, H., & El-Dash, H. (2015). Pulse oximetry: Could wrist and ankle be alternative placement sites? Clinical Pediatrics, 54(14), 1375-1379. Sahinovic, M. M., Eleveld, D. J., Kalmar, A. F., Heeremans, E. H., De Smet, T., Seshagiri, C. V., et al. (2014). Accuracy of the composite variability index as a measure of the balance between nociception and antinociception during anesthesia. Anesthesia and Analgesia, 119(2), 288-301. Sarafidis, P. A., Georgianos, P. I., Karpetas, A., Bikos, A., Korelidou, L., Tersi, M., et al. (2014). Evaluation of a novel brachial cuff-based oscillometric method for estimating central systolic pressure in hemodialysis patients. American Journal of Nephrology, 40(3), 242-250.

108

Särkelä, M., Mustola, S., Seppänen, T., Koskinen, M., Lepola, P., Suominen, K., et al. (2002). Automatic analysis and monitoring of burst suppression in anesthesia. Journal of Clinical Monitoring and Computing, 17(2), 125-134. Sato, T., Nishinaga, M., Kawamoto, A., Ozawa, T., & Takatsuji, H. (1993). Accuracy of a continuous blood pressure monitor based on arterial tonometry. Hypertension, 21(6 Pt 1), 866-874. Saugel, B., Meidert, A. S., Hapfelmeier, A., Eyer, F., Schmid, R. M., & Huber, W. (2013). Non-invasive continuous arterial pressure measurement based on radial artery tonometry in the intensive care unit: A method comparison study using the T-line TL-200pro device. British Journal of Anaesthesia, 111(2), 185-190. Saugel, B., Meidert, A. S., Langwieser, N., Wagner, J. Y., Fassio, F., Hapfelmeier, A., et al. (2014). An autocalibrating algorithm for non-invasive cardiac output determination based on the analysis of an arterial pressure waveform recorded with radial artery applanation tonometry: A proof of concept pilot analysis. Journal of Clinical Monitoring and Computing, 28(4), 357-362. Schallom, L., Sona, C., McSweeney, M., & Mazuski, J. (2007). Comparison of forehead and digit oximetry in surgical/trauma patients at risk for decreased peripheral perfusion. Heart & Lung : The Journal of Critical Care, 36(3), 188-194. Schein, R. M., Hazday, N., Pena, M., Ruben, B. H., & Sprung, C. L. (1990). Clinical antecedents to in-hospital cardiopulmonary arrest. Chest, 98(6), 1388-1392. Schiffrin, E. L. (2004). Vascular stiffening and arterial compliance. Implications for systolic blood pressure. American Journal of Hypertension, 17(12 Pt 2), 39S-48S. Schmid, M., Prettenthaler, H., Weger, C., & Smolle, K. H. (2013). Evaluation of a novel automated non-invasive pulse pressure variation algorithm. Computers in Biology and Medicine, 43(10), 1583-1589. Schreiber, J. U. (2014). Management of neuromuscular blockade in ambulatory patients. Current Opinion in Anaesthesiology, 27(6), 583-588. Sebel, P. S., Bowdle, T. A., Ghoneim, M. M., Rampil, I. J., Padilla, R. E., Gan, T. J., et al. (2004). The incidence of awareness during anesthesia: A multicenter United States study. Anesthesia and Analgesia, 99(3), 833-839. doi:10.1213/01.ANE.0000130261.90896.6C Seitsonen, E. R., Korhonen, I. K., van Gils, M. J., Huiku, M., Lotjonen, J. M., Korttila, K. T., et al. (2005). EEG spectral entropy, heart rate, photoplethysmography and motor responses to skin incision during sevoflurane anaesthesia. Acta Anaesthesiologica Scandinavica, 49(3), 284-292. Schordes, R., Barbeito, A., Bar-Yousef, S., & Mark, J. (2010). Cardiovascular monitoring. In R. D. Miller (Ed.), Millers's anesthesia (7th ed., pp. 1270-1285). Philadelphia, USA: Churchill Livingstone elsevier. Shepherd, J., Jones, J., Frampton, G., Bryant, J., Baxter, L., & Cooper, K. (2013). Clinical effectiveness and cost-effectiveness of depth of anaesthesia monitoring (E-entropy, bispectral index and narcotrend): A systematic review and economic evaluation. Health Technology Assessment (Winchester, England), 17(34), 1-264. Silke, B., & McAuley, D. (1998). Accuracy and precision of blood pressure determination with the finapres: An overview using re-sampling statistics. Journal of Human Hypertension, 12(6), 403-409. Simpao, A. F., & Gálvez, J. A. (2016). When seconds count, buy more time: The oxygen reserve index and its promising role in patient monitoring and safety. Anesthesiology, 124(4), 750-751. doi:10.1097/ALN.0000000000001036

109

Singleton, R. J., Webb, R. K., Ludbrook, G. L., & Fox, M. A. (1993). The Australian Incident Monitoring Study. Problems associated with vascular access: An analysis of 2000 incident reports. Anaesthesia and Intensive Care, 21(5), 664-669. Slogoff, S., Keats, A. S., & Arlund, C. (1983). On the safety of radial artery cannulation. Anesthesiology, 59(1), 42-47. Smith, M., Dobbs, P., & Eapen, G. (2015). Abnormal bispectral index values associated with the presence of periodic lateralized epileptiform discharges. Journal of Neurosurgical Anesthesiology, 27(1), 73-74. doi:10.1097/ANA.0000000000000069 Smolle, K. H., Schmid, M., Prettenthaler, H., & Weger, C. (2015). The accuracy of the CNAP(R) device compared with invasive radial artery measurements for providing continuous noninvasive arterial blood pressure readings at a medical intensive care unit: A method-comparison study. Anesthesia and Analgesia, 121(6), 1508-1516. Soeding, P., Deppe, M., & Gehring, H. (2010). Pulse-oximetric measurement of prilocaine- induced methemoglobinemia in regional anesthesia. Anesthesia and Analgesia, 111(4), 1065-1068. doi:10.1213/ANE.0b013e3181eb6239 Sola, J., Chetelat, O., & Krauss, J. (2007). On the reliability of pulse oximetry at the sternum. Paper presented at the pp. 1537-1537. doi:10.1109/IEMBS.2007.4352595 Solana, M. J., Lopez-Herce, J., Fernandez, S., Gonzalez, R., Urbano, J., Lopez, J., et al. (2015). Assessment of pain in critically ill children. is cutaneous conductance a reliable tool? Journal of Critical Care, 30(3), 481-485. Sørensen, H., Grocott, H. P., & Secher, N. H. (2016). Near infrared spectroscopy for frontal lobe oxygenation during non-vascular abdominal surgery. Clinical Physiology and Functional Imaging, 36(6), 427-435. doi:10.1111/cpf.12244 Sorvoja, H. (2006). Noninvasive blood pressure pulse detection andblood pressure determination. Faculty of Technology, University of Oulu). Spires, J., Lai, N., Zhou, H., & Saidel, G. M. (2011). Hemoglobin and myoglobin contributions to skeletal muscle oxygenation in response to exercise. Advances in Experimental Medicine and Biology, 701, 347-352. doi:10.1007/978-1-4419-7756-4_47 Staehr-Rye, A. K., Rasmussen, L. S., Rosenberg, J., Juul, P., Lindekaer, A. L., Riber, C., et al. (2014). Surgical space conditions during low-pressure laparoscopic cholecystectomy with deep versus moderate neuromuscular blockade: A randomized clinical study. Anesthesia and Analgesia, 119(5), 1084-1092. Stoelting, R. K., Longnecker, D. E., & Eger 2nd, E. I. (1970). Minimum alveolar concentrations in man on awakening from methoxyflurane, halothane, ether and fluroxene anesthesia: MAC awake. Anesthesiology, 33(1), 5-9. Stone, K., Fryer, S., Ryan, T., & Stoner, L. (2016). The validity and reliability of continuous-wave near-infrared spectroscopy for the assessment of leg blood volume during an orthostatic challenge. Atherosclerosis, 251, 234-239. doi:10.1016/j.atherosclerosis.2016.06.030 Strehle, E. M., & Gray, W. K. (2013). Comparison of skin conductance measurements and subjective pain scores in children with minor injuries. Acta Paediatrica (Oslo, Norway : 1992), 102(11), e502-6. Struys, M. M., Vanpeteghem, C., Huiku, M., Uutela, K., Blyaert, N. B., & Mortier, E. P. (2007). Changes in a surgical stress index in response to standardized pain stimuli during propofol-remifentanil infusion. British Journal of Anaesthesia, 99(3), 359-367.

110

Sugino, S., Kanaya, N., Mizuuchi, M., Nakayama, M., & Namiki, A. (2004). Forehead is as sensitive as finger pulse oximetry during general anesthesia. Canadian Journal of Anaesthesia = Journal Canadien D'Anesthesie, 51(5), 432-436. Sun, S., & Huang, S. (2014). Role of pleth variability index for predicting hypotension after spinal anesthesia for cesarean section. INTERNATIONAL JOURNAL OF OBSTETRIC ANESTHESIA, 23(4), 324-329. doi:10.1016/j.ijoa.2014.05.011 Szental, J. A., Webb, A., Weeraratne, C., Campbell, A., Sivakumar, H., & Leong, S. (2015). Postoperative pain after laparoscopic cholecystectomy is not reduced by intraoperative analgesia guided by analgesia nociception index (ANI(R)) monitoring: A randomized clinical trial. British Journal of Anaesthesia, 114(4), 640-645. Szmuk, P., Steiner, J. W., Olomu, P. N., Ploski, R. P., Sessler, D. I., & Ezri, T. (2016). Oxygen reserve index: A novel noninvasive measure of oxygen Reserve—A pilot study. Anesthesiology, 124(4), 779-784. doi:10.1097/ALN.0000000000001009 Takamatsu, I., Ozaki, M., & Kazama, T. (2006). Entropy indices vs the bispectral index for estimating nociception during sevoflurane anaesthesia. British Journal of Anaesthesia, 96(5), 620-626. Talke, P., & Stapelfeldt, C. (2006). Effect of peripheral vasoconstriction on pulse oximetry. Journal of Clinical Monitoring and Computing, 20(5), 305-309. doi:10.1007/s10877-006- 9022-3 Tanaka, P. P., Tanaka, M., & Drover, D. R. (2014). Detection of respiratory compromise by acoustic monitoring, capnography, and brain function monitoring during monitored anesthesia care. Journal of Clinical Monitoring and Computing, 28(6), 561-566. Thee, C., Ilies, C., Gruenewald, M., Kleinschmidt, A., Steinfath, M., & Bein, B. (2015). Reliability of the surgical pleth index for assessment of postoperative pain: A pilot study. European Journal of Anaesthesiology, 32(1), 44-48. doi:10.1097/EJA.0000000000000095 Theilade, S., Joergensen, C., Persson, F., Lajer, M., & Rossing, P. (2012). Ambulatory tonometric blood pressure measurements in patients with diabetes. Diabetes Technology & Therapeutics, 14(6), 453-456. Tirkkonen, J., Nurmi, J., & Hoppu, S. (2014). Medical emergency treatment is here to stay. [Sairaalansisainen ensihoito on tullut jaadakseen] Duodecim; Laaketieteellinen Aikakauskirja, 130(22-23), 2311-2317. Upton, H., Ludbrook, G., Wing, A., & Sleigh, J. (2017). Intraoperative “Analgesia nociception Index”–Guided fentanyl administration during sevoflurane anesthesia in lumbar discectomy and laminectomy: A randomized clinical trial. Anesthesia & Analgesia, 125(1), 81-90. Vakkuri, A., Yli-Hankala, A., Sandin, R., Mustola, S., Hoymork, S., Nyblom, S., et al. (2005). Spectral entropy monitoring is associated with reduced propofol use and faster emergence in propofol-nitrous oxide-alfentanil anesthesia. Anesthesiology, 103(2), 274-279. Vakkuri, A., Yli-Hankala, A., Talja, P., Mustola, S., Tolvanen-Laakso, H., Sampson, T., et al. (2004). Time-frequency balanced spectral entropy as a measure of anesthetic drug effect in central nervous system during sevoflurane, propofol, and thiopental anesthesia. Acta Anaesthesiologica Scandinavica, 48(2), 145-153. Valjus, M., Ahonen, J., Jokela, R., & Korttila, K. (2006). Response entropy is not more sensitive than state entropy in distinguishing the use of esmolol instead of remifentanil in patients undergoing gynaecological laparoscopy. Acta Anaesthesiologica Scandinavica, 50(1), 32-39.

111

Valkenburg, A. J., Niehof, S. P., van Dijk, M., Verhaar, E. J., & Tibboel, D. (2012). Skin conductance peaks could result from changes in vital parameters unrelated to pain. Pediatric Research, 71(4 Pt 1), 375-379. van der Lee, R., Jebbink, L. J., van Herpen, T. H., d'Haens, E. J., Bierhuizen, J., & van Lingen, R. A. (2016). Feasibility of monitoring stress using skin conduction measurements during intubation of newborns. European Journal of Pediatrics, 175(2), 237-243. Vassiliadis, M., Geros, D., & Maria, K. (2007). Awareness despite low spectral entropy values [3]. Anesthesia and Analgesia, 105(2), 535. doi:10.1213/01.ane.0000265663.88180.86 Vegfors, M., Ugnell, H., Hok, B., Oberg, P. A., & Lennmarken, C. (1993). Experimental evaluation of two new sensors for respiratory rate monitoring. Physiological Measurement, 14(2), 171-181. doi:10.1088/0967-3334/14/2/008 Vereecke, H. E. M., Struys, M. M. F., & Mortier, E. P. (2003). A comparison of bispectral index and ARX-derived auditory evoked potential index in measuring the clinical interaction between ketamine and propofol anaesthesia. Anaesthesia, 58(10), 957-961. doi:10.1046/j.1365-2044.2003.03403.x Vesoulis, Z. A., Lust, C. E., Liao, S. M., Trivedi, S. B., & Mathur, A. M. (2016). Early hyperoxia burden detected by cerebral near-infrared spectroscopy is superior to pulse oximetry for prediction of severe retinopathy of prematurity. Journal of Perinatology, doi:10.1038/jp.2016.131 Viertiö-Oja, H., Maja, V., Sarkela, M., Talja, P., Tenkanen, N., Tolvanen-Laakso, H., et al. (2004). Description of the entropy algorithm as applied in the datex-ohmeda S/5 entropy module. Acta Anaesthesiologica Scandinavica, 48(2), 154-161. Visser, K. R., Lamberts, R., Korsten, H. H. M., & Zijlstra, W. G. (1976). Observations on blood flow related electrical impedance changes in rigid tubes. Pflügers Archiv European Journal of Physiology, 366(2-3), 289-291. doi:10.1007/BF00585894 Vivien, B., Di Maria, S., Ouattara, A., Langeron, O., Coriat, P., & Riou, B. (2003). Overestimation of bispectral index in sedated intensive care unit patients revealed by administration of muscle relaxant. Anesthesiology, 99(1), 9-17. doi:10.1097/00000542-200307000-00006 von Dincklage, F., Hackbarth, M., Mager, R., Rehberg, B., & Baars, J. H. (2010). Monitoring of the responsiveness to noxious stimuli during anaesthesia with propofol and remifentanil by using RIII reflex threshold and bispectral index. British Journal of Anaesthesia, 104(2), 201-208. von Dincklage, F., Send, K., Hackbarth, M., Rehberg, B., & Baars, J. H. (2009). Comparison of the nociceptive flexion reflex threshold and the bispectral index as monitors of movement responses to noxious stimuli under propofol mono- anaesthesia. British Journal of Anaesthesia, 102(2), 244-250. von Dincklage, F., Velten, H., Rehberg, B., & Baars, J. H. (2010). Monitoring of the responsiveness to noxious stimuli during sevoflurane mono-anaesthesia by using RIII reflex threshold and bispectral index. British Journal of Anaesthesia, 104(6), 740- 745. Wang, T., Ge, S., Xiong, W., Zhou, P., Cang, J., & Xue, Z. (2013). Effects of different loading doses of dexmedetomidine on bispectral index under stepwise propofol target-controlled infusion. Pharmacology, 91(1-2), 1-6. doi:10.1159/000343634

112

Wang, Y., Tao, J., Dong, Y., Chen, S., Gao, X., Ji, C., et al. (2014). Effect of different levels of systolic blood pressure on brachial-ankle pulse wave velocity. Zhonghua Liu Xing Bing Xue Za Zhi = Zhonghua Liuxingbingxue Zazhi, 35(6), 655-659. Wax, D. B., Rubin, P., & Neustein, S. (2009). A comparison of transmittance and reflectance pulse oximetry during vascular surgery. Anesthesia and Analgesia, 109(6), 1847-1849. Weissman, C. (1990). The metabolic response to stress: An overview and update. Anesthesiology, 73(2), 308-327. Wennervirta, J., Hynynen, M., Koivusalo, A. M., Uutela, K., Huiku, M., & Vakkuri, A. (2008). Surgical stress index as a measure of nociception/antinociception balance during general anesthesia. Acta Anaesthesiologica Scandinavica, 52(8), 1038-1045. Wheeler, P., Hoffman, W. E., Baughman, V. L., & Koenig, H. (2005). Response entropy increases during painful stimulation. Journal of Neurosurgical Anesthesiology, 17(2), 86- 90. Wilson, S. (1993). Facial electromyography and chloral hydrate in the young dental patient. Pediatric Dentistry, 15(5), 343-347. Wimmer, N. J., Scirica, B. M., & Stone, P. H. (2013). The clinical significance of continuous ECG (ambulatory ECG or holter) monitoring of the ST-segment to evaluate ischemia: A review. Progress in Cardiovascular Diseases, 56(2), 195-202. Wolf, A. R. (2012). Effects of regional analgesia on stress responses to pediatric surgery. Paediatric Anaesthesia, 22(1), 19-24. Wolff, H. S. (1969). Automatic measurement of blood pressure without arterial puncture. Proceedings of the Royal Society of Medicine, 62(10), 1019-1022. Won, Y. J., Lim, B. G., Lee, S. H., Park, S., Kim, H., Lee, I. O., et al. (2016). Comparison of relative oxycodone consumption in surgical pleth index-guided analgesia versus conventional analgesia during sevoflurane anesthesia: A randomized controlled trial. Medicine, 95(35), e4743. doi:10.1097/MD.0000000000004743 Wrench, I., Hammon, L., Handa, S., & Mahajan, R. (2015). Changes in pleth variability index and detection of hypotension during spinal anaesthesia for caesarean section. International journal of obstetric anesthesia, 24(4), 388-389. doi:10.1016/j.ijoa.2015.07.001 XU, L., WU, A. -., & YUE, Y. (2009). The incidence of intra-operative awareness during general anesthesia in China: A multi-center observational study. Acta Anaesthesiologica Scandinavica, 53(7), 873-882. doi:10.1111/j.1399-6576.2009.02016.x Yamashina, A., Tomiyama, H., Arai, T., Koji, Y., Yambe, M., Motobe, H., et al. (2003). Nomogram of the relation of brachial-ankle pulse wave velocity with blood pressure. Hypertension Research : Official Journal of the Japanese Society of Hypertension, 26(10), 801-806. Yli-Hankala, A. (2008). Awareness despite low spectral entropy values. Anesthesia and Analgesia, 106(5), 1585; author reply 1586. Yli-Hankala, A., Lindgren, L., Porkkala, T., & Jantti, V. (1993). Nitrous oxide-mediated activation of the EEG during isoflurane anaesthesia in patients. British Journal of Anaesthesia, 70(1), 54-57. doi:10.1093/bja/70.1.54 Yokose, M., Mihara, T., Sugawara, Y., & Goto, T. (2015). The predictive ability of non‐ invasive haemodynamic parameters for hypotension during caesarean section: A prospective observational study. Anaesthesia, 70(5), 555-562. doi:10.1111/anae.12992 Yu, Y., Dong, J., Xu, Z., Shen, H., & Zheng, J. (2015). Pleth variability index-directed fluid management in abdominal surgery under combined general and epidural anesthesia.

113

Journal of Clinical Monitoring and Computing, 29(1), 47-52. doi:10.1007/s10877-014- 9567-5 Zhang, C., Xu, L., Ma, Y., Sun, Y., Li, Y., Zhang, L., et al. (2011). Bispectral index monitoring prevent awareness during total intravenous anesthesia: A prospective, randomized, double-blinded, multi-center controlled trial. Chinese Medical Journal, 124(22), 3664-3669. Zhang, Z., Pi, Z., & Liu, B. (2015). TROIKA: A general framework for heart rate monitoring using wrist-type photoplethysmographic signals during intensive physical exercise. IEEE Transactions on Bio-Medical Engineering, 62(2), 522-531. Zhou, C., Feng, J., Hu, J., & Ye, X. (2016). Study of artifact-resistive technology based on a novel dual photoplethysmography method for wearable pulse rate monitors. Journal of Medical Systems, 40(3), 1-10. doi:10.1007/s10916-015-0412-2 Zhou, J., & Han, Y. (2016). Pleth variability index and respiratory system compliance to direct PEEP settings in mechanically ventilated patients, an exploratory study. SpringerPlus, 5(1), 1371-016-3008-5. eCollection 2016. Zimmermann, M., Feibicke, T., Keyl, C., Prasser, C., Moritz, S., Graf, B. M., et al. (2010). Accuracy of stroke volume variation compared with pleth variability index to predict fluid responsiveness in mechanically ventilated patients undergoing major surgery. European Journal of Anaesthesiology, 27(6), 555-561. doi:10.1097/EJA.0b013e328335fbd1 Zou, G. Y. (2013). Confidence interval estimation for the Bland–Altman limits of agreement with multiple observations per individual. Statistical Methods in Medical Research, 22(6), 630-642.

114

Original publications

115 British Journal of Anaesthesia, 117 (3): 358–64 (2016)

doi: 10.1093/bja/aew215 Paediatrics

Surgical pleth index in children younger than 24 months of age: a randomized double-blinded trial J. Harju1,*, M.-L. Kalliomäki1, H. Leppikangas1, M. Kiviharju1 and A. Yli-Hankala1,2 Downloaded from 1Department of Anaesthesia, Tampere University Hospital, PL2000, Tampere 33521, Finland, and 2Medical School, University of Tampere, Tampere, Finland

*Corresponding author. Email: jarkko.harju@ﬁmnet.ﬁ http://bja.oxfordjournals.org/

Abstract Background: The surgical pleth index (SPI) is a measurement of intraoperative nociception. Evidence of its usability in children is limited. Given that the autonomic nervous system is still developing during the ﬁrst years of life, the performance of the SPI on small children cannot be concluded from studies carried out in older age groups. Methods: Thirty children aged <2 yr, planned for elective open inguinal hernia repair or open correction of undescended testicle, were recruited. The children were randomized into two groups; the saline group received ultrasound-guided saline injection in the ilioinguinal and iliohypogastric nerve region before surgery and ropivacaine after surgery, whereas the block by guest on August 19, 2016 group received the injections in the opposite order. The SPI was recorded blinded and was analysed at the time points of intubation, incision, and when signs of inadequate anti-nociception were observed. Results: There was a signiﬁcant increase in the SPI after intubation (P=0.019) and after incision in the saline group (P=0.048), but not at the time of surgical incision in the block group (P=0.177). An increase in the SPI was also seen at times of clinically apparent inadequate anti-nociception (P=0.008). The between-patient variability of the SPI was large. Conclusions: The SPI is reactive in small children after intubation and after surgical stimuli, but the reactivity of the SPI is rather small, and there is marked inter-individual variability in reactions. The reactivity is blunted by the use of ilioinguinal and iliohypogastric nerve block. Clinical trial registration: NCT02045810.

Key words: monitoring, intraoperative; nociception; paediatrics

Editor’s key points Nociception during general anaesthesia can elicit signiﬁcant autonomic, hormonal, and metabolic changes. Marked changes • The surgical pleth index (SPI) has been developed as a in heart rate, blood pressure, or patient movement during anaes- measure of nociception in adults. thesia are considered signs of inadequate anaesthesia. A variety • It is calculated from analysis of the heart beat interval and of opioids are used during surgery in order to prevent these plethysmographic pulse wave amplitude. changes.1 The use of opioids can lead to signiﬁcant postoperative • The value of the SPI in small children, in whom the sympa- respiratory depression, especially in small children.2 thetic system is not fully developed, is unknown. Traditionally, the signs of inadequate anti-nociception have • The authors found that the SPI did react to noxious stimuli guided opioid administration. The surgical pleth index (SPI, for- in children younger than 2 yr. merly surgical stress index) was originally introduced in 2007.

358 Surgical pleth index in children younger than 24 months | 359

It is calculated from normalized photoplethysmographic wave- given an injection of NaCl 0.9% into the ilioinguinal and iliohypo- form analysis and normalized analysis of the heart rate.3 gastric nerve region before surgery and an injection of levobupi- − The SPI has been used to study opioid administration during vacaine 2.5 mg ml 1 (Abbvie, Helsinki, Finland) after surgery, – surgery,4 7 but evidence of its efficacy in paediatric populations is before extubation. The block group (BG) received the injections scarce. Only two studies have been conducted in paediatric popu- in the opposite order. The order of the injections was blinded to lations, and the children were older than 3 yr in both.89 the patient and the study personnel. Development of the nervous system continues after birth,10 and heart rate is an age-dependent parameter in children.11 As Induction of general anaesthesia the heart rate is higher when compared with adults, and the sci- − − entific background on the behaviour of the plethysmographic Fentanyl 2 μgkg 1 i.v., glycopyrrolate 5 μgkg 1 i.v., and thiopental − waveform as a function of age is very limited, evidence from 4–5mgkg 1 i.v. were used for induction of general anaesthesia adult populations cannot be directly adapted to children. and to enhance tracheal intubation. When clinically necessary, − In this prospective, randomized, double-blinded study, we succinylcholine 1–1.5 mg kg 1 i.v. was also given. wanted to test the performance of the SPI to detect nociception in small children at the time of intubation, surgical incision, Delivery of local anaesthesia and at signs of inadequate anti-nociception. Patients received an ilioinguinal and iliohypogastric nerve block Methods under ultrasound guidance (Sonosite S-nerve; Sonosite Inc., Bothell, WA, USA) during anaesthesia, before and after surgery. Downloaded from The study was registered at ClinicalTrials.gov (NCT02045810). One author (J.H.) performed all of the preoperative blocks and After receiving approval from the local ethics committee (ETL most of the postoperative blocks. A visual assessment with ultra- R13137), 30 patients were randomized into the study groups sound was used to ensure the necessary amount of local anaes- (Fig. 1). Written informed consent was obtained from both parents thesia to produce an effective block. However, the maximal − before enrolment. The study population consisted of children aged amount of levobupivacaine (1.5 mg kg 1 for children <6 months − <2 yr, with ASA classification I–III. They were undergoing surgery or 3 mg kg 1 for children >6 months) was not exceeded. http://bja.oxfordjournals.org/ for inguinal hernia repair or correction of undescended testicles. Both uni- and bilateral procedures were included. Patients with cardiac problems or known ECG disturbances were excluded. Maintenance of anaesthesia Anaesthesia was continued with sevoflurane (2–5% end tidal) −1 Randomization and additional fentanyl boluses of 0.5–1.0 μgkg i.v., if clinically necessary. As the concentration of sevoflurane was considered Randomization information was kept in opaque, sealed, num- sufficient, any additional medication was at the discretion of bered envelopes. Before the patient entered the operating room, the practising anaesthetist. In most instances, a bolus of fentanyl an independent nurse, who was otherwise not involved in the was given in the event of movement or a rapid 15% increase in by guest on August 19, 2016 study, opened the envelope and labelled two syringes based on heart rate (HR) or non-invasive blood pressure (NIBP) during a the randomization. The syringes were marked as number one 5 min period if clinically necessary. A fentanyl bolus was not or two, indicating the order of usage. The saline group (SG) was judged as mandatory treatment in all episodes of movement or changes in haemodynamics. The delivery of sevoflurane was adjusted based on clinical decision and primarily not as a reaction to the signs of inadequate anti-nociception. Tracheal extubation was performed in children aged <6 months when fully awake, Assessed for and in older children after the return of spontaneous breathing. eligibility (n=40) Monitoring ended in all patients upon leaving the operating theatre.

Declined to participate Monitoring (n=10) The ECG and the photoplethysmography were continuously monitored. For ECG monitoring, a standard three-lead ECG was used. Measurement of the SPI was performed from a ﬁnger on Randomized the side opposite to the NIBP measurement, with a TS-PAW (n=30) adhesive saturation sensor™ (GE Healthcare, Helsinki, Finland). The NIBP measurement was set to 5 min intervals. The SPI and the entropy values were kept hidden from the personnel during surgery. An anaesthesia nurse or attending anaesthetist simultaneously recorded the time points of interest and the concentrations of sevoﬂurane throughout the study. The numerical values of HR, NIBP, entropy, and SPI were monitored with a Carescape Block Group Saline Group B850 or B650 monitor (GE Healthcare) and recorded using the (n=15) (n=15) S5 Collect software (GE Healthcare) at 10 s intervals. The plethysmographic waveform was continouously recorded, and the manufacturer reanalysed the curve to produce data for the com-

Fig 1 Flowchart. ponents of the SPI, normalized pulse plethysmographic amplitude (PPGAnorm), and normalized RR interval (RRInorm).3 360 | Harju et al.

Power analysis Results The power analysis was based on the results of a previous study Data were collected between January 2014 and March 2015. comparing the surgical stress index in a group of adult female pa- The patient characteristics and relevant intraoperative data are 12 tients, with or without epidural analgesia at skin incision. The shown in Table 1. A significant increase in the SPI was observed primary outcome measure of our study was the change in SPI at the time of intubation and at the time of incision in the SG. At value at the time of the surgical incision. Assuming similar re- the time of incision in the BG, and when the signs of inadequate sponses [20 units difference (Δ) in SPI and standard deviation anti-nociception were present, no significant changes in the SPI () of 14.4], nine patients per group would be needed to reach a were observed (Fig. 2 and Table 2). The PPGAnorm changed sig- power of 0.80 (P<0.05). However, given that the physiology of nificantly at the times of intubation and incision in the BG, and small children differs from that of adults, and because previously the RRInorm at the time of incision in the SG (Fig. 3 and Table 2). there was no knowledge of the magnitude of  in small children, Eighteen patients experienced 26 events of inadequate anti- the study size was increased to 15 patients in each group. nociception, fulfilling the predetermined criteria. As a reaction to the operation, fentanyl was administered 12 times; six times in both groups (in nine instances soon after Statistical analysis incision). The main reason for administration was movement The results were analysed using IBM SPSS statistics version 23 (IBM, (five times) or heart rate and blood pressure reaction (seven Armonk, NY, USA). Two-tailed P-values <0.05 were considered times). A significant change in the SPI from a median (IQR) of significant. The data were found to be scattered and are reported 56 (45, 67) to 78 (67, 84; P=0.008) was observed (Fig. 2). Downloaded from as the median and quartiles Q1 and Q3 (IQR). Between-group differ- At the time of tracheal intubation, the sevoflurane end-tidal ences were analysed using Wilcoxon signed rank tests. concentration was between 1.8 and 5.3%. No correlation between Our primary outcome measure was the difference in the SPI at the ΔSPI and age was found (data not presented). the time of incision. As a secondary outcome measure, the SPI va- The changes in heart rate, response entropy (RE), NIBPsys, and lues were also compared at the time of intubation or when the NIBPmean after the incision are shown in Table 3. There were no fi signs of inadequate anti-nociception [movement, somatic arou- signi cant differences between the groups. http://bja.oxfordjournals.org/ sal (coughing, grimacing etc.), 15% elevation in NIBP or HR] The mean amount of local anaesthetic used was 0.2 ( 0.1) ml − were observed. Our special interest was in instances when the kg 1 for each block. patient received fentanyl in association with a reaction. Add- itionally, the components of the SPI, PPGAnorm, and RRInorm Discussion were analysed at the time points of intubation and incision. To prevent inaccuracy in the choice of analysed time windows, the The main finding in our study was that the SPI was able to detect median value within a selected time window of 30 s before and nociceptive stimuli in children aged younger than 24 months. A 30 s after the point of interest was used for the statistical calcula- significant increase in the SPI was seen in both groups at the

tions of SPI, RRInorm, and PPGAnorm values. The signs of inad- time of intubation and at the time of incision in the SG. In the by guest on August 19, 2016 equate anti-nociception were analysed only between the time of BG, the SPI did not change significantly at the time of incision, in- intubation and the end of surgery. In the event of several signs of dicating a blunted reaction with ilioinguinal and iliohypogastric inadequate anti-nociception occurring within 10 min intervals, block. Moreover, a significant change in the SPI was also seen only the first was included in the analysis. when the decision to administer fentanyl was made.

Table 1 Patient characteristics and details of anaesthesia and type of surgery presented as numbers or means ( or range). The P-values were calculated using Student’s unpaired t test

Characteristic Block group Saline group P-value

Age (weeks) 21.6 (6‒78) 23.7 (7‒80) Gestational age (weeks) 57.1 (38.7‒117) 61.3 (40‒120) Sex (n) Female 2 2 Male 13 13 Height (cm) 58.0 (11.9) 63.0 (2.2) Weight (kg) 5.8 (3.1) 6.8 (2.2) ASA I/II/III (n) 2/12/1 5/10/0 Surgery (n) Inguinal hernia 14 13 0.559 Undescended testicle 1 2 Type (n) Unilateral 13 10 0.208 Bilateral 2 5 Intubation attemps (n) 1.6 (1.0) 1.5 (0.9) 0.849 Time from block to incision (min) 23:33 (7:40) 18:57 (4:41) 0.057 − Fentanyl (μgkg 1) 3.0 (0.8) 2.4 (0.64) 0.386 Sevoﬂurane at incision (end-tidal %) 2.9 (0.4) 2.9 (0.4) 0.930 Surgical pleth index in children younger than 24 months | 361

AB100 100

80 80

60 60

40 40 Downloaded from 20 20 –120 0 120 –120 0 120 SPI

C D 100 100 http://bja.oxfordjournals.org/

80 80

60 60 by guest on August 19, 2016

40 40

20 20 –120 0 120 –300 0 300

Time (s)

Fig 2 The median value of the surgical pleth index (SPI) at incision in the block group (), the saline group (), at intubation (), and at times when fentanyl was administered (). Blue lines represent each individual, and the green line shows the median of all measurements. The time point of interest is located at the middle of the x-axis (0). The total time line is 240 s (–) and 600 s ().

The heart rate and NIBP changed at all measured points when There also seemed to be a large variance in the baseline of the defined by maximal change within a reasonable time frame. Al- PPGAnorm, and the duration of the reactions was short. Interest- though traditionally used as measurement of nociception, these ingly, the PPGAnorm decreased in all groups at the times of both have not been very specific markers for nociception.13 14 This also intubation and incision, but the change was not significant in the applies to RE, which is known to reflect muscle activity at the SG at the time of incision. This might be because of the wide IQR forehead. The RE reaction is lost with the use of neuromuscular found with PPGAnorm. Although not statistically significant, the block,15 and entropy has shown unreliable values in infants.16 rate of change was almost the same as found at intubation. The In contrast to earlier methods, the SPI has also proved to be a RR interval, which is the inverse of the HR, increased at intub- promising method for monitoring intraoperative nociception in ation, indicating decreasing HR, and decreased (i.e. HR increased) older paediatric patients.9 However, the paediatric data are in- at incision. This is consistent with an earlier study in adults, consistent and controversial.8 The SPI has not been studied in where the RR interval reacted to tracheal intubation and a stan- children younger than 24 months of age. To our knowledge, dardized tetanic stimulus.17 The strong vagal stimulus caused this is the first study conducted in children aged <2 yr. by intubation might explain why the heart rate decreased as a re- Although significant changes in the SPI were seen, a large interaction to intubation, but increased in association with surgical individual difference was observed in our study (Figs 2 and 3). stimulation, which is known to provoke a sympathetic reaction. 362 | Harju et al.

Table 2 Surgical pleth index, normalized pulse plethysmographic amplitude, and normalized RR interval values at different time points. Each A and B represents a Median value of a time window of 30 s length. The time window was taken ending about 20 s before (A) and starting about 20 s after (B) each time point (like SPI at intubation). For signs of inadequate anti-nociception, the median value at the time point (B) and 5 min before (A) is presented. Δ describes the median magnitude of change at each value. The P-values were calculated with the Wilcoxon signed rank test. Values are presented as median (interquartile range). *P < 0.05. PPGAnorm, normalized pulse plethysmographic amplitude; RRInorm, normalized RR interval; SPI, surgical pleth index

Parameter n ABΔ P-value

SPI Intubation 30 52.5 (40.5‒69.5) 58.5 (50.5‒58.5) 5.8 (−1.5 to 20.3) 0.019* Incision in saline group 15 49.00 (37.6‒67.2) 62.8 (53.0‒69.9) 10.7 (−6.1 to 24.1) 0.048* Incision in block group 15 58.0 (44.9‒60.6) 57.3 (51.8‒68.8) 2.8 (−4.1 to 13.9) 0.177 Signs of inadequate anti-nociception 28 56.0 (47.6‒68.9) 60.0 (52.8‒76.0) 1.3 (−3.9 to 12.5) 0.148 PPGAnorm Intubation 30 522.0 (273.0‒713.2) 426.6 (157.1‒595.9) 1.3 (−3.8 to 12.5) 0.003* Incision in saline group 15 521.0 (252.5‒635.5) 349.4 (214.8‒520.6) −68.3 (−245.1 to 127.7) 0.397 Incision in block group 15 396.0 (274.0‒493.0) 323.9 (178.4‒440.7) −98.1 (−175.5 to −27.1) 0.009* RRInorm ‒ ‒ −

Intubation 30 263.3 (97.0 518.0) 293.3 (129.0 405.5) 2.3 ( 189.3 to 89.0) 0.387 Downloaded from Incision in saline group 15 587.3 (480.3‒690.5) 481.5 (186.4‒543.3) −120.9 (−355.1 to −10.1) 0.007* Incision in block group 15 667.0 (564.8‒690.0) 631.3 (243.4‒715.3) −38.0 (−176.1 to 16.4) 0.087

reacted strongly, but others had almost no reaction. The baseline http://bja.oxfordjournals.org/ Table 3 The maximal change of systolic (NIBPsys), mean values of SPI and PPGAnorm were also variable. The reasons for (NIBPmean) pressures and Response Entropy (RE) within a this variability remain unknown. selected timeline (max2min and max5min). Values are The median SPI value seems to be much higher in our study presented as median (interquartile range). *P < 0.05. BG, block (52.5) compared with the mean SPI value before intubation in group; HR, heart rate; SG, saline group adults (44.2)18 and older children (43.3),9 perhaps because of a Parameter At incision After incision P-value higher HR in the younger age group. At the time of incision, the HR increased in both of our study groups. The SPI consists of max2min HR HR (33%) and change in the pulse plethysmographic amplitude

BG 135.0 (130.0‒150.0) 145.0 (130.0‒160.0) 0.008* by guest on August 19, 2016 (67%). It combines the information as one number from 0 to SG 138.0 (130‒160.0) 155 (145.0‒160.0) 0.001* 100, indicating surgical nociception.3 In other words, the higher REmax2min baseline HR in small children compared with that of adults re- BG 26.0 (17.50‒38.8) 30.0 (22.0‒62.0) 0.008* sults in a higher baseline value of SPI in children. For example, SG 26.0 (17.50‒50.0) 30.0 (22.0‒69.0) 0.001* a 20 unit change in HR in paediatric patients has a smaller rela- NIBPsysmax5min BG 73.70 (49.0‒80.0) 76.0 (54.3‒90.0) 0.003* tive effect on the SPI as compared with adults. All in all, the SG 67.0 (48.0‒76.0) 72.0 (54.0‒82.0) 0.001* role of plethysmography is major in the calculation of the SPI. 8 NIBPmeanmax5min A recent study attempted to show the usefulness of SPI in BG 52.0 (37.5‒58.3) 54.0 (40.5‒69.3) 0.002* guiding intraoperative fentanyl administration in children aged SG 47.0 (37.0‒58.0) 56.0 (41.0‒61.0) 0.001* 3–10 yr undergoing adenotonsillectomy. The study showed that the use of SPI was associated with adequate intraoperative anaesthesia and anti-nociception. Unfortunately, patients had more pain, nausea, and emergence agitation during the post- These findings suggest that the rapid change seen in the SPI at in- operative period, which was interpreted as a failure of the SPI. tubation was most probably caused by a change in PPGAnorm. This underlines the fact that the effect of perioperative treatment The SPI was developed by combining two parameters, RRI- during the postoperative period should not be underestimated. In norm and PPGAnorm, which were found to describe the surgical addition, this unsatisfactory result may have been a consequence nociception best.3 To our knowledge, these components have not of trying to transpose reference adult values directly to children, been studied in detail thereafter, in relation to surgical nocicep- as discussed above. tion. Nobody really knows how these components behave in The SPI has mostly been studied with propofol–remifentanil small children. In the present study, the change in PPGAnorm anaesthesia.4719The SPI guidance in propofol–remifentanil an- was significant in the block group, but not in the saline group, aesthesia compared with standard monitoring resulted in more which is paradoxical. The SPI response was nevertheless signifi- stable anaesthesia and lower consumption of remifentanil.4 It cant in the saline group and not in the block group. This might be has also been shown to be similarly reactive during sevoflur- because of a higher basal heart rate in small children when com- ane–fentanyl anaesthesia20 and sevoflurane–sufentanil-based pared with adults. anaesthesia.6 However, using such anaesthesia, SPI guidance Our patients were very young, especially when defined by ges- did not show any benefits when compared with standard of tational age. The infants were still immature, which might ex- care. Although the reactivity of the SPI to noxious stimulation plain the difference in reactions when compared with adults.10 has been found to be present with all opioids in the adult popu- The individual differences in the values of PPGAnorm at the lation, it remains to be determined whether such reactivity is pre- time of skin incision were large. Several patients in both groups sent in small children when other opioids are used. Surgical pleth index in children younger than 24 months | 363

The ilioinguinal and iliohypogastric block has been shown to be an effective aid for inguinal hernia repair.21 With ultrasound A 1000 guidance, it has shown a 95% success rate in inguinal hernia repair, orchidopexy, or hydrocoele repair for postoperative pain.22 Placing the block before the surgery is common practice at our 800 hospital, and surgeons have not argued about interference with the surgical site when it is placed before surgery. In our study, no difference in the amount of fentanyl used in micrograms per kilogram or the number of administrations during the intrao- 600 perative period were found between the two study groups. These results are also in line with the ﬁndings of a study that found ilioinguinal and iliohypogastric nerve block unable to block all re- 400 actions during inguinal hernia repair.23 A number of confounding factors can be found in our study. Firstly, all patients were intubated, with relatively high amounts 200 of opioid administered. This might have blunted the effect of relatively minor noxious stimulus, namely the surgical incision. Secondly, the patients were also administered a relatively high

sevoﬂurane concentration. Thirdly, the innervation of the testi- Downloaded from –120 0 120 cles and abdominal wall comes from pelvic plexus and thoracic 23 24 B 1000 nerve routes, which are not necessarily affected by the block. This might explain why the fentanyl consumption did not differ between the groups, even though the ilioinguinal and iliohypogastric block should be effective during the incision for both types of 800 surgeries. In relation to this, the groups are only comparable up to http://bja.oxfordjournals.org/ the time of incision and the time interval brieﬂy afterward. Fourth- ly, all patients received glycopyrrolate in order to block possible 600 side-effects caused by succinylcholine. Glycopyrrolate has anticholinergic effects, which might block part of the reactions used

PPGAnorm in the calculation of the SPI.25 However, all the anaesthetics and 400 anaesthesia methods used in this study are typical in paediatric anaesthesia practice; thus, our ﬁndings describe the typical performance of the SPI in this patient group. 200 by guest on August 19, 2016

Conclusions Our study shows that the SPI has the potential to detect nocicep- –120 0 120 tion in small children, and an (blinded) increase in the SPI was as- C sociated with the clinical decision to administer fentanyl. 1000 However, the duration of the reaction was very short, and there were large inter-individual differences among children, which might have interfered with the clinical usefulness of the SPI. 800 This indicates that one should be cautious when considering using the SPI in clinical practice in this age group.

600 Authors’ contributions Study design, writing, and analysis: J.H., M.-L.K., H.L., A.Y.-H. 400 Conduct of the study measurements and blocks: J.H., M.K. Analysis of the results: J.H., M.-L.K., H.L., A.Y.-H. Writing the manuscript: J.H., M.-L.K., H.L., A.Y.-H., M.K.

200 Acknowledgements Pirkanmaa Hospital District ethics committee approved the study. The authors wish to thank GE Healthcare for providing –120 0 120 the data collection software and surgical pleth index study moni- Time (s) tors. The authors also wish to thank Mr Matti Huiku from GE Healthcare Finland for providing the data calculations for compo- Fig 3 The median value of the normalized pulse plethysmographic nents of the SPI. amplitude (PPGAnorm) at incision in the block group (), the saline group (), and at intubation (). Blue lines represent each individual, and the green line shows the median of all measurements. The time point of Declaration of interest interest is located at the middle of the x-axis (0). The total time line is 240 s. None declared. 364 | Harju et al.

Funding 13. Sabourdin N, Arnaout M, Louvet N, Guye ML, Piana F, Constant I. Pain monitoring in anesthetized children: ﬁrst as- Finnish Cultural Foundation, Pirkanmaa Regional Fund. sessment of skin conductance and analgesia-nociception index at different infusion rates of remifentanil. Paediatr Anaesth 2013; 23: 149–55 References 14. Rantanen M, Yli-Hankala A, van Gils M, et al. Novel multi- 1. Glass SA, Shafer SL, Reves JG, eds. Opioids. In: Miller RD, parameter approach for measurement of nociception at Eriksson LI, Fleischer LA, Wiener-Kronish JP, Young WL, skin incision during general anaesthesia. Br J Anaesth 2006; eds. Miller’s Anesthesia, 7th Edn. Philadelphia, PA, USA: 96: 367–76 Churchill Livingstone Elsevier, 2010; 769–823 15. Aho AJ, Lyytikainen LP, Yli-Hankala A, Kamata K, Jantti V. Ex- 2. Wheeler M, Oderda GM, Ashburn MA, Lipman AG. Adverse plaining entropy responses after a noxious stimulus, with or events associated with postoperative opioid analgesia: a sys- without neuromuscular blocking agents, by means of the raw tematic review. J Pain 2002; 3: 159–80 electroencephalographic and electromyographic characteris- 3. Huiku M, Uutela K, van Gils M, et al. Assessment of surgical tics. Br J Anaesth 2011; 106:69–76 stress during general anaesthesia. Br J Anaesth 2007; 98: 16. Klockars JG, Hiller A, Munte S, van Gils MJ, Taivainen T. Spec- 447–55 tral entropy as a measure of hypnosis and hypnotic drug ef- 4. Chen X, Thee C, Gruenewald M, et al. Comparison of fect of total intravenous anesthesia in children during slow surgical stress index-guided analgesia with standard clinical induction and maintenance. Anesthesiology 2012; 116: 340–51

practice during routine general anesthesia: a pilot study. 17. Luginbühl M, Yppärilä-Wolters H, Rüfenacht M, Petersen- Downloaded from Anesthesiology 2010; 112: 1175–83 Felix S, Korhonen I. Heart rate variability does not discriminate 5. Gruenewald M, Meybohm P, Ilies C, et al.Influence of different between different levels of haemodynamic responsiveness remifentanil concentrations on the performance of the during surgical anaesthesia. Br J Anaesth 2007; 98: 728–36 surgical stress index to detect a standardized painful stimu- 18. Ilies C, Gruenewald M, Ludwigs J, et al. Evaluation of the sur- lus during sevoflurane anaesthesia. Br J Anaesth 2009; 103: gical stress index during spinal and general anaesthesia. Br J 586–93 Anaesth 2010; 105: 533–7 http://bja.oxfordjournals.org/ 6. Gruenewald M, Willms S, Broch O, Kott M, Steinfath M, Bein B. 19. Bonhomme V, Uutela K, Hans G, et al. Comparison of the Sur- Sufentanil administration guided by surgical pleth index vs gical Pleth Index™ with haemodynamic variables to assess standard practice during sevoflurane anaesthesia: a rando- nociception–anti-nociception balance during general anaes- mized controlled pilot study. Br J Anaesth 2014; 112: 898–905 thesia. Br J Anaesth 2011; 106: 101–11 7. Struys MMRF, Vanpeteghem C, Huiku M, Uutela K, 20. Mustola S, Parkkari T, Uutela K, Huiku M, Kymäläinen M, Blyaert NBK, Mortier EP. Changes in a surgical stress index Toivonen J. Performance of surgical stress index during in response to standardized pain stimuli during propofol–re- sevoflurane-fentanyl and isofl urane-fentanyl anesthesia. mifentanil infusion. Br J Anaesth 2007; 99: 359–67 Anesthesiol Res Pract 2010; 2010: 810721 8. Park JH, Lim BG, Kim H, Lee IO, Kong MH, Kim NS. Comparison 21. Markham SJ, Tomlinson J, Hain WR. Ilioinguinal nerve block by guest on August 19, 2016 of surgical pleth index-guided analgesia with conventional in children. A comparison with caudal block for intra and analgesia practices in children: a randomized controlled postoperative analgesia. Anaesthesia 1986; 41: 1098–103 trial. Anesthesiology 2015; 122: 1280–7 22. Willschke H, Marhofer P, Bosenberg A, et al. Ultrasonography 9. Kallio H, Lindberg LI, Majander AS, Uutela KH, Niskanen ML, for ilioinguinal/iliohypogastric nerve blocks in children. Br J Paloheimo MP. Measurement of surgical stress in anaesthe- Anaesth 2005; 95: 226–30 tized children. Br J Anaesth 2008; 101: 383–9 23. Netter F. In: Hansen J, Benninger B, Brueckner J, Carmichael S, 10. Davis C, Motoyama, eds. 4. Cardiovascular physiology. In: Granger N, Tubbs S, eds. AtlasofHumanAnatomy,5thEdn. Smith’sAnesthesiaforInfantsandChildren. Philadephia, PA, Philadelphia, PA, USA: Saunders Elsevier, 2011; 159, 260, 389 USA: Elsevier Inc., 2011; 100–7 24. Nan Y, Zhou J, Ma Q, Li T, Lian QQ, Li J. Application of ultra- 11. Daymont C, Bonafide CP, Brady PW. Heart rates in hospita- sound guidance for ilioinguinal or iliohypogastric nerve lized children by age and body temperature. Pediatrics 2015; block in pediatric inguinal surgery. Zhonghua Yi Xue Za Zhi 135: 1173–81 2012; 92: 873–7 12. Yli‐HankalaA,RantanenM,UutelaK,KärkäsP, 25. Hocker J, Broch O, Grasner JT, et al. Surgical stress index in re- Kymäläinen M, Huiku M. Surgical stress index and epidural sponse to pacemaker stimulation or atropine. Br J Anaesth analgesia: A‐92. Eur J Anaesthesiol 2006; 23: A92 2010; 105: 150–4

Handling editor: A. R. Absalom J Clin Monit Comput DOI 10.1007/s10877-017-9984-3

ORIGINAL RESEARCH

Comparison of non-invasive blood pressure monitoring using modified arterial applanation tonometry with intra-arterial measurement

Jarkko Harju1 · Antti Vehkaoja2 · Pekka Kumpulainen2 · Stefano Campadello3 · Ville Lindroos4 · Arvi Yli‑Hankala1,4 · Niku Oksala4,5

Abstract Intermittent non-invasive blood pressure meas- the difference was 4.8 ± 7.7 (LoA − 14.1 to 23.6) mmHg urement with tourniquets is slow, can cause nerve and skin (r = 0.72), and for mean arterial pressure it was 11.18 ± 11.1 damage, and interferes with other measurements. Invasive (LoA − 12.1 to 34.2) mmHg (r = 0.642). Our study revealed measurement cannot be safely used in all conditions. Modi- inaccurate agreement (AAMI) between the two methods fied arterial tonometry may be an alternative for fast and when measuring systolic and mean blood pressures dur- continuous measurement. Our aim was to compare arte- ing post-operative care. The readings for diastolic pressures rial tonometry sensor (BPro®) with invasive blood pres- were inside the limits recommended by AAMI. Movement sure measurement to clarify whether it could be utilized in increased the failure rate significantly (p < 0.001). Thus, the postoperative setting. 28 patients who underwent elec- arterial tonometry is not an appropriate replacement for tive surgery requiring arterial cannulation were analyzed. invasive blood pressure measurement in these patients. Patients were monitored post-operatively for 2 h with standard invasive monitoring and with a study device com- Keywords Blood pressure monitors · Arterial pressure · prising an arterial tonometry sensor (BPro ®) added with a Monitoring, intraoperative three-dimensional accelerometer to investigate the potential impact of movement. Recordings were collected electronically. The results revealed inaccurate readings in method 1 Introduction comparison between the devices based on recommendations by Association for the Advancement of Medical In peri-operative evaluation of blood pressure, the tradi- Instrumentation (AAMI). On a Bland–Altman plot, the bias tional cuff-based non-invasive method has several disad- and precision between these two methods was 19.8 ± 16.7 vantages. It is slow [1], cumbersome, and during prolonged (Limits of agreement − 20.1 to 59.6) mmHg, Spearman use may produce nerve and skin damage [2, 3]. correlation coefficient r = 0.61. For diastolic pressure, Invasive blood pressure monitoring has been used to detect rapid changes in blood pressure. It does not occlude the artery and provides beat-to-beat information on blood Clinical trial registration: NCT02357511 pressure. However, it requires an arterial cannula, which * Jarkko Harju can cause serious complications. In addition, the open [email protected] bloodline is considered a risk in regular wards and needs

1 to be monitored vigilantly [4, 5], limiting its feasibility in Department of Anaesthesia, Tampere University Hospital, many clinical scenarios. PL2000, 33521 Tampere, Finland ® 2 Modified arterial applanation tonometry used inBPro Tampere University of Technology, Tampere, Finland sensor is a novel technology that has been explored as 3 CamsoS Consulting Ltd., Helsinki, Finland an alternative to traditional blood pressure measurement 4 Medical School, University of Tampere, Tampere, Finland [6–8]. It is based on a sensitive pressure sensor placed on a 5 Department of Surgery, Tampere University Hospital, peripheral artery; it does not occlude the artery and causes Tampere, Finland only minimal discomfort [8]. In the commercial version

Vol.:(0123456789)1 3 J Clin Monit Comput radial artery at distal forearm is used. Evidence regarding opposite sides were used for the measurements. To exclude the reliability of the measurements in clinical conditions is marked differences between extremities, a control non- limited, and to our knowledge, no studies of BPro ® tonom- invasive blood pressure was measured with an inflatable etry sensor have been conducted in post-operative settings. cuff once on both sides at the beginning and end of each The aim of the present study was to explore the feasibil- measurement. ity of using applanation tonometry to measure post-oper- The study device consisted of a tonometry sensor, a ple- ative blood pressure in non-selected patients. To investi- thysmography sensor and an attached three-dimensional gate the potential impact of movement on measurements, accelerometer, mounted on a flexible strap and wrapped the tonometry sensor was supplemented with an additional around the wrist on the side opposite to the arterial can- three-dimensional accelerometer. nula. Since the technology for tonometry sensor is different from plethysmography, the results from the plethysmography sensor are reported on a previous publication [9]. The 2 Methods tonometry sensor used was the same as in the commercially available FDA and CE approved BPro ® watch (Healthstats, The study was registered at ClinicalTrials.gov Singapore), and the blood pressure analysis service used (NCT02357511) before recruitment began. Written with our study device was the same as in the commercial informed consent was obtained from all participants before device [8]. The tonometry sensor was placed over the radial the study procedures were initiated. After receiving per- artery by manual palpation of the radial artery by study mission from the local ethics committee (ETL R13145), personnel. Data were obtained via a Wi-Fi connection to an 30 patients were recruited into the study between January Internet remote server (on October 9th 2015). In our study, and May 2015. The Inclusion criteria: patients subjected we had two main differences compared to the commercial to elective surgery requiring arterial cannulation and treat- device. First, the main BPro® watch unit used did not have ment at the post anesthesia care unit of Tampere University a display for data verification. Second, we attached a three- Hospital and willing to participate. This qualified patients dimensional accelerometer (Freescale MMA8452Q; NXP, with high risk or patients with high risk surgery. Exclusion Eindhoven, Netherlands) to monitor the movements of the criteria: patients with cardiac pacemaker. This was rec- arm. By omitting the display from the BPro ® watch, our ommended by the device security unit of the hospital for platform enabled us to combine the Freescale accelerom- minimizing the risk of theoretical adverse events caused by eter for measurement. radio interference. The study was conducted as a 2 h obser- For each blood pressure measurement, 10 s of raw signal vational study during the postoperative care treatment. were collected at 1 min intervals. The device sent the measurement signal to the server, where it was analyzed. Thus, 2.1 Patient monitoring the blood pressure data were intended to be available via a webpage. After the recording, all of the data were down- Patients were monitored with a Carescape B650 moni- loaded from the server. tor (GE Healthcare Finland Oy, Helsinki, Finland) as part The values obtained from the two sources were synced of their routine care. In addition, the measurement device using the local time obtained from the Internet. An inde- under investigation was applied during treatment at a post pendent observer recorded all device and patient move- anesthesia care unit for a continuous measurement of 2 h. ments and verified the data accuracy on the arterial line. Study personnel performed the study measurements dur- Due to technical challenges, the raw waveform signals ing the time window of the routine post-operative period. recorded with the applanation tonometry sensor were con- Due to study staff unavailability, we were unable to record verted to blood pressure readings during post-hoc analysis. data from all recruited patients. All parameters from the The first good-quality pressure waveform signal for each standard monitor were recorded with S5 Collect software patient was visually determined, and the corresponding (GE Healthcare) at 5 s intervals. Additionally, finger pho- arterial measurement was set as a reference pressure. The toplethysmogram, three-lead ECG, and arterial pressure data were thereafter processed at the manufacturer’s server signals were stored at a 300-Hz sample rate. The arterial to provide blood pressure values. cannula was inserted into the radial artery as part of the Accelerometer data were used to test whether ruling out monitoring for the operation. In all cases, the arterial line blood pressure readings obtained during movement would (DTX Plus; Argon Medical devices, Helsinki, Finland) was improve the measurement accuracy. The movement signal still functional at the post anesthesia care unit and provided was calculated from the three-accelerometer channels by the control blood pressure for the study measurements. taking a moving variance with a 40 s window from each of While both devices were placed in distal forearm region, the channels and adding them together. A threshold value we were unable to place sensors at same side. Therefore, for the movement was formed by defining a mean value and

1 3 J Clin Monit Comput standard deviation for the lowest 90% of the summoned calculated with Mann–Whitney or Wilcoxon sign rank test data. The mean plus three times the standard deviation was as intended for scattered data. thereafter calculated and used as a threshold for “slight The trending ability between the devices was tested with movement.” Correspondingly, ten times the standard devia- a four-quadrant pot. A difference within 5 min was used tion added to the mean was calculated and used for the to visualize the direction of changes and concordance was “definite movement” threshold value. Our definition of the measured to describe the frequency of data point with equal limits is a modification of a technique commonly used in direction. The data point with a change of ≤3 mmHg were analytical methods for limits of detection [10]. left out of the analysis [15].

2.2 Power analysis 3 Results According to previous pilot data supplied by the manufac- There were a total of 2449 blood pressure measurements turer, the method has a bias of 3 (SD ± 3) mmHg compared recorded from 30 patients. The pressure signal from the to traditional tourniquet-based measurement. For compar- tonometry device was not measurable in two patients; ing two independent samples a minimum of 17 subjects per these measurements were consequently discarded, leaving group was thus required to provide 80% power with a two- 28 subjects for the final analysis (Fig. 1). Patient charac- sided significance of α = 0.05. Furthermore, on the basis of teristics are described in Table 1. All patients were spon- AAMI recommendations, a minimum of 15 subjects and 10 taneously breathing during the study period. The Table 2 readings per subject should be reported over the proof of describes the number of successful measurements for each accuracy for NIBP devices [11]. Accordingly, the mean and patients and relevant background factors associated. The precision should be within 5 ± 8 mmHg. While we expected types of surgery performed prior to measurements con- higher mean bias, 30 subjects were considered sufficient for sisted of: vascular surgery in 11 patients (39.3%), gastroen- adequate study size. terological surgery in 10 patients (35.7%), urologic surgery in 2 patients (7.1%), plastic surgery in 3 patients (10.7%), 2.3 Statistical analysis and orthopedic surgery in 2 patients (7.1%).

The data were analysed with IBM SPSS statistics version 23 (IBM, Chicago, IL, USA) and for Bland–Altman plot calculated using Microsoft Excel 2010 (Microsoft, Red- mond, WA, USA) as described by Zou et al. [12]. MAT- LAB version 2014b (The MathWorks Inc, Natick, MA, USA) was used for trending analysis. The collection software averages the invasive blood pressure measurements over 5 s time interval and therefore the zero-zone approach recommended by AAMI was not used [11]. Phases of obvious artefacts such as invasive pressure line flushing were manually removed from the data. The results are reported by patient-weighted means of the Bland–Altman method for repeated measures [13] and as Spearman correlation coefficient. We visualized the bias of the novel method compared to arterial invasive blood pressure measurement. The data were found to be nonparametric and are reported as median and interquartile range (IQR = 25–75th percentile) or frequencies except for the Bland–Altman plots as mean (±precision) and 95% limits of agreement added with corresponding 95% confidence intervals. The percentage error (PE) was calculated with a formula {2 SD of bias/ [(invasive pressuremean+ tonometry pressure mean)/2] × 100} for all values [14]. The root-mean-square error (RMSE) was calculated as square-root for the mean of second expo- nent for each individuals mean bias. A statistical difference of p < 0.05 was considered significant, and values were Fig. 1 Study flowchart

1 3 J Clin Monit Comput

Table 1 Patient characteristics, described as mean (standard devia- The Fig. 2 represents the Bland–Altman plots for sys- tion or range) or count (%) tolic (SAP), diastolic (DAP) and mean (MAP) arterial Characteristics pressures as described for multiple measurements and corresponding Spearman correlation coefficients for all meas- Age (years) 67.2 (39–87) urements. The median value (interquartile range) for all Male/female gender 12/16 (42.9/57.1%) measurements is described in Table 3. The trending analy- Height (cm) 172.2 (10.6) sis is shown in Fig. 3. Weight (kg) 80.4 (15.6) 2 The Bland–Altman plot for SAP revealed a mean bias BMI (kg/m ) 27.3 (4.2) and precision of 19.8 mmHg ± 16.7 and limits of agree- ASA I/II/III/IV (n) 1/7/19/1 ment, LoAupper 59.6 mmHg (CI 44.4–67.8), LoAlower −20.1 Peripheral arterial disease 11 (39.3%) (CI − 4.9 to 28.3). Additionally Spearman correlation coef- Atrial fibrillation 3 (10.7%) ficient was r = 0.61 and percentage error 32%. Coronary artery disease 1 (3.6%) The Bland–Altman plot for DAP revealed a mean bias Wrist diameter (cm) 18.3 (2.0) of 4.8 mmHg ± 7.7, LoAupper 23.6 mmHg (CI 16.9–27.4), LoAlower −14.0 (CI − 17.8 to − 7.3) with a Spearman

Table 2 The proportion of successful measurements for each patient and relevant background factors associated Patient Measure- Succesful Wrist diameter Gender Peripheral Atrial Non-selective i.v. Noradrenalin ments total measurements arterial disease fibrillation b-blocker during infusion during PACU PACU

1 93 81 24 Male Yes 2 60 23 20 Male Yes 3 61 23 19 Male Yes 4 93 90 19 Male 5 93 74 22 Male 6 93 78 19 Male Yes Yes 7 93 82 18 Male Yes Yes 8 93 77 15 Female Yes 9 93 90 18 Male 10 93 34 19 Female 11 93 87 19 Female Yes 12 55 39 19 Male Yes Yes 13 93 65 20 Female 14 89 53 17 Male Yes 15 77 72 20 Male Yes 16 74 54 15 Female Yes Yes Yes 17 93 56 18 Male 18 93 88 20 Male Yes Yes 19 93 83 15 Female Yes Yes 20 93 57 20 Male Yes Yes 21 93 85 17 Female Yes 22 93 78 17 Female 23 0 0 15 Female Yes 24 0 0 17 Male Yes 25 93 92 19 Female 26 93 84 18 Female 27 93 61 17 Female Yes 28 93 86 19 Female 29 93 54 19 Male Yes 30 90 48 17 Male Total 2459 1894

PACU post anesthesia care unit

1 3 J Clin Monit Comput

Fig. 2 Bland–Altman Plot and Spearman’s correlation coefficient for Bland–Altman plot black lines describe the mean and limits of agree- systolic (a), mean (b), and diastolic (c) pressures (Bland–Altman Plot ment. The grey areas indicate corresponding 95% CI n = 28 patients and for correlation n = 1894 measurement pairs). In

correlation coefficient of r = 0.72 and percentage error correlation coefficient was r = 0.72 and percentage error 31%. 31%. Respectively for MAP there was a mean bias of The tonometry sensor could not measure pressure in 11.1 mmHg ± 11.1, LoAupper 34.3 mmHg (CI 21.6–40.2), 532/2449 measurements (21.6%). All movements, indi- LoAlower −12.1 (CI − 18.0 to 0.6). Additionally Spearman cated by the accelerometer, were associated with an

1 3 J Clin Monit Comput

Table 3 Median pressure (mmHg) for systolic (SAP), diastolic Table 4 Characteristics for measurements when study device was (DAP), and mean (MAP) arterial pressure described as median (IQR) unable to measure blood pressures and Spearman’s correlation coefficient between devices (n = 1894) Characteristics n Failed (IQR) p value Pressure BPro® sensor Invasive Spearman p value correlation Measurements total 2449 21.6% Slight movement Yes 336 35.1% <0.001* SAP 118 (104–134) 139 (119–156) 0.61 <0.001 No 2113 19.6% MAP 76 (67–90) 87 (77–99) 0.64 <0.001 Definite movement Yes 73 34.2% 0.008* DAP 56 (47–68) 60 (53–71) 0.72 <0.001 No 2376 21.3% Peripheral arterial Yes 11 27.0% (16.2–40.4) 0.042* disease No 17 8.6% (3.7–26.3) Wrist circumference >19 7 20.4% (4.3–38.7) 0.699 (cm) 17.5–19 11 11.8% (4.3–38.7) <17.5 10 22.0% (9.4–40.0) BMI (Kg/m2) <25.5 10 13.5% (8.6–30.2) 0.350 25.5–29 8 30.1% (8.1–58.4) >29 10 16.0% (3.8–29.8)

Data are described as numbers and percentage or percentage (interquartile range, IQR) as appropriate. p value was calculated between failure rates. n indicates number of measurements or patients * p < 0.05

mmHg] and highest [31.7 (22.5–45.1) mmHg] quartiles of systolic pressures (p < 0.001). The absence of marked differences between extremities was verified with cuff-based blood pressures. There were Fig. 3 Four-quadrant plot visualizing the trending ability in mean no significant differences between the right and left extrem- arterial pressures between invasive blood pressure (ΔBP−IVMAP) ® measurements and BPro sensor (ΔBP−BProMAP). Horizontal line ities: difference for SAP 1.2 mmHg (95% CI − 11.4 to 8.9 describes the change within 5 min interval in invasive pressure while p ® and = 0.799), DAP 0.2 mmHg (95% CI − 2.8 to 2.3 and vertical axis describe simultaneous change in BPro sensor meas- p grey = 0.835), and MAP 1.5 mmHg (95% CI − 0.9 to 3.9 and urements. The exclusion zone of 3 mmHg is visualized with p area. Each asterisk represents a data point. Data points at lower left = 0.211). or upper right corner indicate a change of equal direction. Concord- ance rate indicates the percentage of data points with equal direction. Included data points n = 1128, excluded data points n = 348 4 Discussion increased proportion of failed measurements (p < 0.001) The main finding of our study was that the BPro® sensor (Table 4). The rate of failed measurement was associated was unable to detect systolic and mean pressures accu- with peripheral arterial disease (PAD) (p = 0.042), but not rately. To our knowledge, this is the first study to compare a with circumference of the wrist (p = 0.699) or body mass recently developed wrist arterial modified tonometry sensor index (BMI) (p = 0.350). with invasive arterial line measurement in high risk patients The accuracy of the tonometry sensor was further evalu- in post-operative care conditions. The recommendations by ated by root-mean-square error (RMSE) (Table 5). BMI the AAMI suggest a mean difference of ±5 mmHg or less, (p = 0.024–0.050) and wrist circumference (p = 0.003–0.83) with a standard deviation of ±8 mmHg or less for method were associated with error. The lowest accuracy with comparison studies. Based on this the SAP and MAP pres- RMSE was 36.7 mmHg for SAP, and it was associated sure comparisons were inaccurate in our study, while DAP with the highest quartile of BMI. Movement also had a was just inside the recommended limits [11]. Also when strong association with error (p < 0.001); however, the error defined by percentage error, the error was out of the accept- was smaller with movement. SAPs were further divided able limit of 30%. It is originally described for cardiac out- into three equal groups, and there The exclusion zone of put measurements [16] but also reported for blood pressure 3 mmHg is visualized withwas a significant difference comparisons [17–19]. The trending analysis revealed a very between median RMSESAP for the lowest [17.1 (14.6–21.1) inaccurate trending ability with very low concordance rate

1 3 J Clin Monit Comput

meaning that a change in one device wasn’t associated with a same directional change in the other. The rate of successful measurements (78.4%) can be considered acceptable. Although MAP is often considered 0.003* 0.781 0.004* the most important indicator for tissue perfusion, the lim- p value <0.001* <0.001* its to avoid post-operative bleeding after high-risk surgery such as neurosurgery or vascular surgery are often defined by systolic pressures. Based on our findings, the sensor cannot be utilized to directly replace invasive arterial cannulation in the post-operative care period. 5.5 (5.2–7.7) 8.0 (5.8–10.5) 5.5 (4.9–6.7) 7.9 (5.5–11.3) 7.5 (5.6–9.8) 7.5 (6.0–10.2) 5.5 (5.1–7.9) 5.5 (5.2–7.5) 7.5 (5.5–10.2) 7.5 (5.5–10.2) 10.8 (7.8–14.2) 11.3 (9.3–14.4) DAP Our study revealed a much weaker performance of the BPro® tonometry sensor for SAP compared to previous studies that reported differences of 2.1 mmHg [6] and 1.3 mmHg [7] for SAP. The difference in DAP that we found was comparable to that found by other studies: 0.049* 0.487 0.015* p value <0.001* <0.001* 7.4 mmHg [6] and 1.6 mmHg [7]. Furthermore, the non- invasive measurement was found to be less accurate with increasing pressures. This might explain the difference in accuracy compared to previous studies. In addition, the inter-study differences might be due to differences in 9.7 (8.5–14.4) 9.7 (8.5–11.0) study populations and methods. The previous studies were 19.3 (12.6–26.7) 12.2 (7.2–16.9) 10.9 (9.1–15.0) 15.5 (10.1–17.5) 12.3 (8.1–19.3) 19.3 (16.6–29.1) 12.3 (7.4–15.9) 11.2 (9.2–14.7) 13.0 (9.7–17.4) 12.5 (9.7–17.4) MAP carried out with relatively young and healthy individuals (mean 32 years old) with minor concomitant diseases. This is in contrast with our study, which included a much older population (mean 67 years), high American Soci- ety of Anesthesiologist (ASA) classifications, and a large 0.083 0.817 0.090 0.015* p value <0.001* proportion of PAD as was expected by the elective arterial cannula as an inclusion criteria. Our study population represents a high risk patient population and typical challenges for intraoperative measurement. In addition, instead of non- invasive pressure measurement, the reference in our study 36.7 (19.4–49.8) 18.5 (11.4–28.2) 18.1 (16.7–27.8) 18.1 (16.8–21.7) 25.7 (16.7–31.1) 21.1 (15.2–37.3) 36.7 (21.1–41.2) 16.6 (11.8–31.1) 21.7 (16.5–27.6) 21.7 (15.8–31.7) 18.1 (16.8–18.1) 21.7 (15.8–31.7) SAP was invasive arterial pressure measurement, which is considered the gold standard. The rate of success in measuring pressure was sub- stantially higher in our study than in previous studies [6], 9 9 7 although it was not actually reported by many previous 10 11 17 11 10 47 210 1673 1836 n studies. In that study, ambulatory 24-h measurement was used and the accuracy was significantly lower during sleep, when the device was not monitored. In our study, the rate of successful measurements was significantly lower with the presence of PAD or movement. Even slight movement raised the number of unsuccessful measures markedly >29 25.5-29.0 <25.5 Yes Yes No >19 17.5-19.0 <17.4 No Yes No (p < 0.001). The effect of movement was not evaluated in previous studies, while information on movement is not available for the standard BPro ® device. Our patients were also observed by an independent observer, who was constantly able to correct placement of the device if it became misplaced. Our study setting corresponds well to clinical situ- 2 ations where arterial line is typically used. Though, the Root-mean-square error (RMSE) calculated for systolic (SAP), diastolic (DAP), and mean (MAP) arterial in different (DAP), pressures groups diastolic (SAP), systolic error (RMSE) calculated for Root-mean-square number of failed measurements might be even higher. A technology similar to BPro ® device is seen in the Tensys ® ® Slight movement Yes Slight movement value calculated between or within groups. n indicates number of patients or measurements calculated between RMSE described p value with median (interquartile range). patient’s Each * p < 0.05 5 Table Peripheral arterial diseas Wrist circumference (cm) circumference Wrist BMI (Kg/m Definite movement Yes Definite movement T-Line (Tensys Medical, San Diego, CA, USA) where

1 3 J Clin Monit Comput a moving sensor constantly locates the best signal and It has been stated previously that although non-invasive the distal forearm is fixed into extension to enhance the and invasive pressures are interchangeable when defined measurement. The reported accuracy has been relatively by MAP values, there are marked differences when meas- good, but the number of studies committed and patients uring the extremes of hypo- or hypertension. The trend in included is still rather small. Furthermore, the fixation differences between the two methods was also consistent of the wrist makes the device much more uncomfortable for DAP and MAP [26]. Non-invasive measurements have [17, 19–21]. This is in contrast to BPro ® design, where been argued to be interchangeable with invasive blood pres- armband fixates the sensor at a location where the pulsa- sure measurements mostly for MAP [27, 28]. In relation to tion is best palpable and a relatively free movement of this, if the signal is validated against non-invasive measure- wrist is allowed. ment data, it might show inaccurate data compared to inva- Accuracy of the tonometry sensor was associated with sive measurements. There are also several other limitations BMI and wrist circumference, partly describing the same for non-invasive cuff-based methods to be used as refer- patients. The worst results were obtained for patients with ence. Factors such as arm diameter, cuff size and placement BMI >29 kg m− 2 and wrist circumference >19 cm. In pre- of cuff ass well as differences between manufacturers may vious studies, diabetes that may harden the arteries was not affect the measurement [29]. associated with poorer performance of the device [22]. The Our study has a number of limitations. First, we did not poor accuracy found in our study is also presented in the use the actual BPro ® device for the measurements. Instead, trending analysis. After removing data points located at an we embedded the sensor in our own measurement platform exclusion zone of 3 mmHg (348 data points), 1128 meas- and used the signal analysis service independently. This left urements were left for the analysis and within those, the us unable to define the reference pressure simultaneously probability to predict the direction of change was less than with the measurements due to technical difficulties; instead, 50%. However, due to the relatively stable blood pressure it was done manually during post-hoc analysis. Therefore, readings obtained from the patients, the results provided by we had to choose a good reference point where the study the trending analysis should not be given too much value device and the reference both showed good signal qual- when evaluating the performance of the measurement ity. This corresponds to what takes place with the BPro ® method. The exclusion zone of 3 mmHg is similar to that device, where the device is calibrated at the beginning of used previously [19], but it may be questioned whether it the measurement and is intended to function independently is correct. thereafter. Second, the measurements were performed on Atherosclerosis reduces arterial wall compliance [23] opposite sides while both devices were placed in the wrist while simultaneously increasing systolic pressure. There- area. However, we verified the absence of a pressure dif- fore, it can be assumed to affect tonometry measurements ference between arms by measuring non-invasive blood by decreasing movement in the arterial wall. The accu- pressure with a tourniquet from both arms prior to and after racy of the device was also slightly lower in patients with study measurements, and we found a low mean. Thirdly, PAD, although the power of the study was insufficient to synchronization of the measurements was performed dur- show the difference. The effect of movement is not entirely ing post-hoc analysis, which could have led to a systematic clear. One explanation is that the window length for move- temporal error. Blood pressure measurements were rela- ment detection was relatively long at 40 s; thus, the indica- tively stable during all cases, without marked deviations tor may have shown movement even if it was not actually occurring during the measurement period. We also tried to occurring at the time of the pressure measurement. In addi- optimize the synchronization using the local time signal. tion, even if a small part of the 10 s measurement period Fourthly, the number of arrhythmias and use of vasoactive contains good signal, the calculation for the blood pressure agents was minor limiting the generalization of our results. may be successful with no major effect on the accuracy. Lastly, the recordings were performed under constant mon- The BPro® sensor has also been used to measure cen- itoring of patients typical to postoperative care. Thus, the tral aortic blood pressure. One study compared the BPro® results might be worse when the accuracy of the measure- device to SphygmoCor ™ and invasive measurement [24], ments is not constantly verified. where patients were measured during elective cardiac catheterization. Patients were included who had cardiac diseases but were not reported to have PAD. A second study 5 Conclusions compared BPro® directly to the SphygmoCor™ device [25]. The comparison for central blood pressure showed good Our study shows inaccurate agreement with respect to agreement between the devices or with traditional meas- systolic and mean pressure measurements between the urement in both studies. In both studies, the mean systolic BPro® tonometry sensor and invasive arterial measure- pressures were lower compared to our study. ment in a post-operative care setting. Diastolic pressure

1 3 J Clin Monit Comput measurements were within the recommendations by 10. Shrivastava A, Gupta V. Methods for the determination of limit AAMI. Even slight movement affected the rate of suc- of detection and limit of quantitation of the analytical methods. Chron Young Sci. 2011;2:21–5. cessful measurements. Caution should be taken, espe- 11. Non-invasive sphygmomanometers—Part 2: Clinical investi- cially when estimating systolic pressures based on gation of Automated Measurement type, ANSI/ AAMI/ ISO tonometry measurements. 81060–2:2013. Arlington: Association for the Advancement of Medical instrumentation; 2013, pp 1–22. Acknowledgements The authors wish to thank Finnish Cultural 12. Zou GY. Confidence interval estimation for the Bland–Altman Foundation, Pirkanmaa Regional Fund (4/2014 J.H.), The Finnish limits of agreement with multiple observations per individual. Society of Anaesthesiologists and Paulo Foundation (12/2015 J.H.) Stat Methods Med Res. 2013;22:630–42. for grants, as well as the Medieta Oy (Helsinki, Finland), for provid- 13. Bland JM, Altman DG. Measuring agreement in method com- ing the study device. parison studies. Stat Methods Med Res. 1999;8:135–60. 14. Ilies C, Bauer M, Berg P, Rosenberg J, Hedderich J, Bein B, Compliance with ethical standards Hinz J, Hanss R. Investigation of the agreement of a continuous non-invasive arterial pressure device in comparison with invasive radial artery measurement. Br J Anaesth. 2012;108:202–10. Conflict of interest NO has been a shareholder on former company 15. Saugel B, Grothe O, Wagner JY. Tracking changes in cardiac Medieta. Antti Vehkaoja, Pekka Kumpulainen and Stefano Campadel- output: statistical considerations on the 4-quadrant plot and the lo have been former employees on former company Medieta. Jarkko polar plot methodology. Anesth Analg. 2015;121:514–24. Harju, Ville Lindroos and Arvi Yli-Hankala declare no conflicts of 16. Cecconi M, Rhodes A, Poloniecki J, Della Rocca G, Grounds interest. RM. Bench-to-bedside review: the importance of the precision of the reference technique in method comparison studies–with spe- Ethical approval This study was approved by the Pirkanmaa Hos- cific reference to the measurement of cardiac output. Crit Care. pital district ethics committee. All procedures involving human par- 2009;13:201. ticipants were performed in accordance with the ethical standards of 17. Saugel B, Meidert AS, Hapfelmeier A, Eyer F, Schmid RM, the institutional and/or national research committee, as well as with Huber W. Non-invasive continuous arterial pressure measure- the 1964 Helsinki declaration and its later amendments or comparable ment based on radial artery tonometry in the intensive care unit: ethical standards. a method comparison study using the T-Line TL-200pro device. Br J Anaesth. 2013;111:185–90. 18. Meidert AS, Huber W, Hapfelmeier A, Schofthaler M, Muller JN, Langwieser N, Wagner JY, Schmid RM, Saugel B. Evalu- ation of the radial artery applanation tonometry technology References for continuous noninvasive blood pressure monitoring compared with central aortic blood pressure measurements in patients with multiple organ dysfunction syndrome. J Crit Care. 1. Campbell NR, Chockalingam A, Fodor JG, McKay DW. Accu- 2013;28:908–12. rate, reproducible measurement of blood pressure. CMAJ. 19. Meidert AS, Huber W, Muller JN, Schofthaler M, Hapfelmeier 1990;143:19–24. A, Langwieser N, Wagner JY, Eyer F, Schmid RM, Saugel B. 2. Bause GS, Weintraub AC, Tanner GE. Skin avulsion during Radial artery applanation tonometry for continuous non-invasive oscillometry. J Clin Monit. 1986;2:262–3. arterial pressure monitoring in intensive care unit patients: com- 3. Lin CC, Jawan B, de Villa MV, Chen FC, Liu PP. Blood pres- parison with invasively assessed radial arterial pressure. Br J sure cuff compression injury of the radial nerve. J Clin Anesth. Anaesth. 2014;112:521–8. 2001;13:306–8. 20. Janelle GM, Gravenstein N. An accuracy evaluation of the 4. Slogoff S, Keats AS, Arlund C. On the safety of radial artery T-Line Tensymeter (continuous noninvasive blood pressure man- cannulation. Anesthesiology. 1983;59:42–7. agement device) versus conventional invasive radial artery moni- 5. Chim H, Bakri K, Moran SL. Complications related to radial toring in surgical patients. Anesth Analg. 2006;102:484–90. artery occlusion, radial artery harvest, and arterial lines. Hand 21. Kim SH, Lilot M, Sidhu KS, Rinehart J, Yu Z, Canales C, Can- Clin. 2015;31:93–100. nesson M. Accuracy and precision of continuous noninvasive 6. Komori T, Eguchi K, Hoshide S, Williams B, Kario K. Com- arterial pressure monitoring compared with invasive arterial parison of wrist-type and arm-type 24-h blood pressure pressure: a systematic review and meta-analysis. Anesthesiology. monitoring devices for ambulatory use. Blood Press Monit. 2014;120:1080–97. 2013;18:57–62. 22. Theilade S, Joergensen C, Persson F, Lajer M, Rossing P. Ambu- 7. Nair D, Tan SY, Gan HW, Lim SF, Tan J, Zhu M, Gao H, Chua latory tonometric blood pressure measurements in patients with NH, Peh WL, Mak KH. The use of ambulatory tonometric diabetes. Diabetes Technol Ther. 2012;14:453–6. radial arterial wave capture to measure ambulatory blood pres- 23. Briet M, Boutouyrie P, Laurent S, London GM. Arterial stiff- sure: the validation of a novel wrist-bound device in adults. J ness and pulse pressure in CKD and ESRD. Kidney Int. Hum Hypertens. 2008;22:220–2. 2012;82:388–400. 8. Ng KG, Ting CM, Yeo JH, Sim KW, Peh WL, Chua NH, Chua 24. Ott C, Haetinger S, Schneider MP, Pauschinger M, Schmieder NK, Kwong F. Progress on the development of the MediWatch RE. Comparison of two noninvasive devices for measurement ambulatory blood pressure monitor and related devices. Blood of central systolic blood pressure with invasive measurement Press Monit. 2004;9:149–65. during cardiac catheterization. J Clin Hypertens (Greenwich). 9. Harju J, Vehkaoja A, Lindroos V, Kumpulainen P, Liuhanen S, 2012;14:575–9. Yli-Hankala A, Oksala N. Determination of saturation, heart 25. Garcia-Ortiz L, Recio-Rodriguez JI, Canales-Reina JJ, Cabrejas- rate, and respiratory rate at forearm using a Nellcor\texttrade- Sanchez A, Gomez-Arranz A, Magdalena-Belio JF, Guenaga- mark forehead SpO2-saturation sensor. J Clin Monit Comput. Saenz N, Agudo-Conde C, Gomez-Marcos MA, EVIDENT 2016. Doi:10.1007/s10877-016-9940-7. Group. Comparison of two measuring instruments, B-pro and

1 3 J Clin Monit Comput

SphygmoCor system as reference, to evaluate central systolic pressure in critically ill patients: an observational study. Crit blood pressure and radial augmentation index. Hypertens Res. Care. 2006;10:R43. 2012;35:617–23. 29. Pickering TG, Hall JE, Appel LJ, Falkner BE, Graves J, Hill 26. Wax DB, Rubin P, Neustein S. A comparison of transmittance MN, Jones DW, Kurtz T, Sheps SG, Roccella EJ, Subcommit- and reflectance pulse oximetry during vascular surgery. Anesth tee of Professional and Public Education of the American Heart Analg. 2009;109:1847–9. Association Council on High Blood Pressure Research. Rec- 27. van Egmond J, Hasenbos M, Crul JF. Invasive v. non-invasive ommendations for blood pressure measurement in humans and measurement of arterial pressure. Comparison of two automatic experimental animals: part 1: blood pressure measurement in methods and simultaneously measured direct intra-arterial pres- humans: a statement for professionals from the Subcommittee of sure. Br J Anaesth. 1985;57:434–44. Professional and Public Education of the American Heart Asso- 28. Mignini MA, Piacentini E, Dubin A. Peripheral arterial blood ciation Council on high blood pressure research. Hypertension. pressure monitoring adequately tracks central arterial blood 2005;45:142.

1 3 J Clin Monit Comput (2017) 31:1019–1026 DOI 10.1007/s10877-016-9940-7

ORIGINAL RESEARCH

Determination of saturation, heart rate, and respiratory rate TM at forearm using a Nellcor forehead SpO2-saturation sensor

1 2 3 2 Jarkko Harju • Antti Vehkaoja • Ville Lindroos • Pekka Kumpulainen • 4 1,3 3,5 Sasu Liuhanen • Arvi Yli-Hankala • Niku Oksala

Received: 18 April 2016 / Accepted: 7 October 2016 / Published online: 17 October 2016 Ó Springer Science+Business Media Dordrecht 2016

Abstract Alterations in arterial blood oxygen saturation, man’s correlation (r) of 0.142, and an RMSE of 4.2 points. heart rate (HR), and respiratory rate (RR) are strongly For HR measurements, the mean difference was 0.6 bpm associated with intra-hospital cardiac arrests and resusci- (SD, 2.5), r = 0.997, and RMSE = 1.8. For RR, the mean tations. A wireless, easy-to-use, and comfortable method difference was -0.5 1/min (4.1), r = 0.586, and for monitoring these important clinical signs would be RMSE = 4.0. The SpO2 readings showed a low mean highly useful. We investigated whether the NellcorTM difference, but also a low correlation and high RMSE, OxiMask MAX-FAST forehead sensor could provide data indicating that the NellcorTM saturation sensor cannot for vital sign measurements when located at the distal reliably assess oxygen saturation at the forearm when forearm instead of its intended location at the forehead to compared to finger PPG measurements. provide improved comfortability and easy placement. In a prospective setting, we recruited 30 patients undergoing Keywords Intraoperative monitoring Á Plethysmography Á surgery requiring postoperative care. At the postoperative Pulse oximetry Á Heart rate Á Respiratory rate care unit, patients were monitored for two hours using a standard patient monitor and with a study device equipped TM with a Nellcor Forehead SpO2 sensor. The readings were 1 Introduction electronically recorded and compared in post hoc analysis using Bland–Altman plots, Spearman’s correlation, and Worsening vital signs are strongly associated with intra- root-mean-square error (RMSE). Bland–Altman plot hospital cardiac arrest and resuscitation [1]. The clinical showed that saturation (SpO2) differed by a mean of signs most commonly associated with patient deterioration -0.2 % points (SD, 4.6), with a patient-weighted Spear- are arterial blood oxygen saturation (SpO2), heart rate (HR), respiratory rate (RR), and blood pressure [2]. There is presently no single device capable of wirelessly monitoring all of these parameters, although several reports describe attempts to develop such a device [3–5]. TM & Jarkko Harju The Nellcor Forehead SpO2 sensor is a reflectance Jarkko.harju@fimnet.fi mode photoplethysmography (PPG) sensor that was developed to measure SpO2 and HR at the forehead [6–9]. 1 Department of Anesthesia, Tampere University Hospital, Studies of the saturation measurement of the use of this PL2000, 33521 Tampere, Finland sensor on the forehead report it to be accurate, with mea- 2 Tampere University of Technology, Tampere, Finland surements comparable to those acquired using traditional 3 Medical School, University of Tampere, Tampere, Finland transmittance mode sensors on the finger [6, 9]. Saturation 4 Department of Anesthesia, Helsinki University Hospital, and HR have been measured successfully at wrist and Helsinki, Finland palmar region in infants using sensors based on light 5 Department of Surgery, Tampere University Hospital, absorbance. [10, 11] However, in adults the skin thickness Tampere, Finland is too great to reach accurate measurements. One previous 123 1020 J Clin Monit Comput (2017) 31:1019–1026 study described the performance of saturation sensors on 2.1 Monitoring the wrist area [10], but to our knowledge Nellcor forehead sensor has not been tested on distal forearm. Based on our Standard care included patient monitoring with a GE pilot measurements, distal forearm region showed CarescapeTM B650 monitor (GE Healthcare Oy, Helsinki, promising results for saturation measurement. Another Finland) using a TruSignalTM saturation sensor on the study compared the power spectrum of PPG recordings at finger. Recordings from this monitor were collected at 10-s different sites, reporting the greatest power at the fingers intervals using S5 CollectTM software (GE Healthcare Oy, for HR and at the forearm for RR [12]. The HR signal was Helsinki, Finland) for all of the study parameters. SpO2 and also present at distal forearm. Combinining this informa- HR measurements from the standard monitor were based tion, distal forearm region might have potential to monitor on finger measurements, while RR was based on impe- saturation, HR and RR using just a single sensor. dance measurement from the ECG electrodes. In clinical monitoring, RR is typically estimated using In addition to standard monitoring, patients were the impedance pneumography from electrocardiogram equipped with a study device mounted in a single unit (ECG) leads. However, the measurement of electrical using a rechargeable battery, which was wrapped around impedance has marked limitations associated with chest the forearm with a flexible band. The study device com- wall movements. Multiple factors, such as incorrect elec- prised a NellcorTM OxiMask MAX-FAST forehead sensor trode placement, coughing, and crying, may generate a for SpO2 (Covidien, MN, USA) along with a 3-D high number of inaccurate readings [13, 14]. The respira- accelerometer (Freescale MMA8452Q; NXP, Eindhoven, tion-associated alterations on PPG waveform remain partly Netherlands). A Faros ECG recorder (Mega Electronics, unknown [15], but multiple changes in the waveform Oulu, Finland) was used to obtain a three-lead ECG to caused by RR have been described. [16–19] RR measure- allow additional synchronizing. The study device sensor ment using PPG has been compared to thoracic impedance was placed over the radial bone at the distal forearm near under postoperative care conditions, showing promising the wrist, and correct placement was indicated by an LED results, with a low proportion of false-positive breath controlled by the signal-processing unit of the SpO2 sensor. recordings (4.6 ± 4.5 %) [20]. More recent studies have Optimal placement was determined using pre-clinical reported similar results [21–23]. voluntary pilot measurements, which indicated that place-

SpO2 and HR are typically measured at the finger, ment over the radial bone enabled the strongest signal. The forehead, or ear lobe. While sensor fixation represents a study device recorded SpO2, HR, and RR every 10 s, and marked challenge, the use of a wristband at distal forearm accelerometer data at 10-Hz frequencies. The data were could offer an easier and more comfortable site for the sent via Wi-Fi connection to an Internet remote server, and sensors. Here we investigated off-label use of the Nell- the data were viewed online on a web page and obtained corTM PPG sensor at a previously unreported location—the electronically after measurements. distal forearm. The device was also equipped with an The RR was obtained from the PPG sensor at the fore- accelerometer to evaluate whether accuracy could be arm using a novel algorithm (Oksala N and Liuhanen S, improved by excluding measurements taken during Patent WO2015107268 A1) [24]. Our present report is the movement. This feasibility study aimed to assess whether a first to describe the performance of this algorithm. Briefly, readily available sensor could reliably wirelessly monitor the algorithm extracts a good quality beat series by the common clinical signs of SpO2, HR, and RR at a single removing normal noise, and then interpolating missing location. beats and deleting erroneously detected ones. From the original signal, the algorithm extracts four primary sub- 2 Materials and methods signals and two derived sub-signals, each of which is independently analyzed in both time and frequency This observational study included 30 patients (mean age of domains. The changes in the PPG waveform that were used 67[SD = 13] years), who were recruited between January consisted of baseline modulation, amplitude modulation, and May 2015 at Tampere University Hospital. Each col- respiratory sinus arrhythmia and pulse wave width. Finally, lection peried lasted for two hours during post-operative good quality estimates are combined to obtain a RR care treatment. The patients were undergoing elective estimate. surgery requiring invasive blood pressure measurement. The values obtained from the standard monitor and the Patients with an implanted cardiac pacemaker were study device were synced using the local time shown on the excluded. All patients gave their written informed consent Internet web page. To allow additional syncing, the arterial prior to study entry, and the study was approved by the line was occluded and the ECG connection from the Faros local ethics committee (ETL R13145). simultaneously removed. An independent observer (J.H.

123 J Clin Monit Comput (2017) 31:1019–1026 1021 and V.L.) recorded all device and patient movements, and the Mann–Whitney U or Kruskal–Wallis test for the verified the data collection accuracy. RMSE. p value \ 0.05 was considered statistically signif- In order to compare the test monitor the PPG-derived icant and RMSE [ 4.0 clinically unacceptable. RR was additionally compared against a validation data described by Charlton et al. [19], which is commonly available online for study purposes [25]. The RR from the 3 Results dataset was recalculated using our algorithm [24] and compared against impedance measurement provided along Of the 30 patients studied, 13 had peripheral arterial dis- the data. ease. In two patients, the recording was unsuccessful due to The accelerometer data were used to test whether a breakdown of the sensor cable during measurement, measurement accuracy could be improved by excluding the leaving 28 patients for final analysis. There were a total of readings obtained during movement. We used the sum- 10,767 SpO2, 14,832 HR and 18,857 RR pairs left for mation value for three-dimensional movement. Thereafter, comparisons. Table 1 describes the patients’ characteris- the threshold for movement detection was formed by tics. The following types of surgery were performed prior defining a mean value and standard deviation for the lowest to measurements: vascular surgery in 13 patients (46.4 %), 90 % of the summation of the acceleration data. We cal- gastroenterological surgery in 8 patients (28.6 %), urologic culated the mean plus three times the standard deviation for surgery in 2 patients (7.1 %), plastic surgery in 3 patients use as a threshold for ‘‘slight movement’’, and the mean (10.7 %), and orthopedic surgery in 2 patients (7.1 %). plus ten times the standard deviation for use as the Patients were predominantly lying in the supine position. threshold for ‘‘definite movement’’. This limit definition In the Bland–Altman plot, we found a low mean dif- was performed using a modification of a commonly used ference in SpO2 [-0.3 % points (95 % confidence interval technique for detection limit determination [26]. -3.9 to 5.1), Limits of agreement (LoA)upper: 7.2 (CI 6.1 to 8.1), LoAlower -7.9 (CI -8.8 to -6.8)] with a low patient- 2.2 Power analysis weighted Spearman’s correlation between devices (r = 0.142). The HR showed a low mean difference

The power analysis was based on the results of a previous [0.6 bpm (CI -0.85 to 2.05), LoAupper 5.6 (CI 5.2 to 5.9). study that compared measurements of reﬂectance and LoAlower -4.4 (CI -4.8 to -4.0)] and a good correlation transmittance mode saturation sensors at the forehead and (r = 0.997). RR differed by -0.6 cycles (CI -3.9 to 5.1,

ﬁngers, reporting a bias difference of -1.39 compared to LoAupper 8.0 (CI 7.2 to 8.6), LoAlower -6.8 (CI -7.4 to -2.61 units (SD, 1.3) [9]. Assuming similar bias differ- -6.0), with a moderate correlation (r = 0.586). The RMSE ences in the present study, we would need 19 patients to comparison was 4.2 points for SpO2, 4.0 points for RR, and achieve a power of 0.80 (p \ 0.05). To account for the 1.8 points for HR (Fig. 1; Table 2). possibility of patient drop-out, we recruited 30 patients. There were a total 25,154 data comparisons for 39 subjects in validation data set. The method comparison in 2.3 Statistical analysis validation data for RR revealed a bias of 4 breaths (CI 4.0 to 4.2), while the RMSE comparison was 5.6 (Table 2). RR

Data were analyzed using IBM SPSS statistics version 23 differed by 4.1 breaths (CI -2.7 to 10.8), LoAupper 11.9 (CI (IBM, IL, USA) and with the method described in a pre- 11.1 to 12.7), LoAlower -3.8 (CI -3.0 to -4.7) as descri- vious publication [27] using Microsoft Excel 2010 (Mi- bed in Fig. 2. crosoft, Redmond, WA, USA) for Bland–Altman Plot. Results are reported using the method described by Bland Table 1 Patient characteristics, described as mean (SD) or frequency (%) and Altman for multiple comparisons [28]. To determine the bias of the novel method compared to traditional Characteristics n = 30 measurement, we calculated the Spearman’s correlation Age (years) 67 (13) between the devices and the root-mean-square error Gender F/M 13/17 (43 %/57 %) (RMSE) to describe the mean error. The correlation Height (cm) 172 (10) between datapoints was calculated patient-weighted as a Weight (kg) 80 (16) correlation between each patients values. The sensor ASA I/II/III/IV 1/7/21/1 comparison data were found to be scattered, and are Peripheral artery disease 13 (43 %) reported as median and interquartile range (IQR; 25–75th Atrial ﬁbrillation 3 (10 %) percentile). The patient demographics are reported as mean Coronary artery disease 1 (3 %) (SD) as for normally distributed data. The p value was Forearm circumference (cm) 18 (2) calculated using the Wilcoxon signed-rank test, or using 123 1022 J Clin Monit Comput (2017) 31:1019–1026

Fig. 1 Bland–Altman plot and Spearman correlation for saturation measurements. In BA plot black lines represent mean and limits of (a), heart rate (b) and respiratory rate (c). The data is presented with a agreement, grey areas describe the 95 % conﬁdence intervals Bland–Altman plot (BA) and Spearman correlation between all respectively Slight movement was detected during 12.6 % of the 4 Discussion total recording time. Movement was associated with higher

RMSE in all tested parameters. For SpO2, the difference The present study investigated the performance of a readily was significant (p = 0.037) but of a small magnitude available PPG sensor for measuring common clinical signs (Table 3). RMSE did not differ in association with at a single location. Our main finding was that the forehead peripheral arterial disease, body mass index (BMI), or PPG sensor could not reliably detect SpO2 at the forearm forearm circumference. when compared to finger PPG recordings. In particular, the

123 J Clin Monit Comput (2017) 31:1019–1026 1023

Table 2 Patient-weighted value described as median (IQR) and as root-mean-square error (RMSE) (IQR), Spearman correlation and p value for the difference between test device and reference monitor (B650) (n = 28) or validation data. (n = 38) Plethysmography Min, max References Min, max RMSE Spearman correlation p

SPO2 (%) 97.5 (95.6–99.2) 86, 100 97.0 (95.5–97.8) 91, 100 4.2 (2.8–5.7) 0.142 0.399 HR (bpm) 67.7 (60.0–73.2) 48, 94 67.6 (59.4–73.9) 47, 93 1.8 (1.6–2.9) 0.997 0.001 RR (1/min) 13.2 (12.3–14.4) 11, 28 12.1 (10.5–14.1) 9, 30 4.0 (3.1–4.7) 0.586 0.048 RR validation 12.1 (10.7–14.0) 9, 19 16.5 (13.6–19.0) 7, 26 5.6 (1.7–6.8) 0.416 \0.001

Fig. 2 Bland–Altman plot and Spearman correlation for HR in measurements. In BA plot black lines represent mean and limits of validation data. The data is presented with a Bland–Altman plot (BA) agreement, grey areas describe the 95 % conﬁdence intervals for multiple measurements and Spearman correlation between all respectively

RMSE was outside of the 2 % accuracy range required by a especially when placed on the forehead [6, 23, 31]. Here, previous recommendation [29] and 4.0 % recommended by we instead placed the sensor over the radial bone, a surface ISO standard [30]. Respiratory rate measurements showed that is more rounded and potentially more prone to a moderate patient-weighted correlation, and a moderately movement. Moreover, in the forearm location, there is a high RMSE, and Bland–Altman plots of HR recordings much longer distance between the device and the reﬂecting showed a small mean difference and a very strong corre- bone due to fatty tissue. Our results showed that RMSE did lation between the devices. Thus, our ﬁndings indicated not differ in relation to movement, BMI, or forearm that HR could be reliably measured by a PPG sensor at the circumference. forearm. The accuracy of RR, measured at forearm, was The gold standard method for monitoring SpO2 is arte- moderate. The performance of SpO2 was poor. rial oxygen saturation; however, this requires invasive measurements [29]. Motion artefacts and hypoperfusion

4.1 Saturation are the most common causes of inaccurate SpO2 measurement at the fingers [32]. We detected a small but sig- Our results showed that when placed at forearm instead of nificant change in accuracy associated with movement. intended location at forehead, oximetry measurements During critical illness, centralization of blood circulation were inaccurate compared to finger oximetry measure- decreases the accuracy of the SpO2 measurement at the ments. Previous evaluation of the accuracy of finger sensor fingers [6]. As all of our patients were electively treated, oximetry reported by the manufacturer revealed an RMSE very few SpO2 readings were below 90 %. of ±2(± 3 during motion). The mean difference in our There was a low correlation and high RMSE found in present study was small, but the correlation between the our study. The fingers are sensitive to mild hypothermia devices was poor and the RMSE describing the error in [33], which could result in a lower SpO2 in the fingers forehead sensor measurement was unacceptably high. The compared to in the distal forearm area. During our study NellcorTM OxiMask MAX-FAST sensor detects the measurements, the forearm and the device were constantly reflection of light, and several reports describe its efficacy, visible, which could have influenced the temperature in

123 1024 J Clin Monit Comput (2017) 31:1019–1026

ﬁngers. Unfortunately, we were unable to measure arterial Bonf

- oxygen saturation, which could have shed more light on this subject.

4.2 Heart rate pp

HR measurement was found to be very accurate, with a high correlation between the devices. This is in line with previous ﬁndings suggesting good accuracy [3–5]. Indeed,

RMSE the RMSE for HR measurement was constantly low despite all confounding factors. Movement was the only factor that signiﬁcantly affected HR measurement and the difference

indicates number of patients or measurements. between the devices was still low. Previous studies eval-

n uating wrist-based optical HR monitors have focused on ﬁtness devices used by young healthy people, and there -Bonf).

p exists no prior data in hospital patients.

4.3 Respiratory rate 0.001* 0.005* 799 4.1 (3.1–4.9) 0.007* 0.035* 0.001* 0.005* 539 4.1 (3.1–4.9) 0.037 0.185 p p-Bonf n \ \ Our RR ﬁndings showed a good accuracy and moderate ) in different groups 2 patient-weighted correlation between the devices. These ﬁndings are in line with those of Nilsson et al. [12] who reported that the distal forearm area is a suitable location

RMSE for RR and HR measurement, although they found that spectral power was lower for HR than for RR. Another previous study compared RR calculated from ﬁnger PPG signal with RR from capnogram monitoring [21], and reported difference similar to that found in our study. Fingers are highly susceptible to movement, making Bonf n - measurement more difﬁcult in mobile patients. Our patients were mostly lying, and slight or stronger movement of the forearm was detected during only 12.6 % of 0.001* 0.005* 2340 4.1 (3.3–4.7) 0.001* 0.005* 1811 4.1 (3.7–4.7) the measurement period. pp \ \ The comparison of the study algorithm against valida- value calculated between or within groups, Bonferroni corrected (

p tion data [19] showed worse performance than against our patient data as decribed in Table 2. There might be several reasons for this difference. Our study data is recorded from

RMSE distal forearm region where the RR signal has a greater power range [12]. The validation data patients were also moving which increases the diffulty in measuring the RR

HRn RRadequately. Furthermore, SPO2 in the validation data, the plethysmographic waveform was collected using ﬁnger sensor based on absorption, thus the anatomic site and used 1720 10 2.4 (1.9–3.3) 6 0.166 1.7 (1.6–3.2) 0.83 10 4.0 (3.0–4.2) 0.89 1 6 2.9 (2.5–4.7) 10 4.1 (3.0–5.4) 0.243 1 6 2.9 (1.9–5.0) 2430 9 2.4 (1.6–4.3) 10 1.7 0.31 (1.6–2.8) 1 9 4.0 (3.5–4.9) 10 0.207 3.2 (2.7–4.3) 1 9 3.7 (2.5–4.7) 0.323 1 10 5.3 (2.4–6.6) 17–20 12 1.7 (1.5–2.3) 12 4.3 (3.4–5.1) 12 4.7 (3.4–6.2) 24–30 9 1.8 (1.5–2.2) 9 4.1 (3.2–4.8) 9 4.2 (3.5–5.7) No 12,961 1.9 (1.5–3.0) 16,517 3.8 (3.0–4.5) 9968 4.1 (2.5–5.2) No 13,469 1.9 (1.5–3.0) 17,046 3.8 (3.0–4.6) 10,228 4.1 (2.5–5.3) No 15 1.9 (1.7–3.0) 15 4.1 (3.0–4.8) 15 4.9 (3.2–6.2) \ [ \ [ sensor technique were different to our own measurements. Estimating RR from a peripheral location is a complex

task. As with SpO2, typical obstacles to RR measurement include movement artifacts and signal amplitude due to low perfusion [15, 19]. The peripheral arterial disease may also affect the measurement, although our study could not ) 2 demonstrate the difference. It is often seen that upper Root-mean-square error (RMSE) calculated for heart rate (HR), respiratory rate (RR) and saturation (SpO extremity arteries remain intact, although lower extremity 0.05

\ arteries are seriously affected by peripheral artery disease. p Slight movement Yes 1871 2.4 (1.8–2.8) Forearm circumference (cm) Deﬁnite movement Yes 1363 2.4 (1.9–2.8) Peripheral arterial diseaseBMI (kg/m Yes 13 1.8 (1.5–3.0) 0.467 1 13 4.0 (3.2–4.3) 0.751 1 13 4.0 (2.6–4.6) 0.274 1 Each patient’s RMSE described with median (interquartile range). Table 3 * [34] Here we were able to achieve the reported accuracy by 123 J Clin Monit Comput (2017) 31:1019–1026 1025 utilizing a strategy that treats different features of the PPG Acknowledgments The authors wish to thank the Paulo Foundation, signal as independent components, and by continuously Finnish Society of Anaesthesiologists and the Finnish Cultural Foundation, Pirkanmaa Regional Fund for grants, as well as the selecting which ones to use based on their quality. Medieta Oy (Helsinki, Finland), which provided the study device.

4.4 Limitations Compliance with ethical standards

Conflicts of interest SL and NO have a pending patent on RR The present study has several limitations. Firstly, the measurement. NO has been a shareholder on a former company inclusion of only patients who were undergoing invasive Medieta. AV and PK have been former employees on a former blood pressure measurement selected patients who had company Medieta. JH, VL, and AY-H declare no conflicts of interest. many comorbidities, and were undergoing major surgery with a high operation risk. However, the recordings were Ethical approval This study was approved by the Pirkanmaa Hospital district ethics committee. All procedures involving human taken at a postoperative care unit under relatively participants were performed in accordance with the ethical standards stable conditions; thus, the findings are likely not specific of the institutional and/or national research committee, as well as with the 1964 Helsinki declaration and its later amendments or comparable to conditions of low SpO2 or low blood perfusion. Addi- tionally, all patients were continuously observed, and study ethical standards. personnel screened the signal quality throughout the monitoring period. Secondly, the patients showed rela- References tively stable hemodynamics during the study period, and were lying in a supine position. Therefore, our findings 1. Schein RM, Hazday N, Pena M, Ruben BH, Sprung CL. Clinical may not be generalized to patients with hypotension or antecedents to in-hospital cardiopulmonary arrest. Chest. delirium. Moreover, all patients had an incremental supply 1990;98:1388–92. of oxygen. Thirdly, the recordings were acquired using two 2. Jones DA, DeVita MA, Bellomo R. Rapid-response teams. N Engl J Med. 2011;365:139–46. different devices and were synchronized in post hoc-anal- 3. Fukushima H, Kawanaka H, Bhuiyan MS, Oguri K. Estimating ysis and, thus, we were unable to perform a beat-to-beat heart rate using wrist-type Photoplethysmography and accelera- comparison. However, the recordings were taken over a tion sensor while running. Conf Proc IEEE Eng Med Biol Soc. short time-period, and were started and ended at the same 2012;2012:2901–4. 4. Renevey P, Sola J, Theurillat P, Bertschi M, Krauss J, Andries D, time with both devices, decreasing the likelihood of a Sartori C. Validation of a wrist monitor for accurate estimation of major desynchronization bias. Lastly the comparison to RR intervals during sleep. Conf Proc IEEE Eng Med Biol Soc. validation data was performed at a post hoc analysis and 2013;2013:5493–6. the site of measurement was different than that used in the 5. Zhang Z, Pi Z, Liu B. TROIKA: a general framework for heart rate monitoring using wrist-type photoplethysmographic signals present study Therefore the value of the comparison to during intensive physical exercise. IEEE Trans Biomed Eng. validation data is limited. 2015;62:522–31. 6. Nesseler N, Frenel JV, Launey Y, Morcet J, Malledant Y, Seguin P. Pulse oximetry and high-dose vasopressors: a comparison between forehead reflectance and finger transmission sensors. 5 Conclusions Intensive Care Med. 2012;38:1718–22. 7. Sugino S, Kanaya N, Mizuuchi M, Nakayama M, Namiki A. Forehead is as sensitive as finger pulse oximetry during general Only a limited selection of currently available devices are anesthesia. Can J Anaesth. 2004;51:432–6. capable of wirelessly monitoring vital signs [35, 36], and 8. Agashe GS, Coakley J, Mannheimer PD. Forehead pulse none can collect all commonly monitored clinical signs oximetry: headband use helps alleviate false low readings likely related to venous pulsation artifact. Anesthesiology. from a single site. Here we report that, compared to 2006;105:1111–6. recordings from a finger sensor, SpO2 measurement at the 9. Schallom L, Sona C, McSweeney M, Mazuski J. Comparison of wrist showed a poor correlation and a high RMSE. With forehead and digit oximetry in surgical/trauma patients at risk for regards to RR, the mean difference in the Bland–Altman decreased peripheral perfusion. Heart Lung. 2007;36:188–94. 10. Safar H, El-Dash H. Pulse oximetry: Could wrist and ankle be plot was small and Spearman’s correlation was moderate, alternative placement sites? Clin Pediatr (Phila). 2015;54:1375–9. but the RMSE was markedly high. Also the RMSE was 11. Phattraprayoon N, Sardesai S, Durand M, Ramanathan R. higher in the validation data. The study device’s mea- Accuracy of pulse oximeter readings from probe placement on surement of HR showed good accuracy and correlation. To newborn wrist and ankle. J Perinatol. 2012;32:276–80. 12. Nilsson L, Goscinski T, Kalman S, Lindberg LG, Johansson A. our knowledge, no commercially available sensors indi- Combined photoplethysmographic monitoring of respiration rate cated for medical use can detect SpO2 at the wrist. The and pulse: a comparison between different measurement sites in development of such a device might provide a reliable way spontaneously breathing subjects. Acta Anaesthesiol Scand. to measure common clinical signs at one location. 2007;51:1250–7.

123 1026 J Clin Monit Comput (2017) 31:1019–1026

13. Wiklund L, Hok B, Stahl K, Jordeby-Jonsson A. Postanesthesia 26. Shrivastava A, Gupta V. Methods for the determination of limit monitoring revisited: frequency of true and false alarms from of detection and limit of quantitation of the analytical methods. different monitoring devices. J Clin Anesth. 1994;6:182–8. Chron Young Sci. 2011;2:21–5. 14. Gaucher A, Frasca D, Mimoz O, Debaene B. Accuracy of res- 27. Zou GY. Confidence interval estimation for the Bland–Altman piratory rate monitoring by capnometry using the Capno- limits of agreement with multiple observations per individual. mask(R) in extubated patients receiving supplemental oxygen Stat Methods Med Res. 2013;22:630–42. after surgery. Br J Anaesth. 2012;1(08):316–20. 28. Bland JM, Altman DG. Agreement between methods of mea- 15. Meredith DJ, Clifton D, Charlton P, Brooks J, Pugh CW, Tar- surement with multiple observations per individual. J Biopharm assenko L. Photoplethysmographic derivation of respiratory rate: Stat. 2007;17:571–82. a review of relevant physiology. J Med Eng Technol. 29. Batchelder PB, Raley DM. Maximizing the laboratory setting for 2012;36:1–7. testing devices and understanding statistical output in pulse 16. Johansson A, O¨ berg PA˚ . Estimation of respiratory volumes from oximetry. Anesth Analg. 2007;105:S85–94. the photoplethysmographic signal. Part I: experimental results. 30. Finnish Standards association (2011) Medical electrical equip- Med Biol Eng Comput. 1999;37:42–7. ment—Part 2–61: particular requirements for basic safety and 17. Dash S, Shelley KH, Silverman DG, Chon KH. Estimation of essential performance of pulse oximeter equipment (ISO respiratory rate from ECG, photoplethysmogram, and piezo- 80601-2-61:2011). electric pulse transducer signals: a comparative study of time- 31. Fischer MO, Fornier W, Hanouz JL, Fellahi JL. Cephalic and frequency methods. IEEE Trans Biomed Eng. 2010;57:1099–107. digital pulse oximetry in cardiac surgery: a comparative pilot 18. Lazaro J, Gil E, Bailon R, Minchole A, Laguna P. Deriving study with arterial oximetry. Eur J Anaesthesiol. 2015;32:60–1. respiration from photoplethysmographic pulse width. Med Biol 32. Salyer JW. Neonatal and pediatric pulse oximetry. Respir Care. Eng Comput. 2013;51:233–42. 2003;48:386–96 (discussion 397-8). 19. Charlton PH, Bonnici T, Tarassenko L, Clifton DA, Beale R, 33. MacLeod DB, Cortinez LI, Keifer JC, Cameron D, Wright DR, Watkinson PJ. An assessment of algorithms to estimate respira- White WD, Moretti EW, Radulescu LR, Somma J. The desatu- tory rate from the electrocardiogram and photoplethysmogram. ration response time of finger pulse oximeters during mild Physiol Meas. 2016;37:610–26. hypothermia. Anaesthesia. 2005;60:65–71. 20. Nilsson L, Johansson A, Kalman S. Monitoring of respiratory rate 34. Tendera M, Aboyans V, Bartelink M, Baumgartner I, Clement D, in postoperative care using a new photoplethysmographic tech- Collet J, Cremonesi A, De Carlo M, Erbel R, Fowkes FG, Heras nique. J Clin Monit Comput. 2000;16:309–15. M, Kownator S, Minar E, Ostergren J, Poldermans D, Riambau 21. Addison PS, Watson JN, Mestek ML, Ochs JP, Uribe AA, V, Roffi M, Ro¨ther J, Sievert H, van Sambeek M, Zeller T. ESC Bergese SD. Pulse oximetry-derived respiratory rate in general Guidelines on the diagnosis and treatment of peripheral artery care floor patients. J Clin Monit Comput. 2015;29:113–20. diseases. Eur Heart J. 2011;32:2851–906. 22. Garde A, Karlen W, Ansermino JM, Dumont GA. Estimating 35. Hernandez-Silveira M, Ahmed K, Ang SS, Zandari F, Mehta T, respiratory and heart rates from the correntropy spectral density Weir R, Burdett A, Toumazou C, Brett SJ. Assessment of the of the photoplethysmogram. PLoS ONE. 2014;9:e86427. feasibility of an ultra-low power, wireless digital patch for the 23. Nilsson L, Johansson A, Kalman S. Respiration can be monitored continuous ambulatory monitoring of vital signs. BMJ Open. by photoplethysmography with high sensitivity and specificity 2015;5:e006606. regardless of anaesthesia and ventilatory mode. Acta Anaesthe- 36. Nair D, Tan SY, Gan HW, Lim SF, Tan J, Zhu M, Gao H, Chua siol Scand. 2005;49:1157–62. NH, Peh WL, Mak KH. The use of ambulatory tonometric radial 24. Oksala N, Liuhanen S (2015) Method and device for the detection arterial wave capture to measure ambulatory blood pressure: the of respiratory rate. PCT/FI2015/050023. validation of a novel wrist-bound device in adults. J Hum 25. Charlton PH (2016) In: Synthetic dataset: dataset for verification Hypertens. 2008;22:220–2. of algorhitm implementions. Charlton Peter H. http://peterh charlton.github.io/RRest/synthetic_dataset.html. Accessed 10 Aug 2016.

123