Chapter 1 Introduction

CHAPTER 1

INTRODUCTION

And what of the god of sleep, patron of anaesthesia? The centuries themselves number more than 21 since Hypnos wrapped his cloak of sleep over Hellas. Now before Hypnos, the artisan, is set the respiring flame – that he may, by knowing the process, better the art.

John W. Severinghaus

When, in 1969, John Severinghaus penned that conclusion to his foreword for the first edition of John Nunn’s Applied Respiratory Physiology (Nunn 1969), he probably did not have in mind a potential interaction between surgery, anaesthesia, analgesia and postoperative sleep. It is only since then that we have identified the importance of sleep after surgery and embarked upon research into this aspect of perioperative medicine.

Johns and his colleagues (Johns 1974) first suggested and examined the potential role of sleep disruption in the generation of morbidity after major surgery in 1974. It soon became clear that sleep-related upper airway obstruction could even result in death after upper airway surgery (Kravath 1980). By the mid-eighties, sleep had been implicated as a causative factor for profound episodic hypoxaemia in the early postoperative period (Catley 1985). The end of that decade saw the first evidence for a rebound in rapid eye movement (REM) sleep that might be contributing to an increase in episodic sleep- related hypoxaemic events occurring later in the first postoperative week (Knill 1987; Knill 1990).

Since then, speculation regarding the role of REM sleep rebound in the generation of late postoperative morbidity and mortality (Rosenberg-Adamsen 1996a) has evolved

1 into dogma (Benumof 2001) without any direct evidence to support this assumption. Controversy has also arisen regarding the role of sleep in the causation of early postoperative respiratory disturbances (Rahman 2001; Wu 2003).

Thirty years of research, speculation and comment on the role of sleep and sleep- disordered breathing in the causation of perioperative morbidity have come and gone. In that time almost nothing has been said about the potential for anaesthetists, with the observations they make each day, to make an important contribution to the diagnosis and management of sleep-disordered breathing with its consequent morbidity. The imbalance needs to be redressed.

The research presented in this Thesis was conducted to examine the importance of sleep in postoperative respiratory dysfunction, and to show that the daily observations made by anaesthetists in their management of the airway might be utilised to assist patients with sleep-disordered breathing in the longer term. The former involved two main areas: a search for evidence of a clinically important contribution of REM sleep rebound to postoperative morbidity and a re-examination of the role of sleep in the postoperative causation of episodic hypoxaemic events in general. To assess the latter, a relationship between airway obstruction under anaesthesia and the severity of sleep-disordered breathing was sought.

How ironic it is that Hypnos, the patron of anaesthesia, should also be god of that sleep which could be snuffing out the respiring flame of those same patients. Let us further examine this irony and see if, by knowing the process and bettering the art as Severinghaus exhorted, we may once again bring harmony to the dual roles of Hypnos and keep alight the flame Severinghaus set before him.

2 CHAPTER 2

REVIEW OF THE LITERATURE

Falstaff: Now, Hal, what time of day is it, lad? Prince Henry: Thou art so fat-witted, with drinking of old sack, and unbuttoning thee after supper, and sleeping upon benches after noon…

Act I, Scene II

Poins: Falstaff! - fast asleep behind the arras, and snorting like a horse. Prince Henry: Hark, how hard he fetches breath…

Act II, Scene IV King Henry IV Part I William Shakespeare

Although hypoxaemia in the postoperative period has been recognised for many years (Nunn 1962; Bay 1968; Papper 1971; Spence 1972; Jones 1990), the first examination of ventilatory patterns and their association with sleep in this period was carried out by Catley and his colleagues (Catley 1985). Studying only the first postoperative night, they found that all profound episodic oxygen desaturations in the postoperative period were caused by upper airway obstruction, occurring only during sleep. The effect of sleep on postoperative respiration has been a subject of interest since then. The finding by Knill’s group (Knill 1987) of a rapid eye movement (REM) sleep rebound, after an initial period of postoperative REM suppression, led them and others to speculate on the

3 potential consequences of a REM-related increase in hypoxaemic events in the late postoperative period. Despite a lack of direct evidence, much has since been made of the possible effect of this REM rebound on late postoperative morbidity and mortality.

This review will examine the literature concerning sleep-related respiratory disorders, obstructive sleep apnoea in particular, and investigate what is already known or suspected about the effects of surgery and anaesthesia on sleep and postoperative sleep- related respiratory dysfunction.

SLEEP-RELATED BREATHING DISORDERS

While the first detailed description of obstructive sleep apnoea (OSA) only appeared in 1966 (Gastaut 1966), many had noted its characteristics prior to this. The best known of these descriptions is Dickens' portrayal of Joe the Fat Boy in The Posthumous Papers of the Pickwick Club, although it would seem that Shakespeare had observed the symptoms at least three centuries earlier. Considering Shakespeare's comic intent, it is clear that the audience would have been aware of them also. This suggests the problem to be a common one as several large studies now confirm (Stradling 1991; Young 1993c; Bearpark 1995; Young 2002). These studies demonstrate that between two and four percent of middle aged adults have clinically significant sleep apnoea with a male:female ratio of 2:1.

It has been recently suggested (Sauret Valet 1999) that the symptoms of sleep apnoea had been observed and described nearly two millennia before Gastaut by Pliny the Younger (Gaius Plinius) in his account of the death of his uncle, Pliny the Elder, during the eruption of Vesuvius. A close examination of Pliny’s letter to the historian Tacitus in which he recorded the event, however, indicates Sauret Valet’s suggestion is rather imaginative. In the letter he described the actions of his uncle in Stabiae thus: “To alleviate people's fears my uncle claimed that the flames came from the deserted homes of farmers who had left in a panic with the hearth fires still alight. Then he rested, and gave every indication of actually sleeping; people who passed by his door heard his snores, which were rather resonant since he was a heavy man.” Later in the letter he

4 reports his uncle’s death: “Supported by two small slaves he stood up, and immediately collapsed. As I understand it, his breathing was obstructed by the dust-laden air, and his innards, which were never strong and often blocked or upset, simply shut down. When daylight came again two days after he died…” There is, therefore, little in Gaius Plinius’ letter apart from a description of heavy snoring to suggest his uncle had sleep apnoea.

Definitions

Until recently there has been no broad consensus regarding standard definitions, including thresholds of significance, for many of the terms used to describe sleep- related breathing disturbances. A recent report has addressed these issues (AASM 1999). It is generally but arbitrarily agreed that an apnoea, defined as a cessation of airflow, has to exceed ten seconds duration to be considered significant. Standard definitions for hypopnoea have been even more controversial (Meoli 2001). It is usually defined as a reduction in airflow or respiratory effort for more than ten seconds accompanied by a desaturation of three percent or more and/or electroencephalographic evidence of arousal (AASM 1999; Tsai 1999; Meoli 2001). The apnoea hypopnoea index (AHI) is the number of apnoeas and hypopnoeas per hour of sleep and is used more or less interchangeably with the term respiratory disturbance index (RDI) the term that will be used for the remainder of this thesis.

The apnoeas may be obstructive, central or mixed. Obstructive apnoeas are characterised by persistent effort without airflow, while with central apnoea effort is absent. OSA, where the apnoeas are predominantly obstructive or mixed, is much more common than central sleep apnoea (CSA). Sleep disordered breathing (SDB) is a term commonly used to encompass both these and other related conditions, some of which are mentioned below. The term obstructive sleep apnoea syndrome (OSAS) is applied when OSA is accompanied by daytime sequelae such as excessive daytime sleepiness.

As there is a continuum of possible RDIs from trivial to severe, defining the presence of clinically significant sleep apnoea is somewhat arbitrary. It is generally agreed that the RDI should exceed five to be considered significant, with some advocating an RDI of

5 ten or more. It has been suggested that an RDI of 5 to 15 represents mild sleep apnoea, 15 to 30 moderate and greater than 30 severe (AASM 1999). However, the magnitude of associated symptoms and hypoxaemia also need to be considered when severity is determined (Lugaresi 1983). These definitions of severity have never been considered from the point of view of their relevance to surgery and anaesthesia.

Hoffstein and Szalai (Hoffstein 1993) found that even with the inclusion of a "clinical impression" by the examining sleep physician, clinical features could not reliably predict the presence or otherwise of OSA. Many patients, brought along to clinics by concerned bed partners who have witnessed apnoeas, deny symptoms. Conversely, some patients exhibiting all the daytime features of OSA have few apnoeas or hypopnoeas. Some of these habitual snorers have been found to have recurrent arousals from sleep resulting from increases in upper airway resistance not sufficient to cause apnoeas or hypopnoeas as usually defined, a condition sometimes referred to as upper airway resistance syndrome (Guilleminault 1993). Complicating matters still further is the variation in daytime sequelae, a few patients, and women in particular, presenting not with excessive daytime sleepiness but with other symptoms such as anxiety (Ambrogetti 1991). Nor do the above criteria always apply satisfactorily to children (Rosen 1992). This is of relevance to anaesthetists, less experienced in this area than sleep physicians, who are nevertheless often required to assess the likelihood of OSA in their patients preoperatively.

Pathophysiology

Obstructive Sleep Apnoea

A narrow, floppy upper airway provides the pathophysiological basis for OSA (Kuna 1991; Douglas 1994; Deegan 1995). This may have a congenital or acquired origin (Table 2.1). Usually such an airway does not cause problems during wakefulness. However, with sleep the associated loss of skeletal muscle tone makes the upper airway still narrower and floppier, particularly during rapid eye movement (REM) sleep when muscle relaxation is profound (van Lunteren 1991). This has two important consequences as gas is accelerated through it. First, the structures will tend to vibrate as

6 turbulent flow patterns are produced, with snoring the result. Second, the pharynx will tend to collapse due to the Bernoulli effect, with resultant partial or complete obstruction (Gavriely 1993). A variety of models have been suggested to describe this behaviour of the upper airway (Ayappa 2003) but the end result is the same. Obstruction will persist until sleep is interrupted and muscle tone is restored. Usually these interruptions are momentary arousals and the sufferer is unaware of them. Occasionally, the obstructive event will result in an awakening, and the sufferer may complain of waking suddenly or with a snort or a snore. With arousal breathing is restored and after a few breaths deeper sleep will again resume with recurrence of the problem as the muscles again relax (van Lunteren 1991). In the more severe cases of OSA, this cycle of apnoeas and arousals may occur hundreds of times a night. In the more subtle cases it may only occur in certain sleep stages and postures (Cartwright 1991; Pevernagie 1992; Oksenberg 1997) or after alcohol consumption (Taasan 1981b; Scrima 1982). The result of this constant sleep disruption is lethargy and somnolence during wakefulness.

A variety of pharmacological agents, some depressant and some stimulant, are known to act on the muscles of the upper airway (van Lunteren 1991). As many of the depressants are commonly used in the perioperative setting the potential for exacerbation of upper airway obstruction is significant.

7 Condition Examples Contribution Obesity, body fat distribution adult obesity, Prader-Willi complex and ill-defined (Grunstein 1994; Harris 1996) syndrome Race/genetics (Redline 1992; ?anatomical similarity Ancoli-Israel 1995) Age (Bixler 1998) ?tissue laxity Male gender (Young 1993c) unclear Alcohol (Taasan 1981b), muscle relaxation, depressed sedatives, analgesics, arousal anaesthetics (Robinson 1989) Smoking ?chronic nasal congestion, pharyngeal oedema Nasal obstruction (Miljeteig septal deviation, chronic nasal increased pharyngeal negative 1992) congestion pressure Pharyngeal obstruction tonsillar & adenoidal increased pharyngeal negative (Helfaer 1994) hypertrophy pressure Cranio-facial abnormality Down's, Pierre-Robin, Treacher- mid-face hypoplasia, (Stokes 1983; Roa 1984; Collins, Apert's, Crouzon's, macroglossia or micrognathia Colmenero 1991; Grunstein Beckwith-Wiedemann, 1991b; Marcus 1991; Tirosh achondroplasia, acromegaly, 1992) fragile-X Laryngeal obstruction laryngomalacia, tracheomalacia laryngeal collapse Endocrine/Metabolic hypothyroidism, androgen upper airway infiltration or (Grunstein 1988) therapy, Cushing's myopathy, obesity Neuromuscular disorders stroke, cerebral palsy, head disordered pharyngeal (Guilleminault 1977; injury, Shy-Drager, neuromuscular function Guilleminault 1978; Flavell poliomyelitis, myotonic 1992; Short 1992; Mohsenin dystrophy, dysautonomia, 1995; Hsu 1998) tetraplegia Connective tissue disorders Marfan's abnormal upper airway (Cistulli 1993) connective tissue Storage diseases (Semenza mucopolysaccharidoses macroglossia 1988) Chronic renal failure (Kimmel unclear 1989; Langevin 1993)

TABLE 2.1. Known and suspected predisposing conditions for obstructive sleep apnoea.

8 Central Sleep Apnoea

Inadequate breathing during sleep due to diminished or absent respiratory effort (central sleep apnoea) may occur in association with disorders of ventilatory control or neuromuscular function or where the respiratory musculature is excessively loaded (Table 2.2). Patients with such conditions have diminished ventilatory capacity that may be sufficient for their needs during wakefulness but results in hypoventilation during sleep when the drive to ventilation is reduced and the compensatory mechanisms fail. As with obstructive sleep apnoea, the pharmacological effects of many drugs used perioperatively can contribute to the development or exacerbation of central apnoeic and hypopnoeic events (Robinson 1989).

Condition Examples Contribution Neuromuscular disorders poliomyelitis, respiratory muscle weakness (Gay 1991; Bradley 1992b; amyotrophic lateral Hsu 1998) sclerosis, muscular dystrophy Excessive respiratory load obesity, airways disease, excessive elastic, resistive or (Grunstein 1994) kyphoscoliosis threshold loading of muscles Disordered peripheral cardiac failure, bilateral delay or failure of chemosensitivity (Connolly carotid body excision ventilatory feedback from 1995; Solin 1998) peripheral chemoreceptors Disordered central stroke, head injury impaired ventilatory drive ventilatory control (Bradley 1992b) Endocrine/Metabolic acromegaly ?increased growth hormone, (Grunstein 1991b) insulin-like growth factor 1

TABLE 2.2. Known and suspected predisposing conditions for central sleep apnoea.

9 Symptoms and Signs

The key symptoms present in most cases of sleep apnoea are heavy snoring, occasional sudden awakenings that may be associated with momentary choking, apnoeas witnessed by a bed partner and excessive daytime sleepiness (McNamara 1993). Obtaining a history from the bed partner can be vital in eliciting several of these symptoms. Apart from these cardinal features, other recognised symptoms are listed in Table 2.3 and the signs in Table 2.4. While the symptoms lack specificity, in many cases a reasonably confident diagnosis may be made on history alone.

Adults Children (Hanning 1989; McNamara 1993; Munoz 1998) (Helfaer 1994; Rosen 1996)

Heavy snoring Snoring Excessive daytime sleepiness Restless sleeping Witnessed apnoeas Somnolence Sudden awakenings with "choking" Aggression/behavioural problems Accidents related to sleepiness Hyperactivity Poor memory/concentration Odd sleeping postures Delirium Frequent coughs/colds Gastro-oesophageal reflux Retarded growth Mood/personality changes Nocturnal sweating Restlessness during sleep Nocturia Enuresis (uncommon) Dry mouth on awakening Nocturnal or morning headache Impotence Nocturnal epilepsy

TABLE 2.3. Symptoms associated with sleep apnoea

10 Oedematous soft palate or uvula, enlarged tonsils Long soft palate and uvula Decreased oropharyngeal dimensions Nasal obstruction Maxillary hypoplasia Retrognathia Central adiposity/increased neck circumference Hypertension and other cardiovascular consequences Conditions/syndromes (listed in Tables 2.1 & 2.2) associated with sleep apnoea

TABLE 2.4. Signs associated with sleep apnoea (Hanning 1989; Rosen 1996)

Investigation

The gold standard investigation for sleep apnoea is full overnight polysomnography (PSG) from which the type and severity of any apnoea may be determined (Carskadon 1989). Electroencephalogram (EEG), electro-oculogram and sub-mental electromyogram (EMG) are recorded for the purpose of staging sleep. Respiration is assessed by monitoring oro-nasal airflow (pressure transducer or thermistor), respiratory effort (inductance or impedance pneumography to monitor thoraco-abdominal motion and/or diaphragmatic EMG) and pulse oximetry. Additionally, it is usual to monitor body position, sound and electrocardiogram. Videotape to record body movements and transcutaneous CO2 are also used in selected cases. Subsets of these may be used for screening purposes, an example being the MESAM 4 system using oximetry, heart rate, snoring and position (Stoohs 1992; Ferber 1994).

Originally, the PSG data were printed out in real time using a polygraph. This method has now largely been replaced by digital storage techniques using a variety of commercially available software packages. Either way, the records are examined in thirty second "epochs" and the sleep stage for each epoch is determined using the criteria of Rechtschaffen and Kales (Rechtschaffen 1968). Respiratory events are scored

11 using the definitions listed above and the total number of events, their duration and the degree of desaturation summarised for the whole night and for specific sleep stages.

Nasopharyngoscopy or upper airway imaging (lateral cephalometry or computed tomography) may be performed to guide treatment, for example whether or not surgery will be of any benefit (Fleetham 1992).

The results of these investigations are relevant for anaesthetists as they may indicate the likelihood of difficulty with intubation or airway maintenance and the potential for problems with ventilation in the postoperative period.

Sequelae

Many of the sequelae of sleep apnoea have limited relevance to anaesthesia. A variety of confounding factors make this issue complex (Koskenvuo 1994). The symptomatic accompaniments have already been listed (Table 2.3) and other sequelae are summarised in Table 2.5. The relative risks of these sequelae are difficult to assess due to the broad spectrum in the severity of the disorder. Recent results from the Sleep Heart Health Study suggest the odds ratio for various chronic cardiovascular sequelae of OSA range from approximately 1.5 to 3 (Malhotra 2002).

While yet unproven, several of the acute changes associated with apnoeic or hypopnoeic episodes have the potential to influence perioperative progress. Possible complications include cardiac arrythmias, myocardial ischaemia, cerebrovascular insufficiency, intracranial hypertension, mental dysfunction and poor wound healing (Jennum 1989; Rosenberg 1989; Rosenberg 1990; Reeder 1991b; Gill 1992; Goldman 1993; Rosenberg 1993; Galatius-Jensen 1994; Gogenur 2004).

Chronically, if the sleep apnoea is severe enough, respiratory and right heart failure may develop (Bradley 1992a) as the result of persistent, severe nocturnal hypoxaemia and hypercapnia, further increasing the risk for anaesthesia and surgery. The role of sleep apnoea in the development of polycythaemia is controversial (Carlson 1992; Hoffstein 1994)

12 Neuropsychological (Kudrow 1984; Sleepiness, impaired memory and cognition, Kales 1985; Jennum 1989) decreased vigilance, increased accident risk, anxiety and depression, chronic headache, intracranial hypertension Cardiovascular (Koskenvuo 1987; Hypertension, ischaemic heart disease, Hung 1990; Bradley 1992a; Wilcox cerebrovascular disease, right heart failure 1993; Carlson 1994; Lavie 1995; Hu 1999) Pulmonary (Sajkov 1994; Laks Hypoxaemia, hypercapnia, pulmonary 1995) hypertension Endocrine (Grunstein 1989; Brooks Decreased growth hormone and testosterone 1994; Rosen 1996) levels, diabetic instability Gastrointestinal (Kerr 1992) Gastro-oesophageal reflux

TABLE 2.5. Potential sequelae of sleep apnoea

. Treatment

Obstructive Sleep Apnoea

In mild cases, conservative measures alone may lead to a satisfactory improvement. These measures include weight loss, reduction of alcohol or sedative consumption, sleeping on the side, and cessation of smoking. In most cases, however, these form an adjunct to more aggressive therapy, either because they are insufficient by themselves or because they prove difficult to achieve (Grunstein 1991a).

Trials of drugs that alter sleep architecture or upper airway muscle tone and electrical stimulation of the upper airway muscles during sleep have generally proved disappointing (Grunstein 1991a). Nevertheless, this aspect of sleep apnoea research has great relevance for the postoperative setting, as a number of commonly used drugs have the potential to influence upper airway function, including agents such as antiemetics that act on the serotonergic system (Horner 2001).

13 Introduced by Sullivan in 1981 (Sullivan 1981), nasal continuous positive airway pressure (nCPAP) remains the treatment of choice for OSA of at least moderate severity (Grunstein 1995). This treatment is highly effective and prevents obstructive events by pneumatically splinting the upper airway (Popper 1986). Compliance, however, is variable and in milder forms of sleep apnoea, where daytime symptoms are mild, it is often not well accepted by patients, being moderately intrusive (Grunstein 1995).

In severe OSA, particularly when associated with morbid obesity or other coexisting disease such as chronic airflow limitation, the patient may present in respiratory and right heart failure. In addition to the obstructive apnoeas, central sleep hypoventilation can be present in such cases, particularly during REM sleep. If so, initial control may be achieved with non-invasive bi-level ventilatory assistance (Kryger 1992; Hill 1993). This involves the delivery of intermittent positive pressure ventilation (IPPV) with positive end-expiratory pressure via a nasal or face mask using BiPAP (Bilevel positive airway pressure – a trademarked name) or similar device. Once control of sleep hypoventilation and respiratory failure have been achieved it is often possible to convert to CPAP, a cheaper therapy, if the predominant problem has been OSA (Piper 1994). This also has implications for perioperative management. Patients with severe but untreated OSA may find themselves in a situation of respiratory decompensation after surgery. An appropriate therapy in such cases may be bi-level non-invasive ventilatory support, at least in the short term.

The use of oral appliances that reposition the mandible (forwards), increasing the pharyngeal dimensions, is becoming more common for the treatment of snoring and milder forms of OSA (ASDA 1995; O'Sullivan 1995). Potential complications of these devices such as temporomandibular joint dysfunction have not yet been widely investigated (Grunstein 1995), but there is now evidence that they are associated with dental side-effects which, while generally mild and temporary in nature, may necessitate treatment cessation in some individuals (Pantin 1999).

Palatal surgery is a reasonable treatment alternative for habitual snoring but a less certain treatment for OSA (Grunstein 1991a; ASDA 1994; ASDA 1996; Sher 1996). Surgical correction of nasal obstruction is important but, of itself, does not usually result

14 in resolution of sleep apnoea. Surgical removal of obstructing lesions in the pharynx can be definitive and tonsillectomy/adenoidectomy is a front-line treatment of obstructive sleep apnoea in childhood (Helfaer 1994). Maxillofacial surgery may be necessary where craniofacial abnormalities exist that are associated with OSA (Conradt 1998), but its use is limited (Cistulli 1996). Tracheostomy, the main method of treating sleep apnoea prior to the development of CPAP, is now only indicated in life- threatening OSA when non-invasive forms of respiratory support are not tolerated.

Central Sleep Apnoea

Patients with sleep-related hypoventilation due to neuromuscular disease or one of the other causes listed in table 2.2 may respond to treatment with CPAP or respiratory stimulants. More usually, if sufficiently severe, non-invasive ventilatory assistance is required and IPPV via nasal or face mask is the method of choice (Grunstein 1991a). Patients requiring IPPV for greater than 12 hours a day and those with inadequate airway patency or protection may need a tracheostomy. External negative pressure ventilation, such as with a cuirass, may exacerbate or induce upper airway obstruction (Hill 1992) and the cumbersome nature of this treatment has rendered it largely obsolete.

SLEEP AND ANAESTHESIA – Their Nature and Effects on Ventilation

Sleep

Unlike anaesthesia, sleep is a state of rousable unconsciousness. While much is known about the electro-chemical factors influencing sleep onset and the sleep-wakefulness cycle, the exact function of sleep remains unclear, apart from the fact that it is essential for wellbeing.

15 Electrophysiology of Sleep

The EEG was first used to investigate and characterise sleep by Loomis and colleagues in the 1930's (Loomis 1937). It was not until 20 years later that Aserinsky and Kleitman recognised the association between eye movement and the phases of sleep (Aserinsky 1953). This soon led to the definition of sleep stages based on EEG, eye movements and muscle tone, more or less as we now know them (non-rapid eye movement (NREM) stages 1 through 4 and REM), by Dement and Kleitman in 1957 (Dement 1957). Rechtschaffen and Kales subsequently refined these definitions into guidelines that remain the international standard after 30 years of use (Rechtschaffen 1968).

A single pair of EEG leads may be used to stage sleep. Typically, one electrode is placed adjacent to the vertex (C3 or C4) and another over the contralateral mastoid process (A2 or A1). The differential input from these is referred to a third electrode, often over the other mastoid. An occipital electrode may also be used. For the eye movements, another pair of electrodes is used, one above the outer canthus of one eye, the other below the outer canthus of the other eye. Both are referred to one of the mastoid electrodes. This results in out-of-phase deflections for both horizontal and vertical eye movements, allowing differentiation from artefacts, which are usually in- phase. A third pair of electrodes is placed under the chin to monitor the EMG (Carskadon 1989).

Relaxed wakefulness is characterised by sinusoidal alpha (8 - 12 Hz) and low voltage, mixed frequency activity on the EEG, accompanied by eye movements, blinking, and high submental EMG tone. With sleep onset (stage 1) there is muscle relaxation, slow horizontal rolling of the eyes and a marked reduction in the amount of alpha activity, leaving mainly the low voltage, mixed frequency component (Carskadon 1989).

Stage 2 may be associated with a further reduction in the EMG, but it is particularly defined by the appearance, superimposed upon the stage one type EEG background, of sleep spindles (short bursts of 12 - 14 Hz activity similar to waking alpha) and K- complexes (a sharp negative wave immediately followed by a broader, high voltage positive component). K-complexes may be either spontaneous or a response to an

16 external stimulus, and are frequently closely associated temporally with spindles (Carskadon 1989).

Stages 3 and 4, together referred to as slow wave sleep (SWS), are characterised by high voltage delta (1 - 4 Hz) activity (hence its other less common name, delta sleep). If the epoch has between 20% and 50% of its record consisting of slow waves then it is scored as stage 3. Epochs containing more than 50% SWS are classified as stage 4 (Carskadon 1989).

Stage REM has an EEG pattern similar to stage 1. It is, however, clearly defined by the presence of episodic rapid eye movements, very low EMG amplitude and a variety of other physiological changes as described below (Carskadon 1989).

It has been assumed to date that these electrophysiological definitions of sleep and its stages are appropriate in perioperative studies. From one of the earliest investigations reported (Aurell 1985) through to more recent work (Rahman 2001; Wu 2003) there have been observations that sleep state as defined by its electrophysiological markers may not coincide with behavioural indications of sleep. This ambiguity warrants further investigation.

The Typical Sleep Pattern

Of the few previous depictions in the anaesthetic or surgical literature of the normal human sleep pattern (Kavey 1979b; Rosenberg 1994; Rosenberg-Adamsen 1996a), at least one is quite inaccurate (Rosenberg-Adamsen 1996a) and none mention the changes in this pattern with age. Knowledge of the typical pattern is necessary before assessment of perioperative changes can be made. As a number of assumptions and speculations have been drawn from relatively few observations of perioperative sleep such data is all the more important.

In young adults (Dement 1957; Williams 1964; Kales 1967) a brief initial period of stage 1 is usually followed by stages 2, 3 and 4 in that order. The SWS component normally predominates this first NREM period and after about 70 minutes of sleep the first REM stage occurs, preceded by a period of stage 2. This cycle is repeated,

17 depending upon the total sleep time, up to six times but the later cycles usually lack stage 4. The REM periods tend to lengthen as sleep progresses while the cycle length, averaging 70 - 90 minutes, shortens as the NREM component decreases more than the increase in REM. Stage 2 is the predominant stage for the total period of sleep, usually making up about 50%. Stage 1 totals about 5%, SWS about 20% and REM about 25%. The graphical depiction of sleep in stages is known as the hypnogram and an example from a young adult male is shown in figure 2.1.

FIGURE 2.1. Actual hypnogram of a young adult male medical student – a graphical depiction of the sleep stages during one night’s sleep (recorded and manually scored with a commercial sleep monitoring system – Compumedics, Melbourne, Australia). REM = rapid eye movement sleep; AWK = awake or movement time. The final REM period ended prematurely as a consequence of study termination.

The Influence of Age

The changes in sleep pattern with age are profound (Feinberg 1974) and have the potential to heavily influence interpretation of studies into perioperative sleep. Total sleep time shows a precipitous decline during adolescence from an average of 10 hours/day or more at age six to about 7.5 hours in early adulthood. There is then a plateau until old age when a further but less dramatic decline occurs. The proportion of time spent in bed but awake remains at a few percent until mid-life whereafter it rapidly increases to about 20% or more in old age. The number of arousals per night increases

18 more linearly. As a result of these changes ageing is associated with more frequent and prolonged interruptions to sleep.

REM sleep decreases from more than 50% of total sleep time in neonates (Stern 1969) to about 30% in later childhood before a plateau of about 25% for most of adulthood and a further decline to about 20% late in life. Stage 4 sleep, on the other hand, displays no plateau, its total amount declining sharply during adolescence, then halving again between the ages of 20 and 60 years. This decline in stage 4, about half of which normally occurs in the first sleep cycle on any given night, results in a shorter first cycle and hence a reduction in REM latency, the time to first REM onset. This first REM period also becomes longer in old age, leading to a more even distribution of REM throughout the sleep cycles, the number of which is about the only sleep variable to remain constant with age.

Other Physiological Variables in Sleep

A complete account of the gamut of physiological changes during sleep is outside the scope of this thesis. Each stage of sleep has a fairly distinct pattern of physiological phenomena (Dempsey 1991; Guilleminault 1991) and the complexity is such that any attempt at classifying sleep stages according to some arbitrary measure of "depth", as is commonly done, amounts to gross oversimplification. Nevertheless, in order to examine the impact of sleep in the perioperative period knowledge of some of these changes is required.

Skeletal muscle function. All skeletal musculature, be it postural, chest or abdominal wall, diaphragm or upper airway, is subject to wake/sleep state-related activity changes (Dempsey 1991). There are, however, marked differences between the muscle groups. The tone of postural muscles, compared with wakefulness, is reduced somewhat in NREM and almost completely abolished in REM. This is a consequence of hyperpolarisation of alpha motor neurones (Seigel 1989) which is most marked during the transition from NREM to REM sleep and during bursts of eye movement activity, commonly referred to as phasic REM sleep (McGinty 1985). Despite this, phasic REM is characterised by rapid, random fluctuations in motor neurone membrane potential, hence varying levels of excitation and inhibition resulting in the eye movements and

19 twitches of limbs and facial muscles. This occurs against the background active inhibition of tonic REM sleep. In contrast to non-respiratory muscles, the inspiratory activity of the chest wall, accessory and diaphragm muscles is preserved in NREM, as is the expiratory activity of the abdominal wall (Tabachnik 1981; Skatrud 1985; Skatrud 1988; Henke 1991). During REM, the tonic and phasic activity of all of these respiratory muscles, except the diaphragm, is greatly reduced. The diaphragm's phasic activity is preserved, albeit on a background of reduced tone (Tusiewicz 1977b). This explains the profound hypoventilation seen when patients with diaphragmatic weakness enter REM sleep (Skatrud 1980). The upper airway musculature follows the same pattern as the postural muscles (Sauerland 1976), increasing the tendency to collapse (Hudgel 1984b), especially during REM, leading to the commonly held view that OSA occurs more in that sleep stage. However, as some muscles are constrictors rather than dilators, and as the state-related changes differ from muscle to muscle and from individual to individual (van Lunteren 1991), this tendency is not universal. Moreover, in some individuals respiratory pump muscle function appears to be affected more than that of the upper airway during REM, leading, in fact, to a higher frequency of OSA in sleep stages 1 and 2 (Series 2002; Ayappa 2003). Possibly for the same reasons, OSA appears to be much less common in SWS than other sleep stages. As SWS is usually considered together with stages 1 and 2 as part of NREM sleep, thus reducing the apparent frequency of obstructive events in NREM overall, this may have contributed to the widespread belief that OSA is most common during REM.

Ventilation-perfusion relationship. Functional residual capacity (FRC) is reduced during sleep, presumably as a consequence of sleep related changes in respiratory muscle tone together with gravitational effects of the supine position on the lung and abdominal contents (Hudgel 1984a). This results in atelectasis in the dependent regions of the lung with shunt, particularly in the case of patients with obesity and chronic lung disease (Ballard 1989).

Load Compensation. The application of resistive or elastic respiratory loads during wakefulness leads to a rapid increase in the motor output to the respiratory musculature as well as an increase in the duration of inspiration (Iber 1982; Wilson 1984; Wiegand 1988). In addition to this, increased negative pharyngeal pressure resulting, for example, from increased upper airway resistance leads to an increase in the neural output to upper

20 airway dilator muscles (Mathew 1982; van Lunteren 1984). Sleep not only imposes both resistive and elastic loads on the respiratory muscles, via upper airway narrowing and decreasing FRC respectively, but it also compromises the compensatory mechanisms that cope with these changes. During NREM sleep, load compensation occurs but is slow and incomplete (Wilson 1984; Wiegand 1988; Badr 1990) with increased reliance on chemical drive which itself may be depressed (see below), the end result being a degree of hypoxaemia and CO2 retention. The situation in REM is worse still, with a further increase in loading and a simultaneous failure of intercostal, accessory, upper airway dilator and expiratory muscles to assist in the necessary compensation (Dempsey 1991). The coexistence of either neurological or mechanical respiratory disease, already challenging the compensatory mechanisms, further increases the tendency to hypoventilation.

Ventilatory control. Wakefulness, including transient arousal, has an important stimulatory effect on ventilation (Dempsey 1991; Khoo 1998). While it appears that chemosensitivity is important for maintaining ventilation during sleep, as indicated by the increased sleep-related hypoventilation seen in patients with carotid body denervation (Connolly 1995), the effects of sleep on chemoreception are far more complex and difficult to define. Standard tests of acute ventilatory responses have demonstrated varying degrees of inhibition, particularly of the response to combined hypoxia and hypercapnia, but these may overestimate the reduction in chemosensitivity as other factors such as increased upper airway resistance, impaired load compensation and changes in cerebral blood flow need to be considered (Dempsey 1991). On the other hand, sleep does unmask the "apnoeic threshold", not normally seen in wakefulness (Dempsey 1986). Thus, in sleep, apnoeas or hypopnoeas can be produced by lowering the PCO2, as may occur during hypoxic hyperventilation (Berssenbrugge 1983). This reduction in ventilation may then result in an overshoot into hypoxic hypercapnia again, leading itself to hyperventilation and consequently a cycle of hypoxia-induced periodic breathing with large swings in oxygen saturation – a variant of central sleep apnoea. Ironically, sufficiently large increases in upper airway resistance may be one factor preventing periodic breathing in some subjects by limiting the hyperventilation (Dempsey 1991). Despite the lack of clarity with respect to sleep effects on chemosensation, it appears that there is a reduction in output from medullary respiratory

21 neurones, particularly during NREM, whereas in REM the output from these neurones tends to be related to the variability in breathing pattern (Dempsey 1991). REM, however, is associated with a depression of the arousal responses to hypoxia and hypercapnia, leading to a tendency for apnoeas to be longer and desaturations more severe in that sleep stage.

Autonomic nervous system. Stage-related autonomic changes result in an overall reduction in heart rate during sleep, the lowest rate occurring in SWS, but with considerable variability during REM such that the average REM heart rate is higher than in NREM. Blood pressure follows a similar pattern to heart rate, as does respiration, but marked differences in regional vascular resistance result in a reduction in peripheral blood flow, while coronary perfusion is better maintained. In contrast, REM sleep is associated with a substantial increase in cerebral blood flow, in keeping with a demonstrable increase in cerebral neuronal activity (Guilleminault 1991).

Metabolism and thermoregulation. Sleep is associated with energy conservation (Guilleminault 1991). Increased metabolic expenditure results in compensatory increases in sleep duration while several days of fasting results in increased SWS, the sleep stage with the lowest metabolic rate (Guilleminault 1991). Thermoregulation is even more closely knit to the sleep cycle, the cycle itself taking cues from circadian temperature changes, both within and without the body (Glotzbach 1989). Body temperature, conversely, is affected by sleep stage, in particular the REM inactivation of thermoregulatory mechanisms such as shivering. REM sleep, perhaps as a consequence, becomes less likely as ambient temperature becomes less acceptable (warmer or cooler) (Karacan 1978; Haskell 1981; Muzet 1983; Glotzbach 1989; Guilleminault 1991). The influences of ambient temperature, fever and fasting on the sleep cycle have significant implications for perioperative sleep research.

Anaesthesia

In contrast to sleep, anaesthesia is a state of reversible, drug-induced unrousable unconsciousness.

22 The electrophysiological nature of anaesthesia is an area of intense ongoing investigation, particularly now with a number of devices allegedly able to monitor "depth" of anaesthesia available (Rampil 1998). It is, however, a very complex issue as different anaesthetic agents have different effects on the EEG (Black 1994) so that no unitary pattern indicating anaesthetic "depth" exists. It is therefore very difficult to make any electrophysiologic comparisons between sleep and anaesthesia, although attempts have been made (Sleigh 1999; Nieuwenhuijs 2002), and such comparisons are probably of limited relevance as the two states are quite distinct. With few exceptions, anaesthetic and sedative drugs produce a dose dependent depression not only of consciousness, but also of most other vital functions, including all those related to respiration. Apart from abolition of the stimulatory effects of wakefulness these include depression of hypoxic and hypercapnic responses (Sollevi 1995), load compensation reflexes (Moote 1986) and the arousal responses that normally protect against asphyxia. As with sleep there is depression of skeletal muscle tone with reduction in FRC, predisposing to atelectasis, and upper airway muscle relaxation predisposing to obstruction (Hillman 2003). These effects are compounded by reduction in the phasic activity of intercostal and accessory respiratory muscles, increasing dependence on the diaphragm, and of the upper airway muscles during inspiration, further predisposing to obstruction as this activity acts to stiffen the airway as intraluminal pressure falls (Tusiewicz 1977a; Nishino 1984; Drummond 1989; Ochiai 1989; Drummond 1996).

The presence of a vigilant anaesthetist to monitor and maintain vital functions during anaesthesia protects the patient from these effects. However, drug induced sedation and post-anaesthesia drowsiness, where the borders between wakefulness, sleep and anaesthesia are less distinct and monitoring perhaps less rigorous, present potential danger to the patient with a sleep-related breathing disorder because of the depression of these responses.

Sleep in the postoperative period

There has only been one study examining the effects of general anaesthesia alone (with isoflurane) on subsequent sleep and it would appear that this effect is negligible (Moote 1988). Other studies imply that the type of anaesthesia is also not important (Kavey 1979b; Lehmkuhl 1987). The addition of a surgical insult changes things considerably

23 (Rosenberg 1995). Sleep architecture is disrupted to a degree which is generally proportional to the "magnitude" of the surgery as is the duration of the disruption, but it is important to note that there is considerable inter-individual variation and specific situations where the generalisation may not hold. The disturbance takes the form of reduced nocturnal total sleep time with a disproportionate reduction in REM and SWS (Kavey 1979b; Kavey 1983; Aurell 1985; Knill 1990; Rosenberg 1994). At some point during the first postoperative week there is an inconsistent tendency for these to rebound for one or two nights (Kavey 1979b; Kavey 1983; Knill 1990; Rosenberg 1994). Unfortunately, literature attention has focused on the two studies in which a REM rebound was predominantly demonstrated (Knill 1990; Rosenberg 1994), largely ignoring those in which it was not (Johns 1974; Ellis 1976; Orr 1977; Kavey 1979b; Aurell 1985; Lehmkuhl 1987) or in which it was not specifically addressed (Kavey 1983). This has led to a widely held but probably inaccurate view that REM rebound is common and significant, occurring on or about the third post-operative night and lasting for several nights thereafter (Cronin 1995; Benumof 2001; Gogenur 2004).

Evidence regarding the effect of surgery on total sleep time across the full 24 hour period in the postoperative setting is conflicting. The data of Ellis and Dudley (Ellis 1976) and that of Edéll-Gustafsson and colleagues (Edell-Gustafsson 1999) suggest that total sleep time over the full 24 hours is actually increased on the immediate postoperative day after major abdominal and cardiac surgery, whereas Aurell and Elmqvist (Aurell 1985), using continuous postoperative recording, reported no sleep at all for many of their subjects for 24 hours or more postoperatively.

The precise mechanism by which the surgical insult produces the sleep disruption is not completely clear. It is likely that pain plays a major role although recent work has questioned its importance (Cronin 2001). Other factors, which may be independent of the surgery and thus account for some of the variability, are neuroendocrine, metabolic and psychological responses, opioid analgesia, and environmental factors such as noise, light and nursing activity (Lehmkuhl 1987; Rosenberg 1995; Rosenberg-Adamsen 1996b).

The extent to which these changes in sleep architecture after surgery influence morbidity and mortality is currently unknown but there has been considerable

24 speculation based on indirect evidence (Rosenberg 1995; Rosenberg-Adamsen 1996a; Benumof 2001; Gogenur 2004). Patients with REM predominant apnoea, for example, might be expected to have an increase in the number and degree of desaturations over a night where REM rebound is occurring (Rosenberg 1994; Benumof 2001; Gogenur 2004) but this possibility has been inadequately investigated. Similarly, the occurrence of REM rebound has led to the suggestion that an associated late postoperative increase in nocturnal hypoxaemia could be contributing to mental confusion, wound breakdown, myocardial ischaemia and infarction, stroke and death (Rosenberg 1989; Rosenberg 1990; Reeder 1991b; Gill 1992; Rosenberg 1992; Goldman 1993; Rosenberg 1993; Gogenur 2004). While some relevant associations such as a circadian variation in unexpected postoperative deaths (Rosenberg 1992) have been demonstrated, direct evidence of causation is lacking.

While anaesthesia, of itself, may not affect subsequent sleep once the anaesthetic agents are eliminated, a considerable amount of research over the last 20 years has considered the effects of sub-anaesthetic concentrations on sleep and ventilatory control. Commencing with the work of Knill’s group in the 1970s (Knill 1978b), conflicting results have emerged regarding the effects on ventilatory responses to hypoxia and hypercapnia of subanaesthetic concentrations of potent inhalational agents, such as might be present in the minutes to hours after emergence (Gelb 1978), as well as some other drugs commonly used perioperatively. A number of studies demonstrated clear reductions in ventilatory responses (Knill 1978b; Knill 1979; Knill 1981; Knill 1982; Knill 1983; Knill 1984; Young 1993a; Dahan 1994a; Dahan 1994b; Nagyova 1994; Dahan 1995; Sarton 1996) while others did not (Knill 1978a; Dahan 1991; Temp 1992; Nagyova 1993; Dahan 1994c; Sjogren 1994; Temp 1994; Sjogren 1995; Sollevi 1995; Warren 2000). Apart from differences in methodology and pharmacology (Dahan 2003), one reason for this conflict appears to be the effect of sleep as van den Elsen and colleagues have shown that subjects stimulated and kept awake exhibited more or less normal ventilatory responses despite the presence of a potent inhalational agent whereas those allowed to sleep exhibited ventilatory depression (van den Elsen 1994; van den Elsen 1998). The mechanism by which sleep might contribute to the depression of ventilatory responses by sedative agents has not yet been investigated. Newer, rapidly eliminated volatile agents such as sevoflurane and desflurane are likely to limit the relevance of the effect of this class of drug on postoperative breathing.

25 ANAESTHESIA AND SLEEP DISORDERED BREATHING

Perioperative risks of sleep apnoea

Notwithstanding the relative paucity of specific information, knowledge of their pharmacological effects strongly suggests that anaesthetic, sedative and analgesic agents will aggravate or precipitate OSA by decreasing pharyngeal tone, depressing ventilatory responses to hypoxia and hypercapnia and inhibiting arousal responses to obstruction, hypoxia and hypercapnia. These latter effects frequently result in varying degrees of central respiratory depression.

A variety of surgical factors are also contributory. Surgery of the thorax and upper abdomen compromises ventilatory function (Knudsen 1970; Entwistle 1991), potentially compounding the effects of any OSA or centrally mediated hypoventilation that might occur postoperatively. Surgery involving the upper airway carries the risk of postoperative swelling that can worsen or precipitate obstruction (Kravath 1980; Burgess 1992; McColley 1992; Valnicek 1994). The same applies to situations where the nose is packed or a nasogastric tube is required, as the reduced lumen calibre will necessitate the generation of more negative pharyngeal pressures during inspiration thus promoting collapse (Taasan 1981a). They may also compromise therapy by making nasal CPAP difficult or unusable, and a full-face mask may be required in such circumstances. Patients are frequently nursed supine, sometimes for good reason, and as OSA is often position dependent (Isono 2002) this may contribute to increased risk of upper airway obstruction.

26 speculation about the level of this risk needs to be kept in perspective. Approximately 10% of the population of developed nations undergo surgery each year. Very large numbers of individuals with significant sleep apnoea undergo surgery without significant morbidity or mortality and without their SDB ever being recognised. This suggests the overall risk is actually quite low. On the other hand, the anecdotes about perioperative patients already mentioned (Reeder 1991a; VanDercar 1991; Ostermeier 1997; Cullen 2001) do suggest that SDB plays some role in adverse perioperative events. It is possible that a combination of factors is involved when morbidity occurs. One such additional factor may be obesity (Benumof 2001; Vieito Amor 2002) but this has been inadequately investigated.

To whom might these risks be important?

OSA is common and anaesthetists will often deal with sufferers. There are those who present with a diagnosis of the disorder. A proportion of this group will be on some sort of treatment, usually CPAP, but with a variable degree of compliance (Grunstein 1995). Some will bring their CPAP machines with them to hospital while others will arrive without their equipment, seemingly quite prepared to forego treatment for the duration of their hospital stay. It can reasonably be assumed that many in this latter group are poorly compliant at home. Another group will have been diagnosed with sleep apnoea but either declined treatment from the start or failed a trial of therapy (Meoli 2003).

There are still a large number of people who present for surgery with features suggestive of sleep apnoea but who have either never heard of the condition and/or have not sought diagnosis or treatment (Meoli 2003). There is also a final group of patients who have apnoea but either lack the overt features or have features that are missed perioperatively. Given the high prevalence of OSA in the community (Young 2002), there appears to be little doubt that the number of patients in these last two groups far outweighs the number already diagnosed.

Undiagnosed Sleep Apnoea

While there are considerable perioperative implications for the management of patients with known or suspected OSA, there has been very little mention in the literature about

27 the potential contribution of anaesthetists in the diagnosis and management of the disorder. Because anaesthetists are intimately involved in the management of patients’ airways during periods when muscle tone and central respiratory drive is depressed, they may be in an excellent position to screen patients for sleep apnoea. As this disorder can be associated with substantial morbidity it is not an issue we should ignore, neither clinically nor from a research perspective.

A clinical suspicion of sleep apnoea may first develop at the preoperative consultation, intraoperatively, if the patient proves difficult to intubate (Hiremath 1998) or has an airway that is difficult to maintain, or postoperatively with snoring and obstruction observed in the recovery room or beyond. These considerations are as important to children as they are to adults, with growth and development potentially compromised by untreated sleep apnoea (Rosen 1996). While difficult intubation has been shown to be strongly associated with OSA (Hiremath 1998), to date there has been no direct evidence that a tendency to upper airway obstruction under general anaesthesia similarly indicates the presence of the condition.

Diagnosed Sleep Apnoea

The perioperative management of patients with known (or preoperatively suspected) SDB has been the subject of a recent Clinical Practice Review Committee report (Meoli 2003). This is a complex issue with very little evidence apart from anecdotes and expert opinion to support any particular approach and a complete review of the available literature on the matter is beyond the scope of this thesis. As the role of postoperative analgesia during sleep may be particularly important this aspect of perioperative care warrants some examination.

The use of opioid analgesics in severe untreated SDB appears to be a common factor in most reported cases where complications, including a number of deaths, occurred (Reeder 1991a; VanDercar 1991; Ostermeier 1997; Cullen 2001; Parikh 2002). These reports also suggest that the route of opioid administration is largely irrelevant, serious complications occurring with intermittent intramuscular, patient-controlled intravenous and epidural opioids. Caution is required when interpreting these reports, however, as several involved what would appear to be excessive doses of opioid and some involved

28 inadequate monitoring or protocol failures. There has been no published scientific comparison between opioids and any other form of analgesia for sleep apnoea sufferers. Published opinion consistently suggests that non-sedative analgesia should be considered as an alternative or to help limit the use of opioids. Suggestions include local or regional techniques (bearing in mind the potential for exacerbation of apnoea with neuraxial opioids), non-steroidal analgesics, ketamine and tramadol (Rosenberg- Adamsen 1996a; McArdle 1999; Aspinall 2001). The deliberate avoidance of opioids in OSA, however, has itself resulted in reports of serious complications such as devastating compartment syndromes resulting from epidural motor blockade (Kontrobarsky 1997). The potential benefits of some alternatives must also be weighed against their risks, renal impairment with non-steroidal analgesics for example, in significant subgroups of sleep apnoea sufferers such as the elderly (McArdle 1999).

The use of continuous positive airway pressure (CPAP) or alternative therapy for SDB such as mandibular advancement devices (Benumof 2001; Loadsman 2001), artificial airways (Young 1993b) and sleeping laterally (Isono 2002) may alleviate the risk of sedative analgesia. The utility of CPAP for the prevention of apnoea when it is used appropriately in this setting is supported by case reports (Reeder 1991a; Rennotte 1995; Mehta 2000). There is at least one report of a “near-miss” asphyxia incident with opioid analgesics in a patient already on CPAP preoperatively who had his CPAP discontinued in the postoperative period (Parikh 2002). The effective use of CPAP in the setting of acute pain management may require a higher level of supervision than that available in the general surgical ward. Most reports of the successful use of postoperative CPAP in severe apnoeics also utilised extended periods of high-dependency nursing (Reeder 1991a; Rennotte 1995; Mehta 2000). Concerns about the risk of CPAP causing gastric distension and anastomotic leaks after upper gastrointestinal surgery appear to be unfounded and CPAP should not be avoided postoperatively for this reason (Huerta 2002).

29 SUMMARY

Sleep disordered breathing is a common problem affecting all age groups, particularly in association with certain other medical conditions and syndromes. The pathological consequences of the disorder may be severe, with significant implications for the perioperative management of sufferers.

Our understanding of the implications of sleep disturbance and sleep disordered breathing for perioperative morbidity and mortality is limited. Several observations have led to considerable speculation in the literature but evidence of a causal relationship is still almost non-existent. REM sleep rebound particularly has been stated as a likely cause of late postoperative morbidity but this speculation requires further assessment. It is based on assumptions that REM sleep results in increased sleep- disordered breathing in the majority of individuals and this may not be correct. Most of the associations demonstrated, such as the circadian variation in postoperative mortality, have also not been adequately tested.

The assumption that standard electrophysiological markers of sleep can be reliably used in perioperative investigations has similarly not been tested.

Anaesthetists are ideally placed to screen large numbers of people for sleep disordered breathing, a source of considerable community morbidity. Difficult endotracheal intubation is known to be associated with obstructive sleep apnoea but, while other observation during and after surgery may also suggest the presence of sleep-disordered breathing, direct evidence is lacking.

30 CHAPTER 3

IS OSA A REM-PREDOMINANT PHENOMENON?

The tone of the upper airway musculature is reduced by sleep, particularly during the rapid eye movement stage (REM). It is reasonably assumed from this and commonly believed that obstructive sleep apnoea (OSA) is a phenomenon occurring most frequently during REM in the majority of sufferers. Authors of anaesthesia literature on the subject of sleep apnoea have, to date, universally upheld this assumption (Hanning 1989; Connolly 1991; Rosenberg-Adamsen 1996a; Warwick 1998; Benumof 2001; Hillman 2003). Consequently, it is also widely assumed that REM rebound in the late post-operative period, following the REM suppression shown to occur after some types of surgery, is likely to worsen nocturnal episodic hypoxaemia (Jones 1990; Knill 1990; Cronin 1995; Rosenberg-Adamsen 1996a; Rosenberg-Adamsen 1997; Rosenberg 1999; Benumof 2001). The results of the first preoperative polysomnogram (PSG) performed as part this PhD research suggested the commonly held assumptions might not be completely accurate. This particular patient was found to have previously undiagnosed mild obstructive sleep apnoea, which was non-rapid eye movement (NREM) stage two predominant.

To determine if this was an isolated finding, I retrospectively reviewed the records of a consecutive series of 148 unselected patients I saw in consultation while working, as part of this research project, with the Sleep Disorders Consultative Service at the Royal Prince Alfred Medical Centre (Sydney, Australia).

31 Methods

Patient Selection

The medical records of 148 consecutive new patients seen by one physician (over a period of approximately two years) in a sleep disorders clinic were reviewed. The patients referred to the clinic are assigned randomly, according to time available on a first-come-first-served basis, to one of several physicians by the secretarial staff when the booking is made. The majority of patients are referred to the clinic by their general practitioners. None are self-referred. The clinic is a well-established unit associated with a major teaching hospital and there are a large number of other private clinics in the greater metropolitan area.

Eleven of the records were unavailable at the time of the review. Sixteen patients were still waiting for diagnostic sleep studies (DSS) and three had failed to attend.

Of the 118 patients who had already undergone DSS, twenty had an overall respiratory disturbance index (RDI – the number of apnoeas and hypopnoeas per hour of sleep) of less than ten per hour (considered “normal” in the clinic at the time). Ten had a non- OSA diagnosis (narcolepsy for example) and ten had OSA but the REM/NREM differential for the RDI and minimum blood oxygen saturation (SpO2min) were not reported. One patient had no REM sleep on the night of the study. Seventy-seven patients who had OSA and enough further information to identify any sleep stage predominance therefore remained for inclusion in the study.

The variables recorded were the RDI and the minimum blood oxygen saturation using pulse oximetry (SpO2min) for both REM and non-rapid eye movement (NREM) sleep, as well as any demographic data available. A number of the study reports lacked satisfactory body position data and this factor was not, therefore, included in the analysis. Four patients had a previous diagnosis of asthma. Spirometric data were recorded for all but eleven patients. Three had an FEV1/FVC of less than 70%, two of which had no known previous history of lung disease.

32 Diagnostic Sleep Study

The diagnostic sleep studies performed in each case were overnight polysomnographic studies including electro-encephalogram (C4/A1, O2/A1), electro-oculogram, submental and diaphragm electro-myogram, nasal airflow (thermistor), chest and abdominal strain gauge, pulse oximeter, electro-cardiogram, leg movement sensors and position sensor. All diagnostic studies were carried out in a variety of sleep laboratories independent of the clinic and analysed manually, according to standard criteria, by experienced sleep study technicians and further checked by one of several physicians specialising in sleep medicine. All episodes of oximetry artifact were manually rejected from analysis.

Definitions

OSA was diagnosed if the overall RDI was greater than ten per hour. While a small number of patients experienced some central events there were no patients in this study with enough central apnoea to classify them as having either mixed or predominantly central sleep apnoea (CSA).

Statistics

All values are presented as median (lower quartile – upper quartile) unless otherwise stated. The two-tailed Wilcoxon Signed Rank Test was used for statistical analysis.

Results

Respiratory Disturbance Index

The median values for NREM and REM RDI were, respectively, 27 (17 – 50) and 35 (19 – 58) per hour (P = 0.39). In thirty-seven of the seventy-seven studies (48%) the RDI in NREM was greater than in REM. Thirty-nine (51%) had a larger number during REM. One patient had identical values for both REM and NREM. The REM versus NREM value for each patient is plotted in Figure 3.1.

33 150

100

0 0 50 100 150 NREM-RDI

FIGURE 3.1. The respiratory disturbance index (RDI – events per hour) in rapid eye movement sleep (REM) plotted against the respiratory disturbance index in non-rapid eye movement sleep (NREM) for each individual patient in the study. Those above the line of identity represent patients with a higher RDI during REM and vice-versa.

Minimum Saturation

Two patients did not have the SpO2min recorded for one or both of REM or NREM leaving seventy-five for analysis. The median values for NREM and REM SpO2min respectively were 87 (83 – 90) and 87 (77 – 91) percent (P = 0.03). Thirty-seven patients (49%) had a lower SpO2min in REM while twenty-seven (36%) were lower in

34 NREM. Eleven patients (15%) had identical values in REM and NREM. The REM versus NREM value for each patient is plotted in Figure 3.2. It can readily be seen from the plot that those patients with the lowest overnight saturation have a tendency for this to occur during REM, while the milder apnoeics tend to have similar nadirs for both REM and NREM.

100

30 30 40 50 60 70 80 90 100

NREM SpO2min

FIGURE 3.2. The minimum blood oxygen saturation (percent) by pulse oximetry

(SpO2min) in rapid eye movement sleep (REM) plotted against the minimum saturation in non-rapid eye movement sleep (NREM) for each individual patient in the study. Those above the line of identity represent patients with a higher saturation nadir during REM and vice-versa.

35 Effect of Age and Body Habitus

In order to determine if age or body habitus had any significant influence on sleep stage predominance, the REM values for RDI were both divided by and subtracted from the corresponding NREM values and the results plotted against both the age of the individuals and their body mass index (BMI). Similarly the difference in SpO2min was plotted against age and BMI, a ratio in this case having far less clinical relevance than the difference. The plots are shown in Figures 3.3 (age) and 3.4 (BMI). Age had no discernible effect on sleep stage predominance in OSA with any of the RDI ratio (slope = -0.03 ± 0.02, r2 = 0.024, P = 0.18), RDI difference (slope = 0.27 ± 0.23, r2 = 0.019, P

2 = 0.24) or SpO2min (slope = 0.02 ± 0.09, r = 0.000, P = 0.87). With increasing BMI there would appear to be a trend towards REM predominance, perhaps more so for

2 SpO2min (slope = 0.47 ± 0.17, r = 0.093, P = 0.007) than RDI ratio (slope = 0.13 ± 0.04, r2 = 0.121, P = 0.002) or RDI difference (slope = -1.12 ± 0.43, r2 = 0.083, P = 0.02). From this data, it could be suggested that there is an increased likelihood of REM predominance in terms of minimum saturation particularly if the BMI is greater than 35 kg/m2.

36 15

-5 20 30 40 50 60 70 80 Age

FIGURE 3.3a. The ratio of rapid eye movement sleep (REM) and non-rapid eye movement sleep (NREM) respiratory disturbance indices (events per hour) for each patient plotted against his/her age in years. The solid line is the line of regression and broken lines the 95% confidence intervals thereof.

37 75

-25

-50

-75 20 30 40 50 60 70 80 Age

FIGURE 3.3b. The difference between the non-rapid eye movement sleep (NREM) and rapid eye movement sleep (REM) respiratory disturbance indices (events per hour) for each patient plotted against his/her age in years. The solid line is the line of regression and broken lines the 95% confidence intervals thereof.

38 60 50 40 30 20 10 0 -10 -20 20 30 40 50 60 70 80 Age

FIGURE 3.3c. The difference between the NREM and REM saturation nadirs (percent) for each patient plotted against his/her age in years. The solid line is the line of regression and broken lines the 95% confidence intervals thereof.

39 15

-5 20 25 30 35 40 45 50 BMI

FIGURE 3.4a. The ratio of rapid eye movement sleep (REM) and non-rapid eye movement sleep (NREM) respiratory disturbance indices (RDI – events per hour) for each patient plotted against his/her body mass index (BMI – kg/m2). The solid line is the line of regression and broken lines the 95% confidence intervals thereof.

40 75

-25

-50

-75 20 25 30 35 40 45 50 BMI

FIGURE 3.4b. The difference between the non-rapid eye movement sleep (NREM) and rapid eye movement sleep (REM) respiratory disturbance indices (RDI – events per hour) for each patient plotted against his/her body mass index (BMI – kg/m2). The solid line is the line of regression and broken lines the 95% confidence intervals thereof.

41 60 50 40 30 20 10 0 -10 -20 20 25 30 35 40 45 50 BMI

FIGURE 3.4c. The difference between the non-rapid eye movement sleep (NREM) and rapid eye movement sleep (REM) saturation nadirs (percent) for each patient plotted against his/her body mass index (BMI – kg/m2). The solid line is the line of regression and broken lines the 95% confidence intervals thereof.

Discussion

Almost half (49%) of the subjects in this study had RDIs in NREM that were higher or equal to their RDI in REM. This was reflected in the failure to find a statistically significant difference in the stage-specific RDIs between the groups. The saturation nadirs were statistically worse overall in REM, consistent with the commonly held view. More than half (51%) of the subjects, however, had nadirs in NREM that were equal to or worse than their nadirs in REM. Overall, therefore, the results are not

42 consistent with the view that, for most subjects, sleep apnoea is a REM-predominant phenomenon. Obesity appears to be associated with an increase in the likelihood of REM-predominance. Contrary to published opinion (Loadsman 2001), there is no evidence in this study that mild apnoea is associated with REM predominance particularly (figures 3.1 and 3.2).

It has been well demonstrated that a period of significant alteration in sleep architecture occurs after major non-cardiac (Ellis 1976; Kavey 1979b; Kavey 1983; Aurell 1985; Lehmkuhl 1987; Knill 1990; Rosenberg 1994) and cardiac (Johns 1974; Orr 1977) surgery. Most notably, REM, slow wave sleep (SWS – NREM sleep stages three and four) and total sleep time are variably reduced in the nights immediately following surgery and there is an inconsistent tendency for these to rebound for one or two nights thereafter (Kavey 1979b; Kavey 1983; Knill 1990; Rosenberg 1994). Unfortunately, literature attention has focused on the two studies in which a REM rebound was predominantly demonstrated (Knill 1990; Rosenberg 1994), largely ignoring those in which it was not (Johns 1974; Ellis 1976; Orr 1977; Kavey 1979b; Aurell 1985; Lehmkuhl 1987) or in which it was not specifically addressed (Kavey 1983). This has led to a widely held but probably inaccurate view that REM rebound is common and significant, occurring on or about the third post-operative night and lasting for several nights thereafter (Cronin 1995; Benumof 2001). It has been further extrapolated from this, based on the assumption that sleep apnoea is worse during REM, that REM rebound is likely to be associated with increased risk of late postoperative nocturnal hypoxaemia and consequently ischaemia-related complications.

There is indeed some direct evidence in support of this notion. Knill et al. studied a series of six obese patients for gastroplasty and found, despite a substantially higher number of apnoeas and hypopnoeas in NREM, that there were more severe desaturations during REM postoperatively (Knill 1987). However, the mean BMI of that series was more than 50 kg/m2 and, in view of the data herein concerning BMI, this potentially explains their finding of REM predominance for desaturation. Rosenberg et al. reported a similar finding for a series of ten patients (Rosenberg 1994). There are, however, a number of issues that may have affected their observations. The method of both monitoring (using a modified F3-A2 electroencephalogram channel on a “Somnolog” recording device) and scoring (using 6 instead of 30 second epochs) sleep

43 does not appear to have been standard. Airflow was not measured, leaving the oximetry data open to question with respect to artifact and the oximetry data also appears to have been assessed in a rather unconventional way (using 2 minute epochs for episodic hypoxaemic events). The data is tabulated in such a way as to make it very difficult to assess but it would appear that the greatest number of desaturations actually occurred during the night prior to the main REM rebound night, when there was actually less REM than preoperatively. It is also quite possible that their data were skewed by what appears to be a single patient with severe pre-existing sleep apnoea.

NREM/REM apnoea or desaturation differentials have been reported in only a few studies of OSA and these have usually been incidental findings. Results similar to those in the current study have been reported (Kudrow 1984; Chan 1989). Despite this, REM predominance of OSA still seems to be a widely held assumption, especially in the anaesthesia literature (Hanning 1989; Jones 1990; Connolly 1991; Rosenberg-Adamsen 1996a; Warwick 1998; Rosenberg 1999; Benumof 2001; Hillman 2003). The reasons for this are not entirely clear. The early history of sleep medicine, an area of study that has existed for less than forty years, focused naturally on the worst sufferers, obese individuals with the “Pickwick” syndrome. REM predominance of OSA may be a feature in these individuals and the views formed then, when the spectrum and prevalence of OSA was unknown, probably reflect this. While it is still possible to find generalisations about REM predominance of OSA in sleep literature (Horner 1996; Koenig 2001) it does not appear to be a universal assumption, as it almost certainly continues to be amongst anaesthetists.

One possible explanation for the discrepancy between the findings of this study and the common literary generalisations is a tendency for apnoeas to be longer during REM (George 1988; Jennum 1989), presumably as a consequence of lessened arousability (Berry 1997). In patients having a very large number of apnoeas any increase in the apnoea duration will actually limit the time available for other apnoeas to occur, possibly leading to both an increase in the degree of desaturation and to a reduction in the RDI during REM. This is consistent with the findings of this investigation (Figures 3.1 and 3.2). The overall results, however, are not explained by this mechanism, particularly when REM predominance has been thought to be a feature of milder forms of OSA (Loadsman 2001).

44 As a result of this investigation I formed the view that the most likely explanation for the finding lies in one or both of the following mechanisms. The pharynx contains both dilator and constrictor muscles and the state- and sleep stage-related relaxation of these muscles may have differential effects that vary from individual to individual (van Lunteren 1991). Alternatively and perhaps more likely, in some individuals not only the upper airway musculature but also the inspiratory pump muscles might have a disproportionate reduction in their contractility during REM. Thus a highly negative pharyngeal pressure, resulting in non-REM obstruction because of the maintenance of chest wall function, might not occur during REM when chest wall muscles cease to contribute to the pump. These possible explanations for the findings of this study are consistent with both the “Balance of Forces” (Isono 1997) and “Starling Resistor” (Gold 1996) models of the upper airway in OSA.

Some very recent contributions to the sleep medicine literature have lent support to these speculative explanations, although it would appear that the mechanisms put forward above are almost certainly too simplistic. Sériès, while discussing the OSA- induced morphological changes in the muscle and peri-muscular tissue of the upper airway and the complex interactions between them, implies that there is more to sleep- induced upper airway obstruction than a simple reduction in EMG activity. This supports the view that a markedly reduced muscle tone during REM will not always lead to greater incidence and degree of obstruction in that stage. Moreover, he goes on to discuss the effect of sleep stage on the coordination between upper airway and respiratory muscle activity, suggesting this mechanism as a reason why “obstructed breathing events mostly occur in stages I-II and REM where ventilation is unstable and are absent in slow wave sleep, which is characterised by a remarkable stability in tidal volume amplitude and breathing frequency” (Series 2002).

Ayappa and Rapoport concur with Sériès, suggesting that stage-related differences in the neural drive to the respiratory pump muscles will be important in determining the degree of upper airway collapse (Ayappa 2003). Slow deep breaths during slow wave sleep and rapid breathing with small tidal volumes during REM may explain limitation of obstruction during those stages in some individuals. Indeed, while they also support the view that REM-related apnoeas are longer and more severe, they actually state that

45 obstruction is most commonly seen in non-REM stages 1 and 2. The fact that this statement is not referenced in their review gives some indication, perhaps, of the difficulty in finding literature support for it. This almost certainly explains the overwhelming assumption of REM-predominance for OSA amongst the anaesthesia community.

The current state of knowledge does not allow the direct extrapolation of results in a diagnostic sleep laboratory to the situation in the postoperative ward. Sleep architecture may be affected by a variety of factors in each situation and any effect this might have on stage predominance of apnoea would be a matter of speculation. The effect of concurrent illness is also not known, although it could be assumed that surgical patients would have a similar incidence and degree of respiratory disease to the subjects of this enquiry.

Despite these potential limitations, the findings of this investigation strongly suggest that the effects of postoperative sleep disturbance on sleep and breathing are likely to be far more complex than previously thought. It should no longer be assumed that REM rebound, if and when it occurs, will produce a greater likelihood of obstructive apnoea in any individual, even if the patient is obese, although this latter group would appear to have a higher likelihood of REM predominant apnoea.

Sleep stage may be important for other reasons, however. Garpestad et al. demonstrated that REM apnoeas were associated with a higher blood pressure increase than those occurring in NREM, although the difference was modest (Garpestad 1995). While it was not possible to control for body position in this investigation, others have demonstrated that sleep stage also determines the position dependency of obstructive apnoeas in some subjects (George 1988; Pevernagie 1992). Interestingly, position dependence, the tendency for obstruction to occur mainly while supine, seems to be a largely NREM phenomenon. As patients tend to be nursed supine after many surgical procedures an increase in the NREM/REM ratio of apnoeas might be expected postoperatively in certain people. This is supported by the findings of Rosenberg- Adamsen et al. (Rosenberg-Adamsen 1997) and would also negate, to some extent, the assumption that REM rebound leads to greater risk. NREM supine apnoeic events may be an important factor increasing apnoeic episodes in the early postoperative phase,

46 prior to any period of REM rebound, consistent with the findings of Rosenberg et al. (Rosenberg 1994).

In contrast to OSA, central sleep apnoea is reported to be most common in NREM stages one and two (Burgess 1997) and actually reduced by REM sleep (Wilcox 1998). Less common than OSA, it is nevertheless prevalent in certain conditions, especially men with left ventricular dysfunction (Findley 1985; Javaheri 1995; Javaheri 1998). Even when congestive heart failure is stable and optimally treated the incidence of CSA has been reported to be as high as 75% in this group (Tremel 1999).

Conclusion

We currently have very limited data about the role of sleep in the recovery of patients from surgery and anaesthesia. Most of the views expressed to date, therefore, have been largely speculative. While it is plausible that a REM rebound-related increase in apnoea is a cause of postoperative morbidity, data from the sleep laboratory setting suggest that this assumption is very simplistic, probably incorrect for the most part, and that more direct evidence from the postoperative ward is required. A small number of patients, mainly obese patients with severe OSA, do appear to have a tendency to clinically significant REM predominance of their apnoea. It is possible that REM rebound might be more important to this group, the preoperative identification of which remains a major problem. The effects of sleep stage on factors other than upper airway patency also remain to be determined.

The role of CSA and its sleep stage dependence has not been considered at all in the postoperative setting. Given that CSA is common with patients suffering from cardiac dysfunction, a group already at increased perioperative risk, this disorder warrants further investigation. The reported NREM predominance of this form of sleep apnoea further complicates the issue of postoperative sleep disturbance.

47 CHAPTER 4

CIRCADIAN VARIATION IN UNEXPECTED POSTOPERATIVE DEATH

In a retrospective study published in 1992, Rosenberg et al. reported a circadian variation in unexpected postoperative deaths in Denmark (Rosenberg 1992). Their hypothesis was that nocturnal episodic hypoxaemia might be contributing to myocardial ischaemia, infarction, arrhythmia and sudden death, particularly in the late postoperative period. While they made no comment about it in the report, they also included data suggesting a peak of deaths on the third to fourth postoperative days. In a subsequent review (Rosenberg-Adamsen 1996a), the same group of authors suggested rapid eye movement sleep (REM) rebound, with an associated worsening of nocturnal episodic hypoxaemia in the middle of the first postoperative week, as a potential contributor to these deaths. Moreover, based on that single 1992 paper with very small numbers, it has been recently stated that “after major operations most sudden and unexpected deaths occur during the night” (Gogenur 2002).

To determine if unexpected postoperative deaths at the Royal Prince Alfred Hospital, Sydney, Australia, exhibited a circadian variation, data from all patients who died within one week of surgery under general anaesthesia, in the 6.5 year period from 1 July, 1992 to 31 December, 1998, were reviewed. As Rosenberg’s group had done, it was assumed that any effect of surgery and anaesthesia on sleep architecture that might contribute to postoperative mortality would have passed by the end of the first week. It was also thought that deaths beyond this period would be unlikely to be sudden or unexpected with the exception of those caused by embolism.

48 Methods

Patient Selection

The hospital maintains a computerised database of all admissions, surgical procedures (undertaken in the operating theatres) and deaths. This database was used to identify patients who had died after visiting the operating suite in the 6.5 year period from 1 July, 1992 to 31 December, 1998. Information from the database was used to exclude deaths beyond one week, deaths occurring after procedures performed with local anaesthesia (central venous catheter insertion for example) and expected postoperative deaths such as inoperable ischaemic bowel or inoperable ruptured abdominal aortic aneursym, septic or cardiogenic shock, and catastrophic intracranial events. The medical records of patients identified as unexpected postoperative deaths were subject to independent review by two co-investigators. Both had to agree that the death was an unexpected postoperative event. Times of death were recorded as accurately as possible from the notes although this was not always obvious.

The Royal Prince Alfred Hospital is a 950 bed, tertiary referral hospital that undertakes a wide range of general, cardiothoracic, transplant, gynaecological and neurosurgical procedures. All surgical patients would have received routine thrombo-embolic prophylaxis.

Statistical Analysis

Data were analysed using Oriana for Windows version 2 (Kovach Computing Services, Anglesey, Wales, U.K.), a software package specifically designed for the analysis of data measured on a circular or directional scale. The strength, direction and 95% confidence limits were calculated for the mean vector. Rayleigh and Kuiper tests (Fisher 1993) were used to determine the likelihood that the data exhibited a circadian variation (see discussion below). P < 0.05 was considered significant.

49 Results

1338 postoperative deaths occurred within the 6.5 year period, 709 of these within 7 days of surgery. 244 were identified as potential unexpected postoperative deaths and detailed examination of the medical records of these patients was carried out. Three files could not be found. 37 fulfilled criteria for unexpected postoperative death. 15 of the 37 patients had a witnessed arrest and, where possible, the time of arrest was recorded as the time of death. It could not be determined from the notes of 3 patients whether or not their arrest had been witnessed. The remainder were “found unresponsive”, 6 in asystole, one in ventricular fibrillation and the initial rhythm of the rest was not recorded. The median age of the patients was 72 years (29 – 90), 19 were male, 18 had a known significant history of cardiac or respiratory disease and 15 had undergone surgery for acute conditions.

Times of death are displayed in Figure 4.1. The mean vector for time of death was 0718 h (see Figure 4.1) and the length of the mean vector was 0.12. The 95% confidence limits are also shown in Figure 4.1. The null hypothesis, that the data did not differ from a uniform distribution, could not be rejected using the Rayleigh test of uniformity (P = 0.57). Kuiper tests performed against both uniform and von Mises distributions similarly returned P-values greater than 0.15.

17 of the 37 patients underwent autopsy. Of these, three had pulmonary embolism, two a ruptured aortic aneurysm, one a haemopericardium and haemothorax, one a right coronary artery plaque rupture, one a new left anterior descending artery thrombosis with the rest having a diagnosis of myocardial infarction or ischaemic heart disease. Theoretically, deaths other than embolic ones would be more likely to occur as a result of obstructive sleep apnoea. Reexamination of the data with embolic deaths excluded produced no meaningful difference in the statistics. Including only those patients with autopsy proven non-thrombotic myocardial infarction or ischaemia (n = 9), the mean vector was 0709 h, its length 0.63 and P = 0.02 (Rayleigh). The times of death for this subgroup are shown in Figure 4.2.

50 The distribution of deaths according to the postoperative day is displayed in Figure 4.3. Similarly, the distribution of deaths due to autopsy-proven cardiac ischaemia is displayed in Figure 4.4 showing the same pattern as for the overall group.

Times of All Unexpected Deaths

FIGURE 4.1. Times of all unexpected postoperative deaths displayed on a 24 hour clock, showing mean vector (nominal peak) and 95% confidence interval. The length (not indicated on the graph) of the mean vector is 0.12. The null hypothesis, that the data did not differ from a uniform distribution, could not be rejected using the Rayleigh test of uniformity (P = 0.57). Kuiper tests performed against both uniform and von Mises distributions returned P-values greater than 0.15.

51 Times of Ischaemic Cardiac Deaths

FIGURE 4.2. Times of unexpected postoperative deaths, shown at autopsy to have been due to myocardial infarction or ischaemic heart disease, displayed on a 24 hour clock, showing mean vector and 95% confidence interval. The length (not indicated on the graph) of the mean vector is 0.63 and P = 0.02 (Rayleigh). This suggests a circadian variation consistent with previously published non- surgical population data.

52 All Deaths by Postoperative Day

Number of 6 Deaths

0 1 2 3 4 5 6 Postoperative Day

FIGURE 4.3. The overall number of deaths by postoperative day showing a lack of any mid-week peak.

53 Ischaemic Cardiac Deaths by Postoperative Day

Number of 3 Deaths

0 1 2 3 4 5 6 Postoperative Day

FIGURE 4.4. The number of deaths due to autopsy-proven myocardial infarction or ischaemic heart disease by postoperative day.

Discussion

A circadian variation for the onset of ischaemic chest pain in acute myocardial infarction (Muller 1985; Thompson 1991), time of transient myocardial ischaemia in patients with known coronary artery disease (Rocco 1987), time of onset for ischaemic stroke (Marler 1989) and onset of episodes of asthma (Ballard 1989) has been demonstrated for the general population. In the general population there is a peak of deaths between 0400 and 0600 hours (Muller 1985; Muller 1987; Willich 1987; Thompson 1991). Sudden cardiac death and the onset of acute myocardial ischaemia peaks between 0700 and 0900 hours, and troughs between 0900 and 1300 hours (Muller 1985; Muller 1987; Willich 1987; Thompson 1991). Ischaemic stroke shows a peak

54 incidence between 1000 and 1200 hours (Marler 1989). There are an increased number of asthmatic deaths at night (Ballard 1989).

Hypoxaemia is often accompanied by increases in heart rate and blood pressure, potentially exacerbating myocardial ischaemia. Arterial pressure fluctuations that occur with hypoxaemia may also increase the likelihood of coronary artery plaque rupture (Millar-Craig 1978). It has been demonstrated that constant and episodic nocturnal hypoxaemia (Rosenberg 1989; Reeder 1992a; Reeder 1992b) is common in the postoperative period, and a relationship between hypoxaemic events and ECG abnormalities (Rosenberg 1989; Rosenberg 1990) and myocardial ischaemia (Reeder 1991b; Gill 1992; Goldman 1993) has been suggested.

Sleep architecture is altered after some types of surgery (Rosenberg-Adamsen 1996a; Loadsman 2001). Knill demonstrated a ‘rebound’ of REM sleep with an associated increase in hypoxaemia in a small group of patients after major abdominal surgery (Knill 1987; Knill 1990), a finding apparently confirmed by others (Rosenberg 1994). Thereafter, a postoperative “mid-week” REM rebound appears to have become the assumed norm for most patients and considerable speculation has been made regarding the potential for this to worsen nocturnal episodic hypoxaemia, the assumption being that sleep apnoea is worse during REM sleep (Rosenberg-Adamsen 1996a; Benumof 2001) (see also chapter 3). It has been further suggested that the previous finding of a circadian variation in unexpected postoperative deaths (Rosenberg 1992), most occurring at night and towards the middle of the postoperative week, supports this hypothesis (Rosenberg-Adamsen 1996a).

There are a number of potential problems with these speculations. Of 46 patients having significant surgery with enough postoperative nights studied and published to date, 23 had a REM rebound (Johns 1974; Ellis 1976; Orr 1977; Kavey 1979a; Aurell 1985; Lehmkuhl 1987; Knill 1990; Rosenberg 1994). From these studies the average rebound is between 15 and 20 minutes of 350 minutes total sleep time, or around 5 per cent, lasting perhaps one or, rarely, two nights. Additionally, most patients do not have a significant worsening of their sleep apnoea during REM, although markedly obese patients may be a subgroup in whom REM predominance occurs more frequently (see chapter 3). The relationship between sleep and post-operative episodic respiratory

55 disturbance has also recently been questioned, Drummond et al. finding that many of the episodic events occurred during wakefulness (Drummond 2002).

There are several possible reasons why this examination of a retrospective case series did not demonstrate a circadian rhythm for unexpected postoperative death. A review of a retrospective case series of this nature is likely to suffer a number of errors and biases. Using a computer-generated database relies on accurate data entry by a large number of hospital personnel; doubtless there are some inaccuracies and omissions that become all the more important when analysing a small group. Similarly, the 3 lost and hence unreviewed files could be significant. As with the study of Rosenberg et al. (Rosenberg 1992), the review of patient files, although independent, was not blinded and may have been subject to personal bias. The recorded times of death are also subject to some inaccuracy. At night patients are routinely observed on an hourly basis on a nursing round of the ward and, as such, recorded times of death at night should be accurate to within an hour unless there was an extended resuscitation attempt. However, nursing shortages and intermittently busy periods on the ward mean that patients may go unobserved for longer periods. Indeed, it is my suspicion that the small clusters of deaths occurring at around 0800-1000, 1600 and 2000 hours in this investigation may have resulted from nursing staff discovering dead patients on their routine medication rounds and after nursing shift hand-over.

Examination of the small subgroup of patients with autopsy-proven myocardial infarction or ischaemia as the cause of death did suggest a circadian variation (Figure 4.2) but the mean vector of that group (just after 0700h) is consistent with previously published non-surgical population data (Muller 1985; Muller 1987; Willich 1987; Thompson 1991). This finding, therefore, also fails to support previous claims that most post-operative deaths occur at night.

It can readily be seen from Figures 4.3 and 4.4 that there is no ‘mid-week’ peak in the data, the death rate gradually decreasing from the day of surgery. There is a suggestion of a mid-week peak in the deaths presented by Rosenberg et al. (Rosenberg 1992). Although this was not specifically mentioned in the text of the paper, its presentation implies an association with late postoperative nocturnal hypoxaemia, an implication continued in subsequent discussions of the potential role of REM rebound in

56 postoperative morbidity and mortality (Rosenberg 1994; Rosenberg-Adamsen 1996a; Kehlet 1997). While the lack of any such mid-week peak in this study does not preclude the possibility that sleep-disordered breathing is involved in adverse postoperative outcomes, it does cast some doubt on the relative importance of REM rebound suggested or even stated outright by a number of authors (Rosenberg-Adamsen 1996a; Kehlet 1997; Benumof 2001).

The subset of patients undergoing surgery is not the same as the general population in whom the circadian pattern of cardiac events has been clearly demonstrated. Sympathovagal tone, a major biological determinant of circadian variation in cardiovascular function, is modulated through circadian patterns of sleep-wake activity. Sleep-wake patterns in surgical patients are often grossly disturbed, even before the surgery is performed when anxiety and such things as bowel preparation interfere. Postoperative patients are commonly found asleep during the day and any sleep-related contribution to morbidity or mortality might be expected to occur at any time, rather than just at night. In the postoperative period, regardless of the time of day, pain, surgical stress, fluid and electrolyte disturbances and altered pharmacological regimes contribute to exaggerated disturbances in neurohumoral activity. For these reasons, a circadian variation in unexpected postoperative deaths ought to be unlikely and the data from this investigation supports that conclusion. The recent finding of a lack of circadian variation in autonomic activity after major abdominal surgery by Gogenur et al. (Gogenur 2002), which they claim as a possible reason why “most postoperative deaths occur at night”, is, perhaps, more likely to be one reason why no such circadian variation in postoperative deaths was found in this study.

Discussion of Statistical Methods

Directional data includes such things as the orientation of structures and landforms in geology, wind direction in meteorology, angular measurement in astronomy, and various directional aspects of biology such as the movement of animals and insects (Pochron 2000; Wallace 2002; Mercado-Hernandez 2003), particularly during migration (Etheredge 1999), or other behavioural phenomena like the orientation of ball-rolling by dung beetles (Byrne 2003). Circular data also includes the large number of phenomena

57 exhibiting periodicity. Examples may be found in the measurement of the waves involved with electromagnetic radiation, acoustics and particle physics. Circadian and seasonal variations are another example of circular data and specific circular statistical methods are considered the most appropriate for their analysis.

The reason for the use of special statistical methods is the fact that one ‘end’ of the circular scale is the same as the other. For the purposes of summary statistics, 0o equals 360o and an arithmetic mean, therefore, is largely meaningless. The mean of 2300 hours and 0100 hours cannot be 1200 hours. For this reason, circular statistical methods derive a mean vector of the data that has both direction and length, the latter indicating the directional ‘strength’ of the mean. In circadian terms, this would indicate the strength and direction of any peak in the data, a short mean vector (near zero) suggesting no peak with a longer mean vector (nearer one) indicating a strong circadian peak. A circular range, the shortest arc length including all the data, may also be calculated, as may confidence intervals for the mean vector.

Special significance tests are also required for the analysis of circular data. The most commonly used is the Rayleigh test that tests the data against the null hypothesis of uniformity using the number of observations and the length of the mean vector. The Kuiper test may be used to test the data against any selected null hypothesis including uniformity or a distribution such as von Mises, which is a unimodal distribution equivalent to the “normal” distribution in parametric statistics. Other specific tests may also be applied when comparing two or more sets of circular data but these are not relevant to this investigation.

Some caution is required when applying such tests. A strongly bimodal distribution may result in a short mean vector and therefore a non-significant result using the Rayleigh test. It is fairly clear from the raw data plot (figure 4.1) that this is unlikely to be the case in this investigation.

Circular statistical methods now appear to be commonly utilised in many other branches of science such as those listed above and a few examples have been cited. Their use in medicine is difficult to assess accurately as many papers make no mention of statistics in the searchable information. Thus far, circular methods appear to be underutilised in

58 the medical literature and the examples found mainly deal with spatial and periodic aspects of molecular biology (Bacchi 1999; Audit 2003), neurophysiology (Ross 1980; Hirsch 1983; Drew 1991; Lebedev 1996; Fitzpatrick 2002; Levine 2002; Rizzuto 2003) and tissue structure (Canham 1991a; Canham 1991b; Whittaker 1991; Canham 1997; Mann 2003).

Specific examination of the literature involving circadian and seasonal variations of specific events reveal very few in which circular statistical methods were used (Ghiandoni 1998; Hawley 2001; Rocchi 2001; Rocchi 2002). All of these involved seasonal rather than circadian analyses with all but one conducted by a single group of biomathematicians. A number have used harmonic regression techniques which may also be appropriate but even in the most recent literature there appears to be a significant number in which alternative and perhaps less satisfactory statistical methods have been used to assess circadian and seasonal variations. Of the 25 papers examining circadian variations in specific events published in the medical literature over the three years 2001 – 2003, 12 used grouped time data (division of the day into time intervals) (Bhalla 2001; Cheung 2001; Kinjo 2001; Simantirakis 2001; Bhalla 2002; Gogenur 2002; Stephenson 2002; Stergiou 2002; Yamasaki 2002; Henriques 2003; Khan 2003; Lee 2003) and 10 used harmonic regression with or without grouped data analysis (Gallerani 2001; Gillis 2001; Bellamy 2002; Casetta 2002; Manfredini 2002; Mehta 2002; Watanabe 2002; Aronow 2003; Delle Karth 2003; Rana 2003). The statistical methods were unable to be determined from 3 papers because the full text was not obtainable (Bilora 2001; Fries 2001) or the methods used were not mentioned (Wolpert 2001). Rosenberg’s original study divided the 24 hours into three 8 hour periods, day, evening and night. The number of deaths occurring at night was then multiplied by 2 and compared with the day and evening deaths together. This may not have been the most satisfactory statistical approach, but the distribution of their raw data suggests the analysis would have been significant using circular methods as well (Rosenberg 1992).

While the grouping of time data will simplify its collection, it introduces a number of potential biases. For example, the arbitrary division of periods at a given time might well split a peak in data points, resulting in failure to find a real circadian variation. Given these limitations and the availability of at least one and probably two more satisfactory alternatives (circular analysis and harmonic regression analysis) it is

59 perhaps surprising that this appears to be the first time circular statistics have been used to assess circadian variations of specific events in a medical setting.

Conclusion

To date only two studies have sought a circadian variation in unexpected postoperative death. Both studies found unexpected postoperative death to be a relatively uncommon event and had similarly small numbers within like size populations of total postoperative deaths over 5 to 6.5 year periods. Rosenberg’s study claimed to demonstrate a circadian variation in unexpected postoperative death. No such variation was found in this study and, given the disruption to sleep architecture and various other aspects of circadian function in the postoperative setting, this is hardly surprising.

The lack of a circadian variation in unexpected postoperative deaths does not, by virtue of the same postoperative disruption of sleep patterns, support the conclusion that sleep plays no role in perioperative morbidity and mortality. The findings of this study however, in conjunction with those presented in chapter 3, do introduce considerable doubt regarding the common speculation that a mid-postoperative week REM rebound is a significant contributing factor.

60 CHAPTER 5

METHODS FOR THE STUDIES OF PERIOPERATIVE SLEEP

Most of the chapters that follow involve perioperative polysomnographic recording of sleep. Specific methodological and statistical techniques are noted in the relevant chapters but generally the methods were as described below. It is also necessary to record a number of limitations and difficulties regarding the methodology.

Recruitment

With the approval of the institutional ethics committee (approval number X98-0270), a total of 19 patients were recruited to undergo perioperative polysomnographic studies and all gave written, informed consent. One patient had a sleep study several months preoperatively as part of an assessment for sleep-disordered breathing. This was discovered after that patient had agreed to participate but it may have influenced her decision to enrol.

Recording Methods

Polysomnography included electroencephalography (EEG: C3-A2 and C4-A1), electro- oculography (EOG) and submental electromyography (EMG) for assessment of sleep as described by Rechtschaffen and Kales (Rechtschaffen 1968; Carskadon 1989). An oro- nasal thermistor airflow transducer, thoracic and abdominal inductive plethysmography and pulse oximetry were used to record respiration. Electrocardiography (ECG: Lead I) and leg movement transducers were also used for all subjects. Body position analysis with a mercury position transducer was attempted in the earliest studies but this was abandoned as a result of persistent problems (see below). Audio recording of snoring with a lapel microphone taped to the forehead was used in nearly all cases (again, see

61 limitations below). The actual set-up used for this work, not including the position sensor or microphone, is shown (not, in this instance, on one of the research subjects) in figure 5.1.

All studies were unattended recordings using a P-Series 2 portable polysomnogram recording device (Compumedics, Melbourne, Australia), using, in all cases, the patient leads/electrodes supplied with the equipment. Saturation, heart rate and body position were sampled at 1Hz, sound, airflow and movement of the thorax and abdomen at 10Hz, leg movements at 25Hz, ECG and EOG at 50Hz, EEG and submental EMG at 125Hz. Signal gain in the P-Series Portable is factory preset. Care was taken to ensure impedance was less than 3kΩ for EEG, EOG, EMG and reference electrodes at the time of set-up and signals were visually inspected on the screen of the device. Gold-plated cup electrodes are supplied with the P-Series 2 for these channels and EC2 Electrode Cream (Grass Instrument Division, W. Warwick, Rhode Island, USA) was used as the conductive medium.

The P-Series Portable Manager version 2 software supplied by Compumedics was used to transfer the recorded raw polysomnography data to a computer. The studies were then staged manually in 30-second epochs using the W-Series Replay version 2 software also supplied by Compumedics. A 50Hz digital notch filter was applied to EEG and EMG recordings during scoring to reduce the effect of mains interference. Rarely, a 30Hz digital low-pass filter was also applied during scoring to individual EEG traces for parts of the study where it became clear that the signal had been affected by increased impedance, leading to excessive 50Hz mains intrusion despite the digital notch filter.

I carried out the polysomnogram recording set-up and data analysis in all cases, with the exception of the single study carried out several months preoperatively as mentioned previously. I was trained in both set-up and analysis techniques by the nursing staff and technicians of the Sleep Unit, Royal Prince Alfred Hospital, prior to undertaking any data collection. I also had a full polysomnographic study myself in order to appreciate its effect on the subjects. It was not pleasant.

62 FIGURE 5.1. Photo showing the attachment of the polysomnographic equipment used for this research. EOG = electro-oculogram; EMG = electromyogram. Electroencephalogram electrodes are hidden in the hair and behind the ears and electrocardiogram under the shirt. The abdominal inductive plethysmography band is higher on the abdomen than it normally would be.

Figures 5.2 – 5.6 contain a series of exemplary screen snapshots (in landscape, menu bars removed) of the Replay software, showing subject 1 in, respectively, wakefulness, stage 1, stage 2, stage 3 (slow wave or delta) and REM sleep during his preoperative

63 night. At the top of the screen is the whole-night hypnogram indicating the overall sleep architecture. Below this, in the top window of the split-screen, is one epoch (30 seconds) of the ECG, EEG, EOG, submental EMG and leg movement traces. The bottom window contains 5 minutes of the respiratory-related traces, the vertical solid line indicating the start of the current epoch in the window above.

Figure 5.2 can be clearly identified as wakefulness by the relatively high EMG tone, the movement artifact towards the end of the epoch and blinking in the EOG. The low- amplitude alpha rhythm of relaxed wakefulness with eyes open is not easy to differentiate from the EEG of stage 1 sleep at this resolution (compare with figure 5.3). The breath-to-breath irregularity and movement obvious in the respiratory traces for many of the surrounding epochs is also typical of wakefulness.

In figure 5.3, the alpha-predominant EEG is replaced by low amplitude, mixed- frequency waves and some slow eye movements can also be seen although these are not as prominent in this epoch as elsewhere. More obvious in this stage 1 sleep is the reduction in EMG tone and the clear onset of obstructive and central respiratory events, one of which is associated with an EEG arousal during the epoch shown.

Stage 2 sleep can be identified in figure 5.4 by the high-amplitude delta activity (less than 20% of the epoch) and the K-complex immediately preceding (or actually part of) an arousal. In figure 5.5, high-amplitude delta activity now makes up more than 20% of the epoch, marginally less than 50%, and the epoch is therefore scored as stage 3. Note the relative lack of respiratory events compared with stage 2 with this subject and the appearance of frontal delta EEG activity in the EOG traces, not to be confused with REMs.

Figure 5.6 shows an example of REM sleep identified by the characteristic deflections in the EOG channels, further reduction in EMG amplitude, and marked variability of both heart and respiratory rates, not associated with arousal. Note again the relative lack of upper airway obstruction compared with stage 2 with this subject (see chapter 3).

Appendix 1 contains an example of the three pages of report generated for each study, in this case again the report for the preoperative night of subject 1. Note on the second

64 page the body position trace. Although the transducer was used in this case, it was obvious that the output was unreliable, despite calibration according to the manufacturers instructions (see limitations below).

65 FIGURE 5.2. Subject 1, preoperative study, awake (legend continues over).

66 Figure 5.2 legend continued – an epoch of wakefulness from the preoperative night of subject 1. The epoch is readily identified as wakefulness by the moderate amplitude, high frequency EEG, blinking and high amplitude EMG.

The very top section contains the all-night hypnogram showing (vertical bar) the position of the current epoch in the study. The main window is split into two sections. AC (sleep) signals are displayed above at 30 seconds/page: ECG = electrocardiogram; EEG = left electroencephalogram (C3/A2) and EEG(sec) = right electroencephalogram (C4/A1); EOG = electro-oculogram, (L) left and (R) right; EMG = electromyogram; LEG(R) = right leg movement transducer signal. DC (respiratory) signals are displayed below at 5 minutes/page: SaO2 = oxygen saturation; H.R. = heart rate (from pulse oximetry); SOUND = audio level from microphone; AIRFLOW = thermistor airflow trace; THOR and ABDO RES = thoracic and abdominal plethysmography traces respectively. The solid vertical bar in the lower section indicates the start of the 30-second epoch above.

67 FIGURE 5.3. Subject 1, preoperative study, stage 1 (legend continues over).

68 Figure 5.3 legend continued – an epoch of stage 1 from the preoperative night of subject 1. The reduced EMG amplitude, rolling eye movements (not prominent) and onset of respiratory events identify this as stage 1.

69 FIGURE 5.4. Subject 1, preoperative study, stage 2 (legend continues over).

70 Figure 5.4 legend continued – an epoch of stage 2 from the preoperative night of subject 1. The epoch is identified as stage 2 by high amplitude delta EEG activity (<20% of epoch) and the K-complex preceding the arousal.

71 FIGURE 5.5. Subject 1, preoperative study, stage 3 (legend continues over).

72 Figure 5.5 legend continued – an epoch of stage 3 from the preoperative night of subject 1. The epoch is readily identified as stage 3 (slow wave sleep) by the high amplitude (>75µV) delta EEG activity (20 – 50% of epoch). Note the frontal delta activity appearing in the EOG, not to be confused with rapid eye movements, and the lack of obstructive respiratory events compared with stage 2.

73 FIGURE 5.6. Subject 1, preoperative study, REM (legend continues over).

74 Figure 5.6 legend continued – an epoch of rapid eye movement (REM) sleep from the preoperative night of subject 1. The epoch is identified as REM particularly by the rapid eye movements seen in the EOG channels but also by the very low amplitude EMG. Note the variation in respiratory amplitude, also common in REM.

Technical limitations

Ideally, an occipital EEG derivation should be recorded to assist with the staging of sleep, particularly sleep onset and arousals (Carskadon 1989), although this is by no means considered essential for accurate sleep staging (Rechtschaffen 1968; Carskadon 1989; 1992). The Compumedics P-Series device is supplied with the facility to record two channels of EEG, although it would appear that this might be increased to five. Any more than two channels would require significantly greater memory size than that supplied with the device in order to record for a full night with the resolution and other signals necessary for an adequate study. For this reason, and because a redundant vertex channel was necessary to allow for signal loss and artifact assessment, no occipital channels were recorded for this study. It would also have been essentially impossible to reliably attach occipital electrodes to patients lying flat postoperatively.

75 Another unfortunate limitation was the unreliability of the position sensor supplied with the P-Series device in this setting. This was a cumbersome sensor and proved difficult to calibrate with surgical patients. It frequently became detached and two ceased to function altogether after only a few studies. Its use was therefore abandoned for most patients in this study. Body position would have been a very useful addition to the data collected and future research into perioperative sleep should ideally address the issue of position transduction if possible. Alternatively, monitored sleep studies might avoid the need for position sensors but the necessary resource allocation would be much greater in that case.

Similarly, the audio microphone was used on some but not all subjects. The microphone required a specific 3V button-style cell to function and a fresh one was not always available at short notice. The information gained from this recording was by no means essential to the study but when present it did provide additional information that was useful under certain circumstances, an example being the confirmation of sleep by the presence of snoring when the electophysiologic indications were equivocal.

Practical problems

As recently reported by Drummond et al. (Drummond 2002), the performance of perioperative polysomnograms is technically difficult. Helfaer et al. (Helfaer 1996), for example, lost a sixth of their subjects as a result of technical failures despite what would appear to be supervised recordings. Patient recruitment and compliance are a problem because of the intrusive nature of the monitoring. The thermistor flow transducer was, by far, the biggest issue for the patients in terms of discomfort. One patient with a particularly short upper lip found it almost intolerable and, to my surprise, found this worse than any postoperative pain he experienced. Monitoring of electrical signals was hampered by issues such as sweating as also reported by Drummond et al. Application of abdominal and, to a lesser extent, thoracic plethysmography bands was made difficult by abdominal wounds and dressings where present.

A number of technical and practical aspects of this study rendered many patients unsuitable for inclusion. Nasogastric tubes, for example, made it impossible to study patients postoperatively using the thermistor supplied with the P-series equipment. A

76 large number of otherwise suitable patients, such as those having bowel surgery or major abdominal vascular surgery were, at the time this study was conducted, all given bowel preparations on the preoperative night, making preoperative study a rather messy proposition. Several patients declined further involvement in the study after the preoperative night, mainly as a consequence of the somewhat intrusive nature of the monitoring. The very description of the monitoring set-up alone was enough to make the majority of patients refuse to be involved at all. Recruitment, therefore, proved to be a very difficult matter. The total number of requests was not recorded but I estimate that my success in recruitment was less than a fifth of the potential subjects approached.

Study Protocol

Initially, the intention was to collect data on each subject for the preoperative night, for at least five postoperative nights and during daytime naps if possible. This was noted in the Information for Participants (appendix 2), along with precise details of the recording set-up, as required by the institutional ethics committee. This description deterred many potential subjects, refusal often occurring without reading the rest of the document. As I considered that the preoperative night and at least 5 postoperative nights were important for the examination of postoperative sleep architecture, for many months early in the research only patients admitted prior to surgery with an expected postoperative stay of at least 5 nights were approached. This proved to be an almost fruitless exercise, with just one single patient (subject 4) completing the protocol to any satisfactory degree. At the same time, the hospital moved to a day-of-surgery admission policy and, within a few months of study commencement, preoperative studies became almost impossible, with only patients having bowel preparation or similar complex procedures being admitted prior to their operative day. A few patients already admitted for semi-urgent surgical conditions or illnesses unrelated to their surgery were the only subjects recruited for preoperative monitoring after that. It therefore became obvious that the inclusion criteria had to be significantly relaxed, with the expectation that few patients would come anywhere near completion of the original protocol.

The concept of monitoring daytime naps was quickly abandoned. Normal daytime surgical nursing care in the ward involves bathing and movement of the patients out of

77 bed at the very least, making it impossible to leave the monitoring equipment in place during the day. All polysomnographic recording set-ups had to be entirely reapplied in the evening of each study night. The near impossibility of recording daytime sleep after surgery is reflected in the fact that few studies have ever managed to achieve it, all in an intensive care environment. Johns and coworkers studied patients all day and night for several days after open cardiac surgery (Johns 1974) while Ellis and Dudley (same group of investigators) imply they did the same thing after various abdominal procedures without actually mentioning as much in their methods (Ellis 1976). Their recording technique, however, relied on automated counts of delta wave activity rather than standard sleep polysomnography and this can not be considered satisfactory to assess sleep accurately. Aurell and Elmqvist (Aurell 1985) were able to record, using standard polygraphic techniques, various continuous periods up to three days and nights after major abdominal surgery and, more recently, Edéll-Gustafsson and colleagues continuously recorded for 48hrs after cardiac surgery (Edell-Gustafsson 1999). These studies all relied heavily on the cooperation and involvement of intensive care nursing staff, a contribution unavailable for this study.

It was also initially intended that the subject sample would be drawn entirely at random from the surgical lists of the hospital, and indeed this is the case for subjects recruited preoperatively. It should be noted, however, that several of those subjects were patients I anaesthetised as these proved to be far more likely to agree to participate than patients for whom I had no clinical role. For the same reason, all but one of the subjects commencing study postoperatively were my own patients and could not, therefore, be considered a random sample. Most of the latter subjects agreed to participate on the basis that the study might benefit them personally, an obstruction-prone airway having been identified perioperatively. Several of these patients even refused monitoring for any more than a single night, emphasising the extreme difficulty associated with recruitment and retention of subjects for this study. Table 5.1 indicates the nights studied for all 19 subjects.

78 Subject Nights Studied Preop Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7

1    2   3  4       5     6  7   8   9    10   11  12  13  14  15  16  17   18   19  

TABLE 5.1. Perioperative nights studied for each subject. Preop = preoperative night; Postop = postoperative night;  indicates night studied.

Most regretably, in two cases it would appear that deliberate intervention by ward nurses who were concerned that the monitoring was overly intrusive resulted in premature withdrawal from the study and subsequent refusal of the patients to continue with monitoring in the postoperative period. Total loss of data occurred in one of these cases.

Sleep Staging

Sleep was manually staged in all cases according to the criteria of Rechtschaffen and Kales (Rechtschaffen 1968), with consideration for the effects of age and sleep pathology as recommended by Bliwise, Carskadon and Rechtschaffen (Bliwise 1989; Carskadon 1989). Standard definitions and criteria (ASDA 1992; AASM 1999) were similarly used for the scoring of arousals and respiratory events. In some postoperative

79 studies, however, the electrophysiological staging of sleep was ambiguous, particularly between stages 1, 2 and wakefulness. Rapid eye movement (REM) sleep was much less ambigious while slow wave sleep (SWS, sleep stages 3 and 4) was not ambiguous at all. The uncertainty could often be clarified by simultaneous examination of eye movements and respiratory parameters and where this was possible the epoch was staged as wakefulness or sleep as appropriate, otherwise the epoch was scored as awake. This approach was probably somewhat different to that taken in previous studies and the potential implications of this are discussed in detail in the relevant chapters.

Statistical Analysis

Statistical analysis of preoperative data (chapter 6) is descriptive. Between-night comparisons of postoperative data (chapters 7, 9 and 10) was, ultimately, limited by missing nights and small numbers of subjects. Repeated measures (longitudinal data) analysis, the most appropriate for multiple between-night comparisons, while possible, would have been difficult to perform and of questionable accuracy because the missing nights were unlikely to have been lost as a result of completely random factors (Cnaan 1997; Albert 1999). Bearing in mind the problem of repeated measures, Wilcoxon signed rank test for matched pairs was chosen for the limited between-night comparisons because it could not be assumed the data were Gaussian. While population values for the percentage of most sleep stages and other sleep parameters would probably be normally distributed, postoperative values are highly likely to be skewed. Moreover, the heterogeneity of the subjects and their surgical insults would also predispose the results towards non-normality. The use of a non-parametric test, however, is not as sensitive with small samples such as this, probably explaining the lack of statistical significance for several of the measures that appeared to show fairly distinct postoperative alterations that were consistent with the findings of other studies. All other comparisons of paired non-parametric data were performed using two-tailed Wilcoxon signed rank test and comparison of unpaired non-parametric data with two- tailed Mann-Whitney U test (Prism 3.0a for Macintosh, Graphpad Software Inc., San Diego, California, USA). P < 0.05 was considered significant (Altman 1999).

80 CHAPTER 6

PREOPERATIVE SLEEP AND BREATHING

A number of studies using various polysomnographic techniques have been carried out to assess perioperative alterations in sleep architecture and/or the subsequent possible respiratory consequences associated with major non-cardiac (Ellis 1976; Kavey 1979b; Kavey 1983; Aurell 1985; Catley 1985; Lehmkuhl 1987; Knill 1990; Rosenberg 1994; Cronin 2001; Rahman 2001; Drummond 2002; Wu 2003), cardiac (Johns 1974; Orr 1977; Edell-Gustafsson 1999) and more minor forms (Rosenberg-Adamsen 1996b) of surgery. Studies using polysomnography after upper airway surgery, all involving either adenotonsillectomy or uvulopalatopharyngoplasty, have also been reported (Johnson 1986; Burgess 1992; Helfaer 1996; Terris 1996) as have a number of studies using actigraphic techniques to monitor pre- and postoperative sleep after minor in- and outpatient surgery in adults and children (Bisgaard 2002; Kain 2002; Kain 2003).

Amongst these studies there is wide variation in methodology and few included data from the immediate preoperative period. Detailed recordings of respiratory variables during perioperative sleep are almost non-existent, especially in the preoperative situation. Some of that data is of questionable quality and some has been collected under somewhat artificial perioperative circumstances. It is important to address these issues as assessments of postoperative sleep and its implications rely to some extent on preoperative data. The quality of preoperative sleep itself may impact in a variety of ways on other aspects of perioperative wellbeing. These factors require closer examination of the studies reported to date and the addition of further data to the pool currently available.

81 Methods

As previously described, with the approval of the institutional ethics committee, a total of 19 patients were recruited to undergo perioperative polysomnographic studies and all gave written, informed consent.

Patients commencing study preoperatively were recruited from the routine operating lists of the Royal Prince Alfred Hospital, Camperdown, Australia, with no particular prerequisite conditions or exclusions. The decision to approach the patients was made solely from the operating list with no knowledge of the patients’ prior medical histories. All were accommodated in a normal surgical ward environment in multiple-bed rooms as are the majority of patients in our institution.

The method of both collecting and examining the polysomnographic data is described in chapter 5.

Definitions

Statistical Analysis

Summary statistics are presented as mean ± standard deviation unless otherwise indicated.

Results

Of the 19 patients who enrolled, 9 had a study on the preoperative night and none of these had a prior diagnosis or history suggesting sleep disorders or sleep-disordered

82 breathing. The demographics and medical histories of these 9 subjects, all male, are shown in table 6.1. None of the nine were particularly obese. Several patients were suffering acute conditions for which they required semi-urgent surgery.

Three subjects were admitted to hospital for various periods prior to their immediate preoperative day. Subject 1 had spent six days in hospital already with an acutely ischaemic and painful leg, subject 5 several weeks with cardiac decompensation, and subject 8 one day for radiological investigations relating to his imminent endoluminal aortic aneurysm repair.

As environmental conditions might have affected sleep, the number of patients in the rooms was recorded for each subject and these are also shown in table 6.1. Some of the subjects (1, 2, 3 and 8) were accommodated in air-conditioned wards while the rest were in older wards without air-conditioning. The possible effect of room temperature was not considered until after data collection was complete so it was not contemporaneously recorded. Nevertheless, daily minimum and maximum temperatures for Sydney were subsequently obtained for all study nights from the Australian Bureau of Meteorology (ABOM 1999).

Many of the subjects (1, 2, 3, 5, 7 and 8) had histories involving previous hospital admissions, some involving prior major surgery (subjects 1 and 8).

Three subjects took medications prior to or during the study that may have affected the results (Bauer 1993). Subject 1 had been taking diazepam 2mg morning and night for many years and he also required 60mg of codeine for pain at approximately 0230 hours during the study. Subject 2 had been chronically taking theophylline. Subject 7 was also taking theophylline as well as clonazepam 500mcg/night as long-term regular medications. He additionally consumed 60mg of codeine at 2030 prior to study commencement, a medication he had been using “as required” for some time for reasons that are unclear.

83 Subject Age Weight Height Patients History Concurrent (kg) (m) in room medications 1 52 77 n/a 4 Vasculopath: previous nitrates regularly and as surgery for ischaemic required, diltiazem, legs (aorto-bi-iliac paracetamol 1g + codeine bypass), ischaemic heart 60mg as required for leg disease (AMI, CAGS). pain, diazepam 2mg twice Hospitalised 1 week daily preoperatively with recurrent acutely ischaemic right leg. 2 73 56.5 1.7 2 Left inguinal hernia. verapamil, nitrates, Severe chronic airways theophylline, salbutamol, disease, ischaemic heart ipratropium, disease. beclamethasone

3 64 68 n/a 4 Right inguinal hernia. captopril, frusemide, Moderate to severe heart glibenclamide failure. 4 69 74 1.68 2 Carcinoma of the prostate. Nil Otherwise well. 5 73 57 n/a 2 Hospitalised 3 weeks digoxin, frusemide, preoperatively with atrial salbutamol, paracetamol fibrillation and pericardial 1g + codeine 60mg as and pleural effusions. CCF required (significant use with orthopnoea resolved in preoperative days) preoperatively. Dust- related lung disease. Found incidentally to have right hydronephrosis.

6 73 76 1.78 4 Urethral stricture. naproxen, allopurinol Otherwise well. 7 71 62 1.65 4 Prostatism. Severe salbutamol, ipratropium, chronic airways disease: theophylline, budesonide, FEV1/FVC = 0.5/1.6 ranitidine, fluticasone, salmeterol, terbutaline, clonazapam (500mcg nocte), paracetamol 1g + codeine 60mg as required (fairly regularly)

8 75 67 n/a 4 Abdominal aortic Nil aneurysm. Previous bowel surgery, otherwise well.

9 52 80.7 1.725 2 Metastatic melanoma to Nil left inguinal lymph nodes. Otherwise well.

TABLE 6.1. Subject demographics. AMI = acute myocardial infarction, CAGS = coronary artery graft surgery, CCF = congestive cardiac failure. n/a = not available.

84 A full night study (more than seven hours recording) was achieved in 7 subjects. Recording ended after 5.5 hours in one study (subject 5), presumably as a consequence of battery failure. Subject number 6, encouraged by nursing staff after he had managed only 48 minutes sleep in the first 4.5 hours of the study, removed his electrodes and withdrew from further involvement in the trial.

Overall sleep results for the subjects are shown in table 6.2 and specific sleep stage analysis is recorded in table 6.3. Sleep efficiency (60 ± 23%) varied over a remarkably wide range, from particularly poor (17%) to remarkably good (86%), even with the age of the subjects considered. This is reflected in the broad range of total sleep times (270 ± 119 minutes). Excluding subjects 5 and 6 on the basis of their incomplete recordings, the minimum efficiency improves to 30% which is still quite poor. The mean of the 7 remaining subjects was 67 ± 19%. The range in total sleep time narrows, however, to values similar to those found in other preoperative studies (312 ± 87 minutes). Five subjects exhibited a relatively normal sleep latency of around 20 minutes, but then there is a large gap to the next group of two with latencies over an hour and then two more with latencies longer than 3 hours. Again excluding subject 6 as his data is unhelpful in this respect, the proportions of each sleep stage also varied quite widely but are consistent with the ranges expected for the age group of the subjects. Stage one made up 10 ± 5%, stage two 57 ± 7%, stage three 16 ± 6%, stage four 2.7 ± 2.6% and REM 15 ± 7%. There appears to be a tendency for patients with chronic respiratory disease (subjects 2, 3 and 7) to have a higher proportion of slow wave sleep (SWS – stages 3 and 4 combined) than the rest of the subjects. While not reflected in the data presented, I also noted while scoring these studies that a large proportion of the epochs scored as stage 2 by standard criteria were borderline for stage 3, again suggesting a disproportionately high amount of delta EEG activity in the presence of respiratory disease.

85 Subject Report Sleep REM Sleep Total Sleep time latency latency period sleep efficiency (min) (min) (min) (min) (min) (%) 1 435.5 19 44.5 434 374.5 82.6 2 495 180.5 81 263 150 30.3 3 512 78 66 433 350.5 68.5 4 480 19 143 444.5 310 64.6 5 331 23 65.5 308 194 58.6 6 271.5 222.5 n/a 48 47.5 17.5 7 423.5 28.5 68.5 395 317.5 75 8 486 22 131 463 418.5 86.1 9 442.5 62 150.5 375.5 264 59.7

TABLE 6.2. Sleep wake data (REM = rapid eye movement sleep).

Subject Stage 1 Stage 2 Stage 3 Stage 4 REM Wake/sleep (%) (%) (%) (%) (%) 'transition' (min) 1 4.9 47.8 21.4 0.7 25.2 1 2 3 63.7 25.3 0.3 7.7 36.5 3 16 49.4 18.7 6.1 9.8 30 4 12.3 61.8 7.1 0.5 18.4 38 5 6.2 61.3 17 3.6 11.9 6 6 4.2 85.3 10.5 0 0 2 7 8.2 49.6 17.3 6.6 18.3 5 8 14 61.8 9.7 0.4 14.2 20.5 9 15 58.1 11 3.4 12.5 53.5

TABLE 6.3. Sleep stage data (REM = rapid eye movement sleep).

In the ‘wakefulness’ prior to any electroencephalographic indication of stage 1 sleep, most subjects exhibited other signs suggestive of imminent sleep onset, in some cases (2, 3, 4, 8 and 9) for many minutes. These signs included reduction of EMG amplitude, changes in respiratory pattern and, particularly, slow rolling eye movements. This phenomenon, the total duration of which is recorded for each subject as ‘sleep/wake transition time’ in table 6.3, is recognised in the sleep literature (Carskadon 1989). An example of this phenomenon is shown in figure 6.1 where the onset of hypopnoeic events and slow rolling eye movements precedes any EEG indication of stage 1 sleep.

86 FIGURE 6.1. “Wake/sleep transition” time prior to stage 1 (legend continued over).

87 Figure 6.1 legend continued – an epoch of from the preoperative night of subject 1. Note the wakefulness-type EEG but attenuated EMG and rolling eye movements. Also visible in the airflow trace is the onset of sleep-related oscillations in ventilatory drive.

Parameters associated with sleep-related breathing events for the subjects are shown in tables 6.4 and 6.5. Considering this group of subjects presented for surgery with no prior suspicion of sleep apnoea, sleep-disordered breathing (SDB) was remarkably common. Three subjects (33%) had no or insignificant SDB, one (11%) had mild SDB, three (33%) had moderate SDB and two (22%) fell into the severe category. Two thirds of the subjects (67%) therefore had some degree of sleep apnoea by current criteria. While this incidence might seem surprisingly high from the perspective of anaesthetists, it falls well within epidemiological expectations for males in this age range (Ancoli- Israel 1994).

The number of arousals from sleep occurring overall and as a consequence of respiratory events are listed in table 6.6 with these subdivided according to sleep stage in table 6.7. Arousals were frequent for all subjects but, again, the incidence fell well within the expected ranges for this age group (Mathur 1995). Respiratory event-related arousals predominated in those subjects with SDB, as might be expected.

88 Subject Total RDI RDI Mixed/ Mixed/ RDI NREM REM central central RDI-NREM RDI-REM 1 15.4 16.3 12.7 3.8 1.9 2 4 2.2 26.1 0 0 3 34.6 34.4 36.5 0.6 0 4 28.8 24.9 46.3 0.2 0 5 1.5 0.4 10.4 0 0 6 0 0 0 0 0 7 9.4 8.8 12.4 0.5 0 8 42.9 42.1 47.4 6.3 7.1 9 27 27 27.3 0 1.8

TABLE 6.4. Respiratory disturbance indices (RDI – events/hour of sleep) overall, in non-rapid eye movement (NREM) and rapid eye movement (REM) sleep.

Subject Average Average Average Average Minimum Minimum awake desaturation saturation saturation saturation saturation saturation nadir NREM nadir REM NREM REM (%) (%) (%) (%) (%) (%)

1 96 3 95 95 89 89 2 90 2 90 90 87 87 3 91 5 90 90 82 84 4 95 3 94 94 89 89 5 98 3 98 97 95 95 6 99 0 98 n/a 95 n/a 7 93 2 92 91 89 89 8 95 5 94 92 82 81 9 97 3 96 96 90 90

TABLE 6.5. Oxygen saturation (by pulse oximetry) awake and associated with sleep-related respiratory events overall, in non-rapid eye movement (NREM) and rapid eye movement (REM) sleep.

89 Subject All arousals RDE-related All arousals RDE-related in sleep arousals in per hour of arousals time sleep time sleep per hour

1 125 87 20 14 2 69 11 27.6 4.4 3 207 169 35.4 28.9 4 174 130 33.7 25.2 5 67 3 20.7 0.9 6 12 0 15.2 0 7 82 47 15.5 8.8 8 197 179 28.2 25.6 9 133 115 30.2 26.1

TABLE 6.6. Arousal statistics, all types and respiratory disturbance event (RDE)-related, for sleep overall.

Subject All arousals All arousals RDE-related RDE-related per hour of per hour of arousals arousals NREM REM per hour per hour NREM REM

1 21.8 14.5 15.8 8.2 2 26 47 2.6 26.1 3 37 20.9 29.8 20.9 4 33.4 34.8 23.7 31.6 5 21.4 15.6 0 7.8 6 15.2 n/a 0 n/a 7 17.1 8.3 9.7 5.2 8 28.3 28.2 26 24.2 9 31.5 21.8 26.8 21.8

TABLE 6.7. Arousal statistics, all types and respiratory disturbance event (RDE)-related, for non-rapid eye movement (NREM) and rapid eye movement (REM) sleep.

As subject 1 took paracetamol 1g + codeine 60mg in the middle of his preoperative study, It was possible to examine to a limited extent the effect his consumption of opioid might have had on various aspects of his sleep and respiration. He consumed no opioid whatsoever for at least 14 hours before study commencement. Prior to the codeine, his respiratory disturbance index (RDI - events/hour of sleep) was 18.3 and

90 after the codeine it was 11.1. The minimum saturation associated with these events remained unchanged at 89%. Sleep efficiency was 81.1% before and 83.9% afterwards with sleep latencies of 19 and 24 minutes respectively. Rapid eye movement sleep (REM) made up 16.8% of pre-codeine sleep and 37.6% of codeine-affected sleep and the latency to REM was 44.5 minutes and 38.5 minutes respectively. Interestingly, between asking for the analgesics and actually getting them he slept (polygraphically proven) for a further 10 minutes, suggesting any pain he was experiencing could not have been too severe to prevent sleep, although about half of that was stage 1. A confounding effect of the pre-study diazepam cannot be excluded, although the dose was small (2mg).

Discussion

Implications: sleep

Of the previously published perioperative polysomnographic studies involving patients having procedures other than upper airway surgery (Johns 1974; Ellis 1976; Orr 1977; Kavey 1979b; Kavey 1983; Aurell 1985; Catley 1985; Lehmkuhl 1987; Knill 1990; Rosenberg 1994; Rosenberg-Adamsen 1996b; Edell-Gustafsson 1999; Cronin 2001; Rahman 2001; Drummond 2002; Wu 2003), six included no assessment of preoperative sleep (Aurell 1985; Catley 1985; Lehmkuhl 1987; Rahman 2001; Drummond 2002; Wu 2003). All three studies of cardiac surgical patients assessed sleep during the preoperative week but not the immediate preoperative night (Johns 1974; Orr 1977; Edell-Gustafsson 1999) and one of those reported wakefulness and stage 1 sleep together (Orr 1977), making interpretation difficult if not impossible, especially in the early post-operative phase. Of the 7 studies including an assessment of sleep on the preoperative night, three used non-standard polysomnographic recording and/or scoring techniques (Ellis 1976; Rosenberg 1994; Rosenberg-Adamsen 1996b), as did one of the cardiac studies (Johns 1974), and the results must therefore be viewed with caution. Several of these studies contain findings based upon comparisons between pre- and postoperative sleep and it is from this data that conclusions have been drawn regarding the significance and potential consequences of aspects such as rapid eye movement sleep (REM) rebound.

91 Only four of these studies have included monitoring of respiration other than with pulse oximetry (Catley 1985; Rahman 2001; Drummond 2002; Wu 2003) and none of these include preoperative data. Neither of the two studies recording only oximetric respiratory data involved standard polysomnographic sleep recording/scoring techniques (Rosenberg 1994; Rosenberg-Adamsen 1996b), leaving the data subject to question with respect to both its sleep stage reporting and possible artifacts in the oximetry. The seminal study by Catley et al. (Catley 1985), which included inductive plethysmography as well as oximetry, similarly involved fewer than the minimum number of channels thought to be necessary for accurate staging of sleep (Carskadon 1989). There is, however, a single meeting abstract report of six gastroplasty patients (Knill 1987), a subgroup of the patients subsequently reported without mention of respiratory monitoring in the paper by Knill et al. (Knill 1990), in which more extensive respiratory monitoring was included in the pre- and postoperative polysomnograms performed. This study, published nearly 20 years ago, used criteria for apnoeas, hypopnoeas and desaturation that are no longer considered standard, leading almost certainly to a significant underestimate of their subjects’ sleep apnoea severity. The authors stated none of their subjects had obstructive sleep apnoea (OSA). While there may not have been a history of that diagnosis, the preoperative studies on their patients with a mean body mass index greater than 50kg/m2 showed they had, on average, at least moderate OSA.

Studies using polysomnography after upper airway surgery, all involving either adenotonsillectomy or uvulopalatopharyngoplasty, have also been reported (Johnson 1986; Burgess 1992; Helfaer 1996; Terris 1996). These authors have concentrated upon aspects of sleep-disordered breathing and have either not included or only superficially reported the actual polysomnographic findings with respect to sleep. None involved a polysomnographic study in the immediate preoperative period. Helfaer et al., in their study of children with mild sleep apnoea on the night after adenotonsillectomy, did compare their single night postoperative sleep and respiratory data with studies carried out in the sleep laboratory some time prior to surgery (Helfaer 1996).

A number of studies have used actigraphic techniques to monitor pre- and postoperative sleep (Bisgaard 2002; Kain 2002; Kain 2003). Actigraphy is of no use in determining

92 specific sleep stage and its validity in the perioperative setting even for distinguishing sleep/wake state is dubious (Sadeh 2002).

With the exception of the abstract report involving the six relatively young but morbidly obese patients, this study is, therefore, the first to examine preoperative sleep using full polysomnography including monitoring of airflow and thoraco-abdominal inductive plethysmography. It is also the first study to examine polygraphically the sleep of a group of middle-aged to elderly surgical in-patients with, in some cases, significant intercurrent illnesses. This was not specifically by design and occurred as a result of the case-mix of our institution combined with a policy of day-of-surgery admission for lower-risk patients.

The wide variation in sleep efficiency is a feature of most preoperative studies and almost certainly involved a number of factors that are, at this stage, impossible to identify as single causative agents. These include prior hospital experience, differences in anxiety, preoperative acute surgical conditions with or without pain, preoperative chronic and acute medications, environmental and seasonal conditions in the ward such as temperature or noise and, very likely, different attitudes and responses to the sleep monitoring itself. No patient required any specific observations to be performed overnight so this should not have been a factor. However, other patients in the room may have required such observations so disturbance by light and nursing activity are very possible problems.

The variability in both sleep efficiency and total sleep time for preoperative nights is very similar across nearly all studies (Ellis 1976; Kavey 1979b; Kavey 1983; Knill 1990; Rosenberg 1994; Edell-Gustafsson 1999; Cronin 2001) including this one when the incomplete nights of subjects 5 and 6 are excluded. Even without those incomplete studies, the mean preoperative total sleep time of 312 minutes and efficiency of 67% are less than that found in most other studies. This is almost certainly due, at least in part, to the age of the subjects, as the sleep variables in this study are largely consistent with epidemiological findings in the elderly, both in terms of mean values and variation (Bliwise 1989; Ancoli-Israel 1994).

93 Both subjects (1 and 8) with efficiencies greater than 80% had spent at least one night in hospital prior to their preoperative night and so may have been more accustomed to the hospital environment. Subject 5, however, had also spent many weeks in hospital prior to his surgery and yet his was amongst the worst nights’ sleep. Subjects 1 and 8 both also had considerable prior experience of major abdominal surgery and therefore may have been less anxious about having surgery the following day. Subject 1, however, was experiencing enough ischaemic leg pain to request opioid analgesia in the middle of the night. It is hard to reconcile this with the fact that he slept better than almost anyone else in the study. He did have diazepam but the chronicity of his use of this drug suggests its contribution should have been minimal.

Another aspect of particular interest with subject 1 was the architecture of his opioid- affected sleep. It is widely accepted that µ-receptor agonists inhibit REM (Cronin 1995) and this is one of the main reasons put forward for the reduction in REM in the early postoperative period, pain being the other presumptive main cause. After receiving a substantial dose of one such agent (codeine 60mg), however, subject 1 experienced more than twice as much REM, percentage-wise, than he had in the several hours of sleep prior to its administration. It is possible he is one of the 8-10% of caucasians lacking in cytochrome CYP2D6, necessary for the O-demethylation of codeine to morphine. As codeine itself has only weak affinity for the µ-receptor, its metabolism to morphine may be necessary not only for its analgesic efficacy but also for its effect on REM sleep. This is, however, a speculative and unlikely cause for the observation. Considering this in association with the findings of Cronin and colleagues, that REM was virtually eliminated in the early postoperative setting despite good control of pain using regional techniques both with and without opioid (Cronin 2001), one must consider the possibility that another factor is the main cause of REM suppression after major surgery.

The effect of environmental conditions is very difficult to determine. One potentially important factor in this study was the overnight minimum temperature of 21.7oC after a hot day experienced by subject 6. This may have been a factor in his particularly poor sleep in a room with 3 other men, poor ventilation and no air-conditioning. Under those conditions, the respiratory plethysmography bands and various other electrodes taped to

94 his skin would have been quite uncomfortable, potentially contributing to both his very long sleep latency and his withdrawal from the study in the middle of the night. On the other hand, subject 7 experienced similar conditions for both nights he was studied and this did not seem to affect his sleep to the same degree. Other studies have attempted to control the environment somewhat by placing the patients in single rooms (Kavey 1979b; Kavey 1983; Knill 1990; Rosenberg 1994; Cronin 2001) or even at home for one subgroup (Edell-Gustafsson 1999) but none mention whether or not the temperature was controlled. It is likely that adverse climatic conditions, or at least ones to which the patients are unaccustomed, play a significant role in the disturbance of sleep both before and after surgery as they do at any time.

The possible observation of increased SWS in subjects with chronic respiratory disease would need to be confirmed with larger studies as this does not appear to have been investigated to date. It is possible that this could be a protective mechanism, SWS being that stage in which obstructive hypoxaemic events tend to occur least, despite fairly profound muscle relaxation (Series 2002; Ayappa 2003) (see also chapter 3). Another possible explanation is hypercapnia. Elevated end-tidal CO2 during emergence from anaesthesia is sometimes associated with an increase in delta EEG activity (personal communication, Dr Christopher Thompson, neuroanaesthetist, Royal Prince Alfred Hospital). Could a similar phenomenon occur during sleep?

Implications: breathing

Some of the subjects included in this study were mildly overweight but none were markedly obese, nor did any of them have any particular feature in their history or examination suggestive of SDB. Certainly none were suspected of having SDB prior to their involvement in this investigation. The high incidence of SDB amongst this group was, initially at least, somewhat surprising, especially when two were found to have quite severe sleep apnoea. While it subsequently became clear that these results are fairly consistent with population estimates for this age group, this is almost certainly a finding of considerable interest to anaesthetists and other physicians and health workers involved in the perioperative care of patients. There is controversy about the perioperative management of patients with sleep apnoea and now it is quite clear that large numbers of patients with various other coincident risk factors are presenting for

95 surgery with undiagnosed and often quite severe SDB. This has a number of significant implications.

To date it has been assumed that sleep apnoea is a significant risk factor for postoperative morbidity and mortality and this conclusion is a fairly logical one to draw. Various anecdotal reports (Reeder 1991a; VanDercar 1991; Ostermeier 1997; Cullen 2001) have led to some rather grave warnings about the level of this perioperative risk (Boushra 1996; Parikh 2002). The data from this study, especially when combined with other epidemiological data (Young 2002), suggest a reconsideration of this risk. Approximately 10% of the population of developed nations undergo surgery each year and elderly patients are well represented in that group. Very large numbers of individuals with significant sleep apnoea obviously undergo surgery without significant morbidity or mortality and without their SDB ever being recognised. This suggests the overall risk is actually quite low, an implication supported by the finding that sudden and unexpected postoperative deaths are rare (see chapter 4).

On the other hand, the anecdotes about perioperative patients already mentioned (Reeder 1991a; VanDercar 1991; Ostermeier 1997; Cullen 2001) do suggest that SDB plays some role in adverse perioperative events. It is possible that a combination of factors is involved when morbidity occurs. One such additional factor is almost certainly obesity but this has been inadequately investigated.

It is possible that some of the hypopnoeic events identified in the study of subject 3 were the result of Cheyne-Stokes breathing due to his cardiac failure. Paradoxical thoraco-abdominal movement was a feature of most of his events, however, and there were almost no events during slow wave sleep. This suggests obstruction to be the primary cause of his SDB but does not rule out a central component. Subject 8 also had a significant mixed/central component to his SDB but, again, it was predominantly obstructive in nature. Obstruction was similarly the cause of the overwhelming majority of events for all other subjects in this study who had SDB. This is consistent with epidemiological data concerning the relative prevalence of obstructive and central sleep apnoea/hypopnoea syndromes. Clearly most attention needs to be paid to the issues concerning perioperative hypoxaemia caused by obstructive events. Nevertheless, apnoea of central origin, while less common, almost certainly points to concurrent

96 cardio-respiratory or neurological disorders that may have inordinate impact on perioperative management.

An interesting feature of the period I described as “sleep/wake transition time” was the degree of respiratory instability experienced by a number of subjects. Periodic breathing and partial obstruction began to occur well in advance of EEG sleep onset, with various signs normally associated with arousal from sleep accompanying termination of the respiratory events. In other words, it appeared that ‘sleep’-disordered breathing commenced during many epochs that had to be scored as wakefulness using standard criteria. As a result, these respiratory events are not included in the RDI totals recorded by the sleep software and so the severity of the SDB may have been understated. In some cases the degree of respiratory disturbance was greater during this phase than in other stages of sleep. Ventilatory instability during sleep onset is another phenomenon that has been recognised in the sleep literature (Dunai 1996; Dunai 1999) for some time and more recent work suggests that the associated fluctuations in arousal may, themselves, contribute to the respiratory instability (Younes 2004). This phenomenon may explain some of the postoperative respiratory events described in the recent studies by Drummond et al as occurring in wakefulness (Rahman 2001; Wu 2003). Moreover, it may have substantial implications for the effect of opioids and/or pain on respiratory function. If this type of ‘transitional sleep’ or drowsiness is present for large parts of the early postoperative period, as several previous studies of postoperative sleep suggest it might be, then this could well be a significant factor in the exacerbation of respiratory disturbance, already well recognised in this setting. This phenomenon is potentially very important and should prove to be a fertile research field.

It is of further interest that subject 1 experienced fewer episodes of sleep-related respiratory compromise per hour after the administration of codeine. His SDB was NREM stages 1 and 2 predominant and as his codeine-affected sleep contained a greater proportion of REM this is one possible reason for the reduction. This provides further evidence that the usual assumption of REM predominance of SDB and the consequent increase in SDB assumed to occur as a result of postoperative REM rebound are both likely to be incorrect in many cases (see also chapter 3). With patients such as subject 1, the opposite seems to be the case.

97 Conclusions

The quality or efficiency of pre-operative sleep is highly variable between patients. Interaction between a large number of internal and environmental factors is likely to be the cause. The effect of the sleep monitoring equipment itself is perhaps more important in some patients and this needs to be considered when assessing previous studies as well as during the design of future research into perioperative sleep. Normal values and ranges for the age groups examined also need to be borne in mind when assessing findings on perioperative sleep. The role of opioids and pain in the perioperative suppression of REM sleep needs to be re-examined. Similarly, the finding of an increase in slow wave sleep amongst chronically hypoxaemic subjects warrants further scrutiny.

The most important finding of this study was the remarkably high incidence of significant and undiagnosed sleep apnoea amongst a group of non-obese, middle aged to elderly patients presenting for surgery. While this would appear to be in keeping with prevalence studies for this age group, it is a finding that is almost certainly surprising to anaesthesia personnel, considering the current literature. The finding in one subject of a reduction in the severity of sleep apnoea after opioid administration supports my previous conclusion that the role of sleep architecture in the genesis of OSA is more complex than previously assumed in the anaesthesia literature. This is especially so when the role of the transition from wakefulness to sleep on respiratory control is considered along with the potential effect of perioperative factors, especially opioids, on this transitional phase, an issue that also warrants examination in greater detail.

98 CHAPTER 7

POSTOPERATIVE SLEEP

Various polysomnographic techniques have been used to assess postoperative sleep architecture after major non-cardiac (Ellis 1976; Kavey 1979b; Kavey 1983; Aurell 1985; Catley 1985; Lehmkuhl 1987; Knill 1990; Rosenberg 1994; Cronin 2001; Rahman 2001; Drummond 2002; Wu 2003), cardiac (Johns 1974; Orr 1977; Edell- Gustafsson 1999) and more minor forms (Rosenberg-Adamsen 1996b) of surgery including that of the upper airway (Johnson 1986; Burgess 1992; Helfaer 1996; Terris 1996). All of these studies to date have relied on an important and potentially flawed assumption – that sleep may be reliably defined by its electrophysiology.

The electroencephalogram (EEG) in particular, and with it the electro-oculogram (EOG) and electromyogram (EMG), are routinely used as markers of sleep under normal (or almost normal) sleeping conditions. The standardised and somewhat arbitrary classification of EEG sleep stages according to Rechtschaffen and Kales (Rechtschaffen 1968) has been derived from large numbers of behavioural and electrophysiological observations of normal sleep in both humans and animals. While it might be convenient to assume that these markers of sleep are meaningful in the perioperative setting, and this is indeed what most investigators to date appear to have done, there is in fact no evidence to support this assumption.

It would require, as it did for the classification of normal sleep, very large numbers of concurrent behavioural and electrophysiological observations to accurately characterise postoperative sleep. Nevertheless, this present study was carried out in an effort to gain further insight into the macro- and micro-architecture of sleep and the possible differences between the way ‘normal’ and postoperative sleep should be electrophysiologically examined. This has significant implications for the interpretation of previous and future studies.

99 Methods

Three subsets of patients were included. Patients commencing study preoperatively were recruited from the routine operating lists of the Royal Prince Alfred Hospital, Camperdown, Australia, with no particular prerequisite conditions or exclusions. The decision to approach these patients was made solely from the operating list with no knowledge of the patients’ prior medical histories. Several were suffering acute conditions for which they required semi-urgent surgery. A second group of subjects were recruited preoperatively under the same conditions (prior history unknown, no suspicion of sleep disorders) but did not have a study in the immediate preoperative period. The third group included a number of patients who were recruited postoperatively, mostly on the basis of the behaviour of their airways during the immediate surgical period. None of these had a prior diagnosis of sleep disordered breathing but in several cases there was reason to suspect it.

All subjects were accommodated in a normal surgical ward environment, mostly in multiple-bed rooms, as are the majority of patients in our institution. The method of both collecting and examining the polysomnographic data is described in chapter 5.

Definitions

With additional consideration of possible differences in the nature of postoperative sleep electrophysiology as discussed below, sleep was manually staged in all cases according to the criteria of Rechtschaffen and Kales (Rechtschaffen 1968), with regard for the affects of age and sleep pathology as recommended by Bliwise, Carskadon and Rechtschaffen (Bliwise 1989; Carskadon 1989). Standard definitions and criteria (ASDA 1992; AASM 1999) were similarly used for the scoring of arousals and respiratory events and while these aspects of postoperative sleep are examined in greater detail in the following chapters their occurrence was given due consideration in the staging of sleep itself.

100 Statistical Analysis

Summary statistics are presented as mean ± standard deviation unless otherwise indicated. Between-night comparisons were performed using two-tailed Wilcoxon signed rank test (Prism 3.0a for Macintosh, Graphpad Software Inc., San Diego, California, USA) with P < 0.05 considered significant.

Results

As indicated in chapter 6, nine of the 19 subjects had a study on the preoperative night and none of these had a prior diagnosis or history suggesting sleep disorders or sleep- disordered breathing. The demographics and medical histories of these subjects have already been described (see chapter 6, table 6.1).

Two subjects were recruited pre-operatively but not studied until the first post-operative night. This subset included subject 10 who admitted after recruitment that she had previously had a diagnostic sleep study that confirmed severe obstructive sleep apnoea. She had declined a trial of therapy at that time. Bowel preparation precluded immediate pre-operative study. Her demographics and medical history are given in table 7.1. Subject 11, a 72-year-old male, was studied on the night immediately after anterior resection of the colon, bowel preparation also being the reason for no pre-operative study in his case. The post-operative study was a technical failure and no useful information could be obtained from the data whatsoever. Furthermore, intervention by nursing staff resulted in his refusal to take any subsequent part in the study. No further details are therefore included for this patient. Shortly after rising on the ninth postoperative morning, he suffered a cardiac arrest and died.

Subject 12, a 52-year-old, 115kg male having a percutaneous nephro-lithotomy, was recruited after he proved to be extremely difficult to intubate. Congenitally blind, his EOG signals were non-existent, making the study very difficult to interpret. Regrettably, initial doubts about the utility of his study led me to neglect satisfactory back-up of his polysomnographic data and a computer hard-drive failure shortly

101 thereafter resulted in its irretrievable loss. No further details are therefore included for this subject.

Demographic and medical details for the remaining 7 subjects, all recruited postoperatively, are also given in table 7.1. Subject 15 was a day-of-surgery admission recruited immediately after surgery on the basis that he was somewhat difficult to intubate endotracheally but he declined study until the second postoperative night. The others were all approached after they were found to have obstruction-prone airways under general anaesthesia.

Subjects 13 - 17 were admitted on the day of surgery, the remainder admitted to hospital for various periods prior to their surgery. The relevant lengths of preoperative stay for subjects 1 - 9 have been discussed in chapter 6. Subjects 10, 18 and 19 were admitted the day prior to surgery, in all cases for bowel preparation.

102 Subject Age/Sex Weight Height Patients History Concurrent (kg) (m) in room medications

10 66/F 68 1.59 1 Ischaemic heart disease. insulin, ranitidine, Insulin dependent diabetes. cisapride, verapamil, Prior frontal craniotomy prednisone, methotrexate for meningioma. OSA on previous sleep study. Rheumatoid arthritis.

13 55/M 120 1.8 4 Melanoma, otherwise Nil well. 14 64/M 110 1.73 4 Hypertension and spironolactone, ischaemic heart disease. irbesartan, nitrates, lisinopril, atorvastatin, alpha-methyldopa 15 49/M 88 1.78 4 Melanoma, otherwise Nil well. 16 54/M 109 1.77 4 Melanoma. Tracheal fluoxetine narrowing on X-ray, heavy smoker. 17 76/M 86 1.73 4 Melanoma. Atrial warfarin, rabeprazole, fibrillation, oesophageal digoxin, glimepiride, reflux, non-insulin allopurinol dependent diabetes. 18 59/M 88 n/a 4 Prostatic carcinoma. allopurinol, atorvastatin, Hypertension and non- metformin, glibenclamide, insulin dependent diabetes. perindopril

19 65/M n/a n/a 4 Prostatic carcinoma. enalapril, ranitidine Hypertension.

TABLE 7.1. Demographics and medical histories for subjects commencing study postoperatively. Studies for subjects 11 and 12 were technical failures hence their exclusion (note: demographics for subjects 1-9 are given in table 6.1). n/a = not available.

The number of patients in the rooms was recorded for each subject and these are also shown in tables 6.1 and 7.1. The majority of subjects (1 - 3, 8, 10, 13 - 19) were accommodated in air-conditioned wards while the rest were in older wards without air- conditioning. As mentioned in chapter 6, the possible effect of room temperature was not considered until after data collection was complete so it was not contemporaneously recorded. Nevertheless, daily minimum and maximum temperatures for Sydney were subsequently obtained for all study nights from the Australian Bureau of Meteorology (ABOM 1999). The only postoperative recording that might have been adversely affected is the single study of subject 7 on his immediate postoperative night. After a daytime maximum of 28.8 oC, the overnight minimum of 21.0 oC might have been

103 associated with humid and uncomfortable conditions in his ward. Subject 9 refused to be studied on his immediate postoperative night. The daytime maximum of 37.9 oC was almost certainly a factor in his refusal. None of the subjects suffered significant fevers on any of the nights studied.

The preoperative surgical histories and their potential effect on the preoperative studies of subjects 1-9 were noted in chapter 6. Of the subsequent subjects, only subject 10 had a significant surgical history, having undergone a craniotomy to excise a frontal meningioma.

Subject Surgery Performed: Type of Anaesthesia 1 Embolectomy and revision right femoro-popliteal bypass: general anaesthetic 2 Left inguinal hernia repair (open): spinal anaesthetic 4 Open prostatectomy: general anaesthetic 5 Right ureterectomy (via midline laparotomy): general anaesthetic 7 Transurethral resection of prostate: spinal anaesthetic 8 Endoluminal abdominal aortic aneurysm repair: combined epidural and general anaesthetic 9 Left inguinal lymph node dissection for metastatic melanoma: general anaesthetic 10 Laparoscopic-assisted anterior resection of rectum 13 Left inguinal lymph node dissection: general anaesthetic 14 Excision of left retroperitoneal and thigh tumours: general anaesthetic 15 Right ilio-inguinal lymph node dissection: general anaesthetic 16 Right retroperitoneal lymph node dissection: general anaesthetic 17 Right axillary lymph node dissection: general anaesthetic 18 Prostate brachytherapy: general anaesthetic 19 Prostate brachytherapy: general anaesthetic

TABLE 7.2. Surgery performed for each subject having postoperative studies. Prostate brachytherapy involves insertion under anaesthesia of 12 or more catheters into the prostate via a template sutured to the perineum. These remain in situ for 36 hours (three doses of radiation).

104 Subject Nights Studied Preop Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7 1    2   4       5     7   8   9    10   13  14  15  16  17   18   19  

TABLE 7.3. Perioperative nights studied for each subject. Preop = preoperative night; Postop = postoperative night;  indicates night studied.

The surgical procedure and type of anaesthesia used for each subject is listed in table 7.2. As subjects 3 and 6 withdrew from the study after the preoperative night they, along with subjects 11 and 12, are excluded from this and subsequent data tables. Table 7.3 indicates the specific postoperative nights studied for each subject.

All subjects were administered medications prior to or during at least one of their postoperative studies that may have affected the results (Bauer 1993). Doses, routes and times of administration of all drugs thought to be relevant are listed in table 7.4. With the exception of those patients having abdominal surgery who would have been nil-by- mouth for one or two days, most patients continued their regular medications and received oral paracetamol 1g four times daily.

105 Subject Postop 1 Postop 2 Postop 3

1 codeine 60mg at 2030, n/a diazepam 2mg at 2200 diazepam 2mg at 2200 2 n/a codeine 16mg at 2130, n/a theophylline* 4 pethidine 165mg periop codeine 60mg at 1200, codeine 60mg at 1200, (12 hrs before), codeine 1800 and 2200 1800 and 2200 60mg at 1800, pethidine 75mg at 0500 5 pethidine 155mg periop oxycodone 10mg at 1115, temazepam 10mg at 2100 (12 hrs before), pethidine oxycodone 10mg at 1700, 75mg at 1815, droperidol pethidine 75mg at 0115, 0.25mg at 1815, droperidol 0.25mg at oxycodone 10mg at 0015 0115 7 codeine 60mg at 2100*, n/a n/a clonazepam 500mcg at 2200, theophylline* 8 papaveretum 20mg n/a n/a premed (12 hrs before), epidural fentanyl 50mcg/hr throughout study 9 n/a dextropropoxyphene 65mg dextropropoxyphene 65mg at 1800, at 1130 dextropropoxyphene 65mg at 2200 10 morphine 21mg periop, Nil after 1000 (50mg n/a PCA morphine 16mg to total morphine use to 2200, nil to 0200, 3mg to then) 0400, 3mg to 0600 13 n/a n/a n/a 14 n/a n/a oxycodone 10mg at 1215, alpha methyldopa* 15 n/a morphine 10mg at 1725, n/a oxycodone 10mg at 2145 16 n/a morphine 80mg and n/a oxycodone 20mg from 0210-2130, morphine 15mg at 0210, fluoxetine 20mg daily 17 morphine 10mg intraop n/a Nil (12 hrs before) 18 morphine 7.5mg at 1120, Nil n/a oxycodone 10mg at 1445 19 morphine 6mg introp (12 oxycodone 10mg at 1300 n/a hrs before), oxycodone 10mg at 2330

TABLE 7.4 (part 1).

106 Subject Postop 4 Postop 5 Postop 7

1 n/a n/a n/a 2 n/a n/a n/a 4 codeine 16mg q.i.d. codeine 16mg q.i.d. n/a 5 n/a n/a n/a 7 n/a n/a n/a 8 n/a n/a n/a 9 n/a n/a n/a 10 n/a n/a n/a 13 n/a n/a codeine 60mg at 1400, codeine 60mg at 0130 14 n/a n/a n/a 15 n/a n/a n/a 16 n/a n/a n/a 17 n/a n/a n/a 18 n/a n/a n/a 19 n/a n/a n/a

TABLE 7.4 (continued). Medications administered before or during recordings that may have affected the study. Postop = postoperative night; n/a = study not performed; PCA = patient controlled analgesia; q.i.d. = four times a day; * indicates chronically used agent.

A full night study (more than seven hours recording) was achieved for the majority of post-operative studies, termination mostly occurring when the memory card became full. Subject 5 removed the electrodes after 274 minutes of recording on his third postoperative night, having completed two full postoperative nights and a partial preoperative study prior to this. Subject 14 removed his electrodes after 208 minutes on the only night he was studied (postoperative night 3). Subject 16 removed the nasal flow transducer 95 minutes into his single night 2 study, leaving the rest of his electrodes intact for the remainder of the night. Subject 17 removed his leads after 262 minutes on his second postoperative night. Report times (duration of recording from recording commencement to cessation or removal of electrodes) are listed for all postoperative nights in table 7.5.

107 Subject Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7 1 503 518 2 545 4 559.5 460 487.5 487.5 441 5 559.5 559.5 274.5 7 517.5 8 517.5 9 454 448.5 10 517.5 517.5 13 423 14 208 15 450 16 451 17 451.5 262.5 18 451.5 451.5 19 451.5 451.5

TABLE 7.5. Report times (minutes) for all postoperative nights studied. Postop = postoperative night; space = study not performed.

Sleep latencies for all postoperative nights studied are listed in table 7.6. For the majority of studies, the latency is effectively the time from set-up completion to sleep onset, as recording was commenced manually and the subjects were left in a “lights-off” situation. Subjects 17-19 differed in that their recording was programmed to commence at 2200 hours for subjects 17 and 18 and at 2230 for subject 19, set-up having been completed earlier in the evening than usual. All of these subjects indicated a desire to watch television for some time before lights out and the commencement time reflected the finish of the program they wanted to watch. On two nights, this resulted in recording commencement after sleep onset. Particularly on early postoperative nights, subjects were frequently already asleep and had to be woken for the recording set-up. This is reflected in the relatively short latencies of some patients. Subjects 1 and 7 both required nursing intervention after study commencement on the first postoperative night, including the administration of regular evening medication, and this almost certainly contributed to their longer latencies.

108 Subject Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7 1 48 20 2 84 4 7 21.5 69.5 26 100.5 5 12.5 78.5 27 7 98 8 17.5 9 79 149 10 12 40 13 26 14 24.5 15 21.5 16 2.5 17* 14 0 18* 94 1.5 19* 8 0

TABLE 7.6. Sleep latencies (minutes) for postoperative nights studied. Postop = postoperative night; space = study not performed; * indicates results possibly affected by pre-programmed start time.

Sleep period, the time from first sleep onset to final awakening or study termination, is listed for all postoperative studies in table 7.7. Total sleep time (TST), the time spent asleep during the report or recording time, is shown in table 7.8. There was wide variation in recording time as noted in table 7.5, a few studies ceasing prematurely and even complete ones varying by up to an hour, and this renders between-night comparison of TST inappropriate.

109 Subject Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7 1 454 497 2 436.5 4 552.5 434.5 416 459 291 5 538.5 481 188.5 7 415 8 500 9 364 297 10 505.5 396 13 395.5 14 182 15 428.5 16 449 17 437.5 260.5 18 357.5 450 19 443.5 451

TABLE 7.7. Sleep period, the time in minutes from first sleep onset to final awakening or study termination, for postoperative nights studied. Postop = postoperative night; space = study not performed.

Subject Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7 1 409 446 2 156.5 4 338 332.5 206.5 327 260 5 236.5 143.5 173 7 345.5 8 440.5 9 234 204.5 10 387 380 13 304.5 14 138 15 302.5 16 168.5 17 390.5 207.5 18 213 408.5 19 265 337

TABLE 7.8. Total sleep time, the time in minutes asleep during the recording period, for postoperative nights studied. Postop = postoperative night; space = study not performed.

110 Sleep efficiency, total sleep as a percentage of recording time, is given in table 7.9. Overall, excluding subjects with incomplete studies, sleep efficiency on the first postoperative night was 67 ± 16% while the preoperative equivalent mean was 67 ± 19% (non-identical groups). When the efficiencies of subjects studied on both nights were compared (n = 5) there appeared to be a small reduction in the mean from 73 ± 12% to 67 ± 17% but this failed to reach statistical significance (P = 0.06). There appeared to be an overall tendency towards a reduction in efficiency on the second and third nights (58 ± 23% and 58 ± 24% respectively) but this did not reach statistical significance either when paired samples were examined (P = 0.38 and 0.63 respectively, n = 4 for both nights).

Subject Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7 1 81.3 86.1 2 28.7 4 60.4 72.3 42.4 67.1 59 5 42.3 25.6 63 7 66.8 8 85.1 9 51.5 45.6 10 74.8 73.4 13 72 14 66.3 15 67.2 16 37.3 17 86.5 79 18 47.2 90.5 19 58.7 74.6

TABLE 7.9. Sleep efficiency, total sleep time as a percentage of report time, for postoperative nights studied. Postop = postoperative night; space = study not performed.

Subjects gave many reasons for disrupted sleep when questioned. Some of these are notable. Subject 4 had his worst night, efficiency-wise, on postoperative night 3. This was the result of the admission in the early hours of the morning, to the other bed in his 2-bed room, of a very sick patient requiring considerable nursing and medical attention. Subject 18 was similarly disturbed on his first postoperative night by the terrible stridor

111 of another patient in his 4-bed room, suffering from an advanced upper airway tumour. Interestingly, subject 18 also complained after his second night that he’d been kept awake by the fellow in the bed next to him who, he claimed, watched television all night. Despite this, he still managed a sleep efficiency of 90% with relatively normal architecture!

Sleep stage information for all postoperative studies is recorded in tables 7.10 – 7.16.

Subject Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7 1 10.5 7.8 2 22 4 28 17.3 11.9 12.1 9.1 5 54.1 30 6.1 7 7.4 8 21.8 9 26.9 11 10 42.5 3.9 13 3.6 14 12.7 15 43.3 16 95.3 17 9.3 14.7 18 16.7 6.1 19 52.8 28.6

TABLE 7.10. Stage 1 sleep as a percentage of total sleep time for postoperative nights studied. Postop = postoperative night; space = study not performed.

112 Subject Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7 1 64.1 61.7 2 55.6 4 69.2 62 51.6 52.3 51.7 5 42.1 69.3 48.6 7 49.1 8 71.4 9 59.6 67 10 50.6 69.7 13 70.1 14 35.9 15 56.2 16 4.7 17 79.6 69.4 18 82.9 49.7 19 34.2 33.4

TABLE 7.11. Stage 2 sleep as a percentage of total sleep time for postoperative nights studied. Postop = postoperative night; space = study not performed.

Subject Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7 1 4.5 7.2 2 13.7 4 0.4 2.6 10.2 11.0 13.1 5 0.0 0.0 32.9 7 18.5 8 0.0 9 1.9 7.6 10 0.1 8.0 13 4.4 14 27.9 15 0.0 16 0.0 17 1.2 3.6 18 0.5 21.2 19 10.0 8.9

TABLE 7.12. Stage 3 sleep as a percentage of total sleep time for postoperative nights studied. Postop = postoperative night; space = study not performed.

113 Subject Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7 1 0.0 1.1 2 1.6 4 0.0 0.0 0.5 0.5 1.2 5 0.0 0.0 3.8 7 4.9 8 0.0 9 3.2 8.6 10 0.0 0.0 13 2.6 14 14.1 15 0.0 16 0.0 17 0.0 0.0 18 0.0 8.8 19 3.0 17.4

TABLE 7.13. Stage 4 sleep as a percentage of total sleep time for postoperative nights studied. Postop = postoperative night; space = study not performed.

Subject Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7 1 4.5 8.3 2 15.3 4 0.4 2.6 10.7 11.5 14.3 5 0.0 0.0 36.7 7 23.4 8 0.0 9 5.1 16.2 10 0.1 8.0 13 7.0 14 42.0 15 0.0 16 0.0 17 1.2 3.6 18 0.5 30.0 19 13.0 26.3

TABLE 7.14. Slow wave sleep (stages 3 and 4 together) as a percentage of total sleep time for postoperative nights studied. Postop = postoperative night; space = study not performed.

114 Subject Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7 1 79.1 77.8 2 93.0 4 97.6 81.8 74.1 75.8 75.2 5 96.2 99.3 91.3 7 79.9 8 93.2 9 91.7 94.1 10 93.3 81.7 13 80.8 14 90.6 15 99.5 16 100 17 90.1 87.7 18 100 85.8 19 100 88.3

TABLE 7.15. Non-REM sleep (stages 1-4 together) as a percentage of total sleep time for postoperative nights studied. Postop = postoperative night; space = study not performed.

Subject Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7 1 20.9 22.2 2 7.0 4 2.4 18.2 25.9 24.2 24.8 5 3.8 0.7 8.7 7 20.1 8 6.8 9 8.3 5.9 10 6.7 18.3 13 19.2 14 9.4 15 0.5 16 0.0 17 9.9 12.3 18 0.0 14.2 19 0.0 11.7

TABLE 7.16. REM sleep as a percentage of total sleep time for postoperative nights studied. Postop = postoperative night; space = study not performed.

115 Stage 1 sleep made up 10.0 ± 5.0% of all-night preoperative studies. This more than doubled overall for the first (27.0 ± 18.5%) and second (30.4 ± 27.2%) postoperative nights (non-identical groups), with much greater variability. The latter was markedly skewed by the night of subject 16 who remained in stage 1 for nearly all of his recorded sleep. Comparing subjects with studies on both preoperative and first postoperative nights (n = 5 and excluding subject 16 who was not studied preoperatively) the mean stage 1 was still more than double preoperative levels (9.1 ± 3.5%) on night 1 (24.4 ± 16.6%). This was not, however, statistically significant (P = 0.13). Statistical comparisons of paired studies for night 2 (24.1 ± 5.6%) and night 3 (9.2 ± 2.7%) against the preoperative night were similarly not significant (P = 0.13 and 0.63 respectively) Subsequent nights (single studies) all had similar amounts of stage 1 sleep to the preoperative mean.

The average percentage of stage 2 sleep, whether or not incomplete studies were included, was remarkably constant across all perioperative nights, although the variability was much more marked on postoperative nights 1 and 2, subject 16 again accounting for much of the variability on the second night. None of first (P = 0.81, n = 5), second (P = 0.88, n = 4) or third (P ≤ 1, n = 4) postoperative nights differed significantly from the preoperative night in their amounts of stage 2 sleep.

For comparative purposes, stages 3 and 4 are considered together as slow wave sleep (SWS). Overall (non-matched groups), there was a marked reduction in SWS from the preoperative night (18.6 ± 7.0%) to the first postoperative night (4.8 ± 8.2%), with many subjects having no SWS at all. This recovered to some extent on nights 2 (9.7 ± 11.6%) and 3 (11.7 ± 4.1%). Some subjects had no SWS on night 2 as well as night one but there was insufficient data to assess the possibility of a subsequent rebound for those subjects. Similarly there were others who appeared to have a rebound on night 2 although the latter did not have preoperative studies for individual comparison. Considering matched pairs only, when compared statistically with the preoperative night, none of postoperative nights 1 (P = 0.06, n = 5), 2 (P = 0.13, n = 4) or 3 (P = 0.63, n = 4) differed significantly in their amounts of SWS.

The effect of surgery on REM sleep in this study was not as marked as that on SWS. While it appeared to be reduced overall, (non-matched groups, complete studies) from

116 preoperative levels (14.8 ± 5.6%) on both postoperative night 1 (7.8 ± 7.9%) and on night 2 (8.8 ± 7.3%), neither of these reached statistical significance when matched nights were compared (P = 0.13, n=5 and P = 0.13, n = 4 respectively). A questionable overall rebound on night 3 to 18.0 ± 10.6% was also not significant (P = 0.88, n = 4). There was only one subject (4) who had enough nights to examine the individual presence of a REM rebound and it would appear in his case that it did occur. It must be noted, however, that he had essentially identical fractional amounts of REM for all of his last three postoperative studies (nights 3 – 5), suggesting that his preoperative night might have contained less REM than usual in his case. On the other hand, his third postoperative night study, where his REM recovered, contained one brief episode of sleep where he progressed directly from stage 1 to REM, suggesting a degree of REM deficit with some “urgency” for its recovery.

In the period of ‘wakefulness’ prior to any electroencephalographic indication of stage 1 sleep, most subjects exhibited other signs suggestive of imminent sleep onset, as noted in many of the preoperative studies, although few patients had any significant incidence prior to surgery. These signs included reduction of EMG amplitude, changes in respiratory pattern and, particularly, slow rolling eye movements. The total duration of this phenomenon is recorded for each postoperative night as ‘wake/sleep transition time’ in table 7.17 and the percentage of this time with respect to total recording time is shown in table 7.18 for all nights recorded. While neither reached statistical significance in a matched-pair comparison (P = 0.13, n = 5 and P = 0.13, n = 4 respectively) it appeared that a greater percentage of the first (12.4 ± 9.3%) and second (14 ± 11.9%) postoperative nights were spent in this state of “drowsiness” than the preoperative night (4.4 ± 4.1%). There was a subsequent reduction back towards preoperative levels on the third night (7.4 ± 6.2%).

117 Subject Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7 1 0.0 0.0 2 59.0 4 75.5 43.0 59.5 56.0 24.0 5 121.0 146.0 3.5 7 37.5 8 21.5 9 78.5 50.0 10 63.5 9.5 13 4.5 14 26.0 15 32.0 16 187.0 17 25.0 24.5 18 82.0 18.0 19 130.5 57.5

TABLE 7.17. “Wake/sleep transition” time (see text) in minutes for postoperative nights studied. Postop = postoperative night; space = study not performed.

Subject Preop Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7 1 0.2 0.0 0.0 2 7.0 10.8 3 5.8 4 7.9 13.5 9.3 12.2 11.5 5.4 5 1.8 21.6 26.1 1.2 6 0.7 7 0.2 7.2 8 4.2 4.2 9 12.1 17.3 11.1 10 12.3 1.8 13 1.1 14 12.5 15 7.1 16 41.5 17 5.5 9.3 18 18.2 4.0 19 28.9 12.7

TABLE 7.18. “Wake/sleep transition” time (see text) as a percentage of report time for all nights studied. Preop = preoperative night; Postop = postoperative night; space = study not performed.

118 For several subjects during the early postoperative phase, where pain and/or opioid administration were at their maximum, there were many epochs that contained features normally associated with wakefulness. These included the predominance of alpha (8 – 12Hz), sigma (12 – 16Hz) and beta (16 – 32Hz) frequencies in the EEG as well as moderately high EMG activity. Other features such as respiratory obstruction, however, suggested the subject was actually asleep. The occurrence of K-complexes and spindles in the EEG, along with the respiratory events, were taken as confirmation of sleep, despite the high frequency background. Figure 7.1 is an example of stage 2 sleep from the preoperative night of subject 5, containing a K-complex followed a few seconds later by an obvious spindle. Note the low-amplitude EMG and the mixed-frequency background of the zoomed EEG. Figure 7.2 is an epoch from the first postoperative night of the same subject, after a midline laparotomy, when he would have been experiencing considerable pain. It is notable for its higher EMG tone (compare with figure 7.1) and the clear predominance of sinusoidal alpha rhythm in the zoomed EEG. It is nevertheless recognisable as stage 2 sleep by the K-complex present. Figure 7.3 shows an epoch from the second postoperative night of subject 5. Again, the EMG tone is higher than preoperative levels and the EEG still contains alpha rhythm as the predominant background despite the K-complex, spindle and hypopnoea-related arousal otherwise defining it as stage 2 sleep. Figures 7.4 – 7.7 contain epochs from the pre- and postoperative nights of subject 8 during both stage 2 and REM sleep. Subject 8 was receiving relatively high dose opioid via an epidural catheter after an endoluminal abdominal aortic aneurysm repair (a relatively minor incision in the groin). His postoperative studies were notable for their increase in high frequency (beta) activity in all stages and this can be seen in the zoomed EEGs. Also notable is the increase in EMG tone, not only in stage 2 sleep but also in REM. Note also that obstructive events were associated with a lesser degree of desaturation during the postoperative study as a result of oxygen administration.

119 FIGURE 7.1. Subject 5, stage 2, preoperative night (legend continues over).

120 Figure 7.1 legend continued – an epoch of stage 2 sleep from the preoperative night of subject 5. The epoch is readily identified as stage 2 by the K-complex and sleep spindle. Note the low amplitude mixed frequency background of the zoomed EEG.

121 FIGURE 7.2. Subject 5, stage 2, postoperative night 1 (legend continues over).

122 Figure 7.2 legend continued – an epoch of stage 2 sleep from the first postoperative night of subject 5. The epoch is again identified as stage 2 by the K- complex. Note the predominant alpha rhythm of the zoomed EEG and the higher EMG amplitude.

123 FIGURE 7.3. Subject 5, stage 2, postoperative night 2 (legend continues over).

124 Figure 7.3 legend continued – an epoch of stage 2 sleep from the second postoperative night of subject 5. The epoch is yet again identified as stage 2 by the K-complex, spindle and hypopnoea-induced arousal. Again note the predominant alpha rhythm of the zoomed background EEG and the higher EMG amplitude.

125 FIGURE 7.4. Subject 8, stage 2, preoperative night (legend continues over).

126 Figure 7.4 legend continued – an epoch of stage 2 sleep from the preoperative night of subject 8. The epoch is again identified as stage 2 by the K-complex.

127 FIGURE 7.5. Subject 8, stage 2, postoperative night 1 (legend continues over).

128 Figure 7.5 legend continued – an epoch of stage 2 sleep from the first postoperative night of subject 8. The epoch is somewhat equivocal but can be identified as stage 2 sleep by the respiratory events in the surrounding epochs. Note the increase in beta rhythm of the zoomed EEG and the higher EMG amplitude.

129 FIGURE 7.6. Subject 8, stage REM, preoperative night (legend continues over).

130 Figure 7.6 legend continued – an epoch of REM sleep from the preoperative night of subject 8. The epoch is identified as REM by the eye movements and low amplitude EMG.

131 FIGURE 7.7. Subject 8, REM, postoperative night 1 (legend continues over).

132 Figure 7.7 legend continued – an epoch of REM sleep from the first postoperative night of subject 8. The epoch is again identified as REM by the eye movements. Note however the higher amplitude EMG compared with figure 7.6 and the increase in the beta component of the zoomed EEG segment.

Discussion

The overall quality of sleep in the early postoperative period was, for most subjects, poor. This was evident both in their polysomnography and in their general comments. It was only partly reflected in their sleep efficiency which was somewhat better than expected but, generally, the findings with respect to sleep architecture, bearing in mind the limitations mentioned below, are consistent with and support the cumulated findings of previous investigations of perioperative sleep. In the early postoperative period there was a variable reduction in SWS and REM replaced by “lighter” stages of sleep, particularly stage 1, and the extent of this alteration seems to be proportional to the magnitude of the surgical insult. Smaller surgical procedures appear to be associated with reductions more in SWS than of REM. Inadequate data were collected to enable further conclusions to be drawn regarding previous findings of late postoperative REM or SWS rebound.

133 The relatively short early postoperative latencies noted for several subjects might be partly the result of sedative analgesia and, as noted above, it was common to find subjects asleep already when I arrived to apply the monitoring equipment. Subject 8 in particular fell asleep and obstructed his upper airway on several occasions while the set- up was actually taking place. His latency would have been even shorter than that recorded, but nursing intervention, application of oxygen amongst other things, kept him awake briefly. The issue of napping outside the usual “nocturnal” hours is discussed in greater detail below.

An important observation in this study was the increase, particularly associated with more major surgery and the need for opioid analgesia, in the proportion of the night spent in a transitional state between wakefulness and sleep. This state is commonly accompanied by marked respiratory instability with consequent alterations in the level of arousal and sympathetic activity (noted in this study as heart rate changes). This observation almost certainly explains, at least in part, the finding by Rahman et al. (Rahman 2001) that abnormal breathing was more frequent during “wakefulness” than sleep.

Of particular importance is the observation that the EEG during sleep is altered after surgery.

Implications

Previously published perioperative polysomnographic studies involving patients having procedures other than upper airway surgery (Johns 1974; Ellis 1976; Orr 1977; Kavey 1979b; Kavey 1983; Aurell 1985; Catley 1985; Lehmkuhl 1987; Knill 1990; Rosenberg 1994; Rosenberg-Adamsen 1996b; Edell-Gustafsson 1999; Cronin 2001; Rahman 2001; Drummond 2002; Wu 2003) have involved a wide variety of polysomnographic techniques. Two of the earliest studies used a method reliant upon the recording of an automated count of delta waves in the EEG and could not, therefore, be considered truly polysomnographic (Johns 1974; Ellis 1976). Similarly, for both their studies, Jacob Rosenberg’s group in Denmark used the Somnolog portable sleep monitoring system which records a signal derived from the alpha and delta activity of the EEG (Rosenberg 1994; Rosenberg-Adamsen 1996b). The finding of increased levels of high frequency

134 activity in the postoperative EEGs in this present study suggest caution is required when assessing their findings. The study by Catley et al. (Catley 1985) used fewer than the minimum number of channels thought to be necessary for accurate staging of sleep (Carskadon 1989).

As noted previously, studies using polysomnography after upper airway surgery have also been reported (Johnson 1986; Burgess 1992; Helfaer 1996; Terris 1996) but those have concentrated upon aspects of sleep-disordered breathing and have either not included or only superficially reported the actual polysomnographic findings with respect to sleep and are therefore of limited value in this respect.

Also mentioned already, a number of studies have used actigraphic techniques to monitor perioperative sleep (Bisgaard 2002; Kain 2002; Kain 2003). Actigraphy is of no use in determining the specific sleep stage, and its validity in the perioperative setting even for distinguishing sleep/wake state is dubious (Sadeh 2002). Movement artifacts were very commonly associated with arousals during this study, especially in some individuals and especially where respiratory obstruction was the cause. As many epochs of sleep included these arousal-related movements it is hard to see how actigraphy alone could be used reliably to assess postoperative sleep, even if there is evidence that it has good reliability in other settings. A study comparing actigraphy to polysomnographic assessment of sleep after surgery is clearly required.

Few studies, all from the Edinburgh group, have polygraphically examined perioperative sleep while assessing respiratory movement and/or airflow at the same time (Rahman 2001; Drummond 2002; Wu 2003). These studies used a non-standard EEG derivation, the importance of which is uncertain, and monitored for a single postoperative night. To my knowledge, this present study is, therefore, the first to monitor sleep using standard polysomnographic techniques, including extensive recording of respiratory parameters, commencing preoperatively and extending for several nights into the postoperative period.

The use of respiratory monitoring is an important consideration. A number of subjects in this study had many epochs in the early postoperative period which were somewhat ambiguous for staging in terms of the EEG and EMG. The situation was markedly

135 clarified by the presence of respiratory patterns, upper airway obstruction and periodic breathing mostly, and sometimes snoring on the audio channel, that strongly favoured the conclusion that the epoch was sleep rather than wakefulness. Effectively, I used respiratory parameters as a surrogate behavioural marker of sleep. In some cases, subject 5 for example, the ambiguity was probably the result of the effect of pain. Subject 8, on the other hand, would have had little pain after a relatively minor incision with effective epidural analgesia. The epidural, however, was delivering a considerable dose of opioid (50 micrograms of fentanyl per hour) for a patient of his age and the equivocal nature of his postoperative sleep EEG may well have been due to this. Aurell and Elmqvist (Aurell 1985) made the same observation with one of their subjects, scoring a large percentage (over 6 hours) of that subject’s night as “abnormal stage 1” which they then excluded from analysis.

Again to my knowledge, Drummond’s group in Edinburgh (Drummond 2002; Wu 2003) have been the only investigators to consider a possible effect of perioperative interventions on EEG micro-architecture when staging sleep. They took the interesting approach of using theta frequencies, as well as alpha, as a marker of wakefulness or arousal, based on their proposal that morphine slowed the EEG. While this implies they did not accept the usual assumptions regarding the EEG definitions of sleep, their particular approach may have been problematic for two reasons. From the observations in this study as detailed above it is likely that some factor, possibly pain, has a tendency to increase alpha activity in the EEG during sleep in the postoperative period. Secondly, their observation that respiratory events occurred more frequently during epochs staged as wakefulness (Rahman 2001) may reflect the possibility that the patients were actually asleep but an alpha or even theta predominant EEG led to ambiguity in the staging. This possibility is further supported by another of their observations, that patients were thought by the staff caring for them to be asleep at times when the EEG suggested otherwise, an observation also made by Aurell and Elmqvist (Aurell 1985).

There is indeed direct evidence in support of this notion of increased alpha activity during sleep after surgery. Rains and Penzien (Rains 2003) recently studied the association between chronic pain and a phenomenon known as the “alpha-EEG sleep pattern”. In questioning the association, they concluded that a variety of noxious stimuli, including pain, precipitate the intrusion of alpha EEG activity into sleep. In a

136 previous review of the relevant literature, Pivik and Harman (Pivik 1995) proposed certain fronto-central alpha EEG activity during sleep to be a marker of sleep maintenance processes, enhanced in response to sleep-disturbing events, rather than arousal. In a most interesting and relevant study, Trachsel et al. (Trachsel 1994) administered endotoxin to healthy volunteers and found not only an increase in both alpha and beta activity during sleep but also suppression of REM. That study also provides evidence that neurohumoral factors may be part of the explanation for the finding of Cronin et al. (Cronin 2001), that postoperative sleep disturbance occurs even when opioids are avoided and pain is well controlled. The most direct evidence, however, is the investigation of Drewes et al. (Drewes 1997) in which experimentally induced pain was shown to increase alpha, sigma and beta EEG activity during slow wave sleep in healthy volunteers.

Wu and Drummond (Wu 2003) may have illustrated the importance of respiratory monitoring for the staging of postoperative sleep in their paper. In order to eliminate potential bias, sleep staging and respiratory event scoring were performed independently in a blinded fashion, thus eliminating respiratory monitoring as a potential surrogate behavioural marker of sleep. In their figure 3 (reproduced as figure 7.8 below), the section of EEG marked as an arousal appears no different to that immediately preceding it. By contrast, the section that immediately follows, not originally marked as an arousal, contains EEG evidence of being just that. In addition, the part of the figure I believe to be a record of an arousal is associated with a large increase in respiratory amplitude, despite the assertion that the figure contains no change in breathing pattern, confirming the occurrence of the arousal.

The conclusion drawn in figure 4 (reproduced as figure 7.9 below) of Wu’s paper (Wu 2003) may also be incorrect, again illustrating the utility of staging sleep with the aid of respiratory monitoring. The figure, quite correctly, is said to show a period of wakefulness. The “relief of respiratory obstruction” shown is, I believe, quite typical of the normal respiratory variation seen during wakefulness in most sleep studies, not only postoperative ones. Certainly, this was a common feature of the preoperative studies conducted as part of this investigation, intermittent mouth breathing being just one possible cause of marked alterations in nasal flow when subjects are awake. In postoperative studies, bouts of pain can cause patients to alter their breathing pattern as

137 well. Subject 4, suffering relatively little constant pain after his open prostatectomy, had several episodes of bladder spasm during study set-up. This was associated with long periods of breath-holding on each occasion. Such episodes may also be part of the explanation for the observations in Wu and Drummond’s study (Wu 2003).

Based on these observations, it is my suspicion that many epochs scored as wakefulness in previous studies of postoperative sleep may have indeed been sleep with an “abnormal” EEG or a state of deep sedation or drowsiness bordering on sleep but lacking the usual EEG/EMG indicators of stage 1 sleep. This would be consistent with nursing and personal observations but further investigation, using combined behavioural and polysomnographic study, are necessary to confirm this. It is altogether possible, of course, that the approach taken in this study, that of using respiratory parameters as surrogate behavioural markers of sleep, is itself problematic. It is important to address these issues because such “transitional” sleep, very commonly seen in the postoperative studies of this investigation, appears to be strongly associated with periodic breathing and upper airway obstruction.

138 Reproduction of Figure 3 from Wu and Drummond

FIGURE 7.8a. An annotated reproduction of figure 3 from Wu and Drummond (Wu 2003) (used with permission, legend continues over).

139 Figure 7.8 legend continued. The original legend read, “Fig. 3. An example of a short period of arousal from sleep, indicated by the solid bar, with no change in breathing pattern. The traces are two electroencephalogram channels, a submental electromyogram (EMG), and nasal flow.” The section I have marked with the hollow bar is where a clear arousal occurs. The arrows above the nasal pressure trace have also been added to the original figure. The breath marked “A” is clearly flattened compared with the one preceding it, a feature indicating obstruction, as stated by the authors and illustrated in another figure of the manuscript, also reproduced below (7.8b). The breath marked “B” is further reduced in amplitude and this coincides with the onset of what I believe to be the true arousal, following which there is a markedly larger breath, labeled “C”.

FIGURE 7.8b. See legend above (reproduced with permission). The original legend read, “Fig. 1. Nasal cannular pressure. An example of recovery from partial respiratory obstruction: At the arrow, there is an abrupt change in signal amplitude, associated with an increase in frequency and a change in waveform. The preceding waveform is flattened.”

140 Reproduction of Figure 4 from Wu and Drummond

FIGURE 7.9. Reproduction of figure 4 from Wu and Drummond (Wu 2003) (used with permission). The original legend read, “Fig. 1. Nasal cannular pressure. An example of recovery from partial respiratory obstruction: At the arrow, there is an abrupt change in signal amplitude, associated with an increase in frequency and a change in waveform. The preceding waveform is flattened.” Note the relative lack of flattening of the breath preceding the arrow, when compared to the figures (7.8a and b) above.

141 Limitations of the Study

As noted in chapter 5, it proved impossible to monitor sleep during the day in the postoperative ward, although it was initially intended that this would occur. From personal observations during pain rounds on the first postoperative morning, I would estimate that one quarter of the patients require arousal from sleep for their acute pain consultation. This observation has importance from both a research and clinical perspective. The data of Ellis and Dudley (Ellis 1976), bearing in mind the significant limitations of their recording technique (see chapter 5), and that of Edéll-Gustafsson (Edell-Gustafsson 1999) et al. suggest that total sleep time over the full 24 hours is actually increased on the immediate postoperative day after major abdominal and cardiac surgery while Aurell and Elmqvist (Aurell 1985) reported no sleep at all for many of their subjects for 24 hours or more using continuous postoperative recording. Thus daytime naps could reduce the need for sleep at night and affect the interpretation of nocturnal recordings. In an example of clinical relevance, ward nursing staff are usually well aware that therapy for obstructive sleep apnoea (OSA) needs to be applied at night, but it is extremely rare to find OSA patients using their therapeutic device while napping or in a drug-induced stupor during the day. Further research into this aspect of postoperative sleep is therefore acutely necessary but this will almost certainly have to wait until reliable methods of monitoring sleep are improved and made less intrusive.

All but one of the subjects were male. Whether or not this had any effect on the outcome is a matter of speculation. This was not by intention but likely occurred as a result of two main factors. In the early months of the study, when patients on my own surgical lists were more likely to agree to participate, the only inpatient surgical procedures for which I anaesthetised were urological, mainly prostatectomies. After taking up a melanoma list, again there was a majority of male patients because of the nature of the disease. Later, when patients with upper airway obstruction under anaesthesia became the most likely recruits, again the male preponderance of sleep apnoea was a likely cause. It is of interest that the only female who agreed to participate had, unbeknownst to me at the time, already been investigated for sleep apnoea.

142 In almost every respect other than their sex, the subjects were an extremely heterogeneous group. Most notably, the nature of their surgery, the anaesthesia provided and the consequent need for opioid analgesia varied considerably. Any attempt to control for these factors would have resulted, in this study, in almost complete failure to recruit sufficient numbers. Intercurrent illnesses and regular medications also differed a great deal, as would be expected with a relatively elderly group of subjects. There was also no possible control, as occurred in some other studies, over ward conditions. There is no doubt whatsoever that these factors affected this study and the results must therefore be viewed with this in mind. On the other hand, this heterogeneity may allow some limited consideration of the effect of the magnitude of the surgical insult and opioid dose on postoperative sleep.

Some early postoperative studies contained many of the equivocal epochs described above. These studies were extremely difficult to stage and required multiple re- examinations in most cases including epoch-by-epoch assessment of zoomed-in sections of the EEG. It is possible, likely even, that some of the epochs scored as sleep were actually wakefulness or a state of drowsiness similar to that described in this study as “wake/sleep transition time”. Most, however, contained or were associated with enough features to score them as sleep with reasonable confidence.

Caution is required when interpreting some of the statistical analyses in this study. Repeated measures analysis, the most appropriate for this type of study, would be of questionable accuracy because of missing nights with small numbers of subjects as discussed in the Methods (chapter 5). Wilcoxon signed rank test for matched pairs was chosen for the limited between-night comparisons because it could not be assumed the data were Gaussian. While population values for the percentage of most sleep stages and other sleep parameters would probably be normally distributed, postoperative values are highly likely to be skewed. Moreover, the heterogeneity of the subjects and their surgical insults would also predispose the results towards non-normality. The use of a non-parametric test, however, is not as sensitive with small samples such as this, probably explaining the lack of statistical significance for several of the measures that appeared to show fairly distinct postoperative alterations consistent with the findings of other studies.

143 Conclusions

Sleep macro-architecture is altered after surgery in a variable fashion, mostly dependent on the type of surgical procedure and degree of postoperative discomfort. Major surgery invariably results in poor postoperative sleep, even in the presence, it appears, of adequate analgesia. More importantly, however, there are changes in the micro- architecture of the postoperative sleep EEG that may not only distinguish it from normal sleep but also result in misinterpretation of postoperative studies. These changes, which seem mainly to be an increase in alpha and/or beta activity, appear to be relatively consistent after a significant insult and may be the result of pain and/or opioid administration. It should not be assumed that the EEG and EMG characteristics previously defined for normal sleep adequately define sleep in the postoperative setting. Further definition of the behavioural and electrophysiological characteristics of sleep after surgical stress is therefore required.

Alternative methods for examining sleep in the postoperative setting, such as with actigraphy, are likely to be of limited value, certainly until they are validated in this setting in which sleep appears to have a character entirely different from normal sleep.

144 CHAPTER 8

SPECTRAL ANALYSIS OF THE

ELECTROENCEPHALOGRAM DURING

POSTOPERATIVE SLEEP

Visual analysis of the electroencephalograms (EEG) during pre- and postoperative sleep, performed as part of this investigation, suggest that there are systematic changes in the frequency spectra of the EEG after surgery, particularly when the magnitude of the surgical insult is significant and possibly when higher doses of opioids are present. Investigators in Edinburgh, as previously discussed, suspected there may be a generalised slowing of the EEG in response to morphine administration and adjusted their sleep staging technique accordingly (Rahman 2001; Drummond 2002; Wu 2003). It was observed during this present study, however, that higher than normal EEG frequencies appeared to be associated with postoperative sleep.

In order to clarify this situation, spectral analysis of EEG samples from a number of subjects during different sleep stages was performed and comparisons made between samples taken from pre- and postoperative nights.

Methods

Subject recruitment and the methods for staging sleep have already been described in detail (see chapters 5 and 7). As far as possible sleep was staged according to standard definitions and criteria (Rechtschaffen 1968; Carskadon 1989). In some postoperative studies, however, the electrophysiological staging of sleep was ambiguous, particularly between stages 1, 2 and wakefulness. Rapid eye movement (REM) sleep was much less ambigious while slow wave sleep (SWS, sleep stages 3 and 4) was not ambiguous at all. The uncertainty could often be clarified by simultaneous examination of eye movements

145 and respiratory parameters and where this was possible the epoch was staged as wakefulness or sleep as appropriate, otherwise the epoch was scored as awake.

Subjects with satisfactory studies on both the preoperative night and either the first or second postoperative night were included. The polysomnographic records were examined in order to identify periods of unbroken stage 2, 3 and rapid eye movement (REM) sleep. It was arbitrarily decided that a continuous period of 6 30-second epochs of sleep would be used where possible, although a few samples were necessarily shorter (none less than 4 epochs) and some longer (maximum 10 epochs).

Having identified an appropriate period of sleep, the raw EEG data for that period for both sides of the brain (C3/A2 and C4/A1 derivations) were converted to text file samples of raw data from the Replay software. As previously described, the EEG had been sampled at 125Hz. These raw EEG samples were then imported into MATLAB 6.5 (The Mathworks, Incorporated, Natick, Massachusetts, USA). Power spectral density estimates were plotted for each sample using Welch's averaged, modified periodogram method with Hanning windows of 256 data points (Welch 1967). This window length was selected as it gave acceptable frequency resolution (approximately 0.5Hz) with narrow 95% confidence limits.

As sleep is usually identified by the transition from a predominantly alpha EEG to one in which theta rhythms predominate (Worsnop 1998; Worsnop 2000), it was decided that the ratio of theta power (3 – 8Hz) to alpha + sigma (8 – 16Hz) would be compared from pre- to postoperative studies. In order to do this, the area under the power spectral density curve from 3 – 8Hz was divided by the area under the curve from 8 – 16Hz. The resulting ratios from each side of the brain were averaged to create a single ratio for each subject, in each relevant sleep stage, for each of the pre- and postoperative studies. A reduction in the ratio would indicate a relative increase in power in the 8 – 16Hz range normally associated with arousal. As it was also suspected from visual examination of the EEGs that there might be an increase in the beta frequencies during postoperative sleep, the same process was carried out examining ratios of sub-beta (theta + alpha + sigma) power (3 – 16Hz) to beta power (16 – 32Hz).

146 Statistical Analysis

Between-night comparisons were performed using two-tailed Wilcoxon signed rank test (Prism 3.0a for Macintosh, Graphpad Software Inc., San Diego, California, USA) with P < 0.05 considered significant.

Results

Of the nine subjects with preoperative studies, two (subjects 3 and 6) withdrew after the perioperative night. The preoperative EEG of subject 2 was relatively abnormal, possibly as a result of his chronic theophylline medication, and his first postoperative recorded sleep (night 2) contained inadequate sleep unbroken by arousals. He was therefore excluded. All six remaining subjects had satisfactory periods of stage 2 sleep on both the pre- and postoperative nights and suitable EEG samples were obtained. Two subjects had inadequate slow wave or REM sleep on the postoperative night examined and one, subject 8, had so many respiratory-related arousals that it was even difficult to find a series of four continuous epochs of stage 2 sleep in his case. In five cases, the postoperative samples were taken from the immediate postoperative night. The remaining subject (9) declined study on night 1 and his samples are therefore drawn from postoperative night 2. theta / (alpha + sigma) ratios

The ratio of power in the 3 – 8Hz range to that between 8 – 16Hz was reduced from the preoperative night to the postoperative night for all subjects in all sleep stages. Only the reduction during stage 2 was statistically significant (P = 0.03), there being inadequate samples in the other stages. The relevant ratios over both nights for stages 2, 3 and REM sleep are shown in figures 8.1, 8.2 and 8.3 respectively.

147 (theta + alpha + sigma) / beta ratios

This ratio, inversely related to the power in the beta EEG range, was also reduced in stage 2 (P = 0.03) and stage 3 sleep in all subjects, the latter not reaching statistical significance. These ratios are displayed in figures 8.4 and 8.5. Figure 8.6 shows the ratios during REM, pre- and postoperatively, and it can be seen that the effect here is inconsistent.

Values representing the relative postoperative change in both of these ratios for stage 2 sleep are given in Table 8.1. Each value is the relevant preoperative ratio divided by the corresponding postoperative ratio for each subject. Thus, a larger number represents a higher relative change in the ratio concerned, indicating a higher relative increase in the 8 – 16Hz or the 16 – 32Hz bands respectively.

Subject 3-8/8-16 3-16/16-32 1 1.320 1.558 4 1.144 1.421 5 1.408 4.628 7 1.042 1.487 8 1.060 1.344 9 1.157 1.772

TABLE 8.1. The relative change in power ratios after surgery for stage 2 sleep. The value in the 3-8/8-16 column for each subject is the preoperative ratio of theta power to alpha + sigma divided by the corresponding postoperative ratio. A higher value represents a greater relative postoperative increase in alpha + sigma power. The 3-16/16-32 column contains values representing the equivalent calculation for (theta + alpha + sigma)/beta power, the number being proportional to the relative increase in beta power from the pre- to postoperative night.

148 FIGURE 8.1. The ratio of theta (3 – 8Hz) to alpha + sigma (8 – 16Hz) EEG power during stage 2 sleep for six subjects, pre- and postoperatively.

149 FIGURE 8.2. The ratio of theta (3 – 8Hz) to alpha + sigma (8 – 16Hz) EEG power during stage 3 sleep for three subjects, pre- and postoperatively.

150 FIGURE 8.3. The ratio of theta (3 – 8Hz) to alpha + sigma (8 – 16Hz) EEG power during stage REM sleep for three subjects, pre- and postoperatively (note, this figure has a difference scale to those previously).

151 FIGURE 8.4. The ratio of theta + alpha + sigma (3 – 16Hz) to beta (16 – 32Hz) EEG power during stage 2 sleep for six subjects, pre- and postoperatively.

152 FIGURE 8.5. The ratio of theta + alpha + sigma (3 – 16Hz) to beta (16 – 32Hz) EEG power during stage 3 sleep for three subjects, pre- and postoperatively.

153 FIGURE 8.6. The ratio of theta + alpha + sigma (3 – 16Hz) to beta (16 – 32Hz) EEG power during REM sleep for three subjects, pre- and postoperatively (note the change in scale from the corresponding figures showing sleep stages 2 and 3).

154 The potential effect of pain and/or opioids

Recognising the large potential role of confounders, it might be possible to make some assessment of the effect of opioids in these EEG changes. Subjects 1 and 7 received the same dose of codeine at the start of their postoperative study as they did on the preoperative night so any effect on their EEGs should have been negated. Subject 4 received no opioid for over 4 hours prior to his study. Subjects 5 and 9 were given the oral equivalent of approximately 3mg of parenteral morphine shortly before the periods from which their EEG samples were drawn. Subject 8 was continuously receiving 50µg of fentanyl per hour, effectively the highest dose of opioid administered to any of the subjects by far. While he also had the smallest change in both the ratios assessed, it is difficult to draw any clear conclusion from this.

Again considering the potential for confounders, and the lack of any specific information regarding the degree of painful stimulus in each case, it might be possible to make a very limited assessment of the role of pain in these EEG changes. Subject 5, after his midline laparotomy, would almost certainly have suffered the most discomfort of the subjects in this study. The relative change in both his ratios was the largest, indicating the largest relative increase in his entire 8 – 32Hz EEG frequency range. Subject 8, in contrast, with arguably the least likelihood of significant pain, had the lowest change in the ratio assessing a beta increase, and the second lowest change in the ratio assessing the 8 – 16Hz band.

Discussion

The results of this study suggest that there are systematic changes in the EEG during sleep after surgery. There appears to be an increase in relative power in both the alpha + sigma (8 – 16Hz) and beta (16 – 32Hz) frequency ranges, although the latter may be inconsistent in REM sleep when it occurs. There is also some indication that the degree to which this increase occurs may be proportional to the painful stimulus.

155 The changes observed in this study are consistent with the only other published spectral analysis of postoperative sleep EEGs. Lehmkuhl et al. (Lehmkuhl 1987) mentioned in a single sentence without further discussion that they had found an increase in fast beta frequencies during sleep stages 1 and 2 in all their post-surgical groups. They are also consistent with the finding of an increase in the alpha, sigma and beta activity during slow wave sleep with experimentally induced pain (Drewes 1997).

Implications

Aurell and Elmqvist (Aurell 1985) observed that one of their subjects had ambiguous electroencephalographic activity during the first postoperative night after extensive abdominal surgery. They scored a large percentage (over 6 hours) of that subject’s night as “abnormal stage 1”. It is quite likely that their observation is explained, at least in part, by the findings of this present study. Similarly, Rahman et al. (Rahman 2001) observed that respiratory events occurred more frequently during epochs staged as wakefulness. There are a number of possible explanations for that observation but some of the epochs staged thus may have been rendered ambiguous by an increase in alpha and beta EEG activity.

There are few published investigations into the effect of opioids on the electrophysiologic markers of sleep. Kay et al. (Kay 1979) found that morphine, methadone and heroin all resulted in an increase in EMG and EEG alpha activity while decreasing theta and spindling. This is consistent with the findings of the present study, although few of the subjects had significant amounts of opioid in the immediate study period. It is also somewhat contrary to the approach taken by Drummond et al. (Drummond 2002) in which they used frequencies above 4Hz as a marker of arousal.

As discussed in chapter 7, there is also evidence in support of the occurrence of increased alpha activity during sleep in the presence of noxious stimulation (Pivik 1995; Rains 2003) as well as higher frequency bands (Drewes 1997). This is in keeping with the observation, in this study, of a possible association between the degree of postoperative pain and the extent of the effect on the EEG.

156 An alternative explanation lies in the effect of drowsiness on the EEG. While Pivik and Harmon (Pivik 1995) suggested fronto-central alpha rhythm, 1 – 2Hz slower than occipital waking alpha, as a possible marker of sleep maintenance processes in the presence of noxious stimuli, Santamaria and Chiappa (Santamaria 1987) identified the same type of fronto-central alpha activity as being one feature typical of drowsiness. If both of these propositions are correct, postoperative subjects might pass from a state of opioid-induced drowsiness into sleep and back again without any noticeable alteration in their EEG. This would help to explain many of the observations made during studies of postoperative sleep, as fronto-central EEG derivations have been most commonly used in this setting.

Further evidence for this possibility lies in the effect of anaesthesia on the EEG, and, in particular, the changes that occur in the EEG spectrum as subjects move from wakefulness through sedation into anaesthesia (Rampil 1998). Light sedation is associated with an increase in beta power while during heavier sedation alpha rhythms predominate. Patients requiring opioid analgesia after surgery are universally sedated to some extent by that analgesia and, after major surgery, it may well be that deep sedation and sleep become indistinguishable both behaviourally and electrophysiologically.

This observation has important implications both for previous work on postoperative sleep and for this present study. The use of frequencies above 4Hz to distinguish wakefulness from sleep, the approach taken by Drummond et al. (Drummond 2002), is likely to understate the amount of sleep. In contrast, the approach taken in this study may have identified as sleep, epochs that were actually sedated wakefulness. It is similarly impossible to determine what effect this ambiguity may have had on the results of previous studies involving electrophysiological analysis of postoperative sleep. It is very likely that there were some effects that were probably significant especially in studies without surrogate behavioural markers of sleep or with non- standard or automated EEG processing techniques.

Limitations of the Study

It must be recognised that the findings of this study are potentially self-fulfilling. The study is based entirely on what is assumed to be appropriate staging of sleep in the

157 postoperative setting, using a combination of both electrophysiological and surrogate behavioural markers to stage sleep that is sometimes ambiguous. Should the original staging of sleep be inaccurate, and epochs scored as sleep were, in truth, wakefulness, then it would be no surprise to find an increase in higher frequency components of the EEG. The periods of sleep selected for this analysis, however, were chosen on the basis that the staging was as certain as it could be for that subject. There could be little doubt about any of the epochs scored as stage 3 and while statistically not significant, the findings in that sleep stage were entirely consistent with those during stage 2 for the subjects concerned. While these results therefore need to be viewed with some caution, this lack of certainty again highlights the need for combined behavioural and polysomnographic studies of sleep in the postoperative setting.

Postoperative studies, as noted by Drummond et al. (Drummond 2002), are prone to difficulties with sensor adhesion. Mismatched impedances can result in unsatisfactory common-mode noise rejection and this could be a factor in the apparent alteration in postoperative sleep EEGs, although this should be made evident by an unusually large 50Hz mains peak in the spectrum. Another possible explanation for changes in the beta EEG range particularly might be increased scalp EMG activity, itself possibly as a consequence of pain.

Conclusions

There appears to be a systematic increase in the relative amounts of alpha (8 – 12Hz) + sigma (12 – 16Hz) and beta (16 – 32Hz) EEG activity during sleep after surgery, throughout most sleep stages, although the effect is less and/or inconsistent in REM. The cause of these EEG changes during sleep is unclear but it is possible that pain plays a more important role than opioid medication. The potential role of environmental factors such as light and noise must also be considered. While this finding needs to be confirmed by further investigation using behavioural as well as electrophysiologic monitoring of postoperative sleep, it may help to explain a number of previous observations in this setting. Further investigations into the individual effects of opioids and experimentally induced pain on the macro and micro-architecture of sleep may be very helpful.

158 Under certain postoperative circumstances, particularly after major surgery and with heavy opioid analgesia, sedation and sleep may blur into each other. The very definition of sleep under these circumstances may need to be re-examined. The behavioural correlates of both sleep and deep sedation are likely to be so similar that investigation of this issue may be difficult and contentious.

159 CHAPTER 9

POSTOPERATIVE BREATHING DURING SLEEP

Although hypoxaemia in the postoperative period has been recognised for many years (Papper 1971; Spence 1972; Jones 1990), the first examination of ventilatory patterns and their association with sleep was carried out by Catley et al. (Catley 1985). Studying only the first postoperative night, they found that all profound episodic oxygen desaturations in the postoperative period were caused by upper airway obstruction, occurring only during sleep. The effect of sleep on postoperative respiration has been a subject of interest since then. Knill’s group’s finding (Knill 1987) of a rapid eye movement (REM) sleep rebound, after an initial period of REM suppression, led them and others to speculate on the potential consequences of a REM-related increase in hypoxaemic events. Despite a relative dearth of direct evidence, much has since been made of the possible effect of this REM rebound on late postoperative morbidity and mortality.

This investigation was carried out to further characterise the role of sleep and alterations in its architecture in the generation of postoperative hypoxaemic events.

Methods

160 Statistical Analysis

As discussed in previous chapters, small numbers of heterogeneous subjects and missing data render the use of inferential statistics difficult and unreliable in this study. Analysis of the data is therefore primarily descriptive. Summary statistics are presented as mean ± standard deviation unless otherwise indicated. Between-night comparisons of paired data, where possible, are performed using two-tailed Wilcoxon signed rank test and comparison of unpaired data is performed with two-tailed Mann-Whitney U test (Prism 3.0a for Macintosh, Graphpad Software Inc., San Diego, California, USA). P < 0.05 is considered significant.

Results

As indicated in chapter 6, nine of the 19 subjects (1-9) had a study on the preoperative night and none of these had a prior diagnosis or history suggesting sleep disorders or sleep-disordered breathing. Two of these (subjects 3 and 6) withdrew after the preoperative study. Two subjects (10 and 11) were recruited pre-operatively but not studied until the first or second post-operative night. Subject 10 admitted after recruitment that she had previously had a diagnostic sleep study that confirmed severe obstructive sleep apnoea. She had declined a trial of therapy at that time. Subject 15 was a day-of-surgery admission and he was recruited immediately after surgery on the basis of a difficult intubation (as was subject 12) but he declined study until the second postoperative night. The remaining 6 subjects (13, 14, 16 – 19) were all approached postoperatively after they were found to have obstruction-prone airways under general anaesthesia.

For reasons indicated in chapter 7, the postoperative studies of subjects 11 and 12 were technical failures and, along with subjects 3 and 6, are not considered any further in this study.

The demographics, medical histories, surgical details and sleep results of the subjects have already been described in chapters 6 and 7.

161 Subject 16 removed the nasal flow transducer 95 minutes into his single night 2 study, leaving the rest of his electrodes intact for the remainder of the night. The results of his respiratory analysis must therefore be considered with some caution as the majority of it had to be carried out using oxygen saturation and abdominal and thoracic plethysmography alone, doubtless introducing some inaccuracy.

The average awake oxygen saturations for all subjects having postoperative studies are given in table 9.1. Overall, there appears to be little difference between the means of any of the first three postoperative nights (96 ± 2, 95 ± 2, 95 ± 2 respectively) and the preoperative night (95 ± 3), despite minimal use of oxygen therapy. This is supported by statistical comparison of paired data when night 1 (P = 0.37, n = 5), night 2, (P = 0.5, n = 4) and night 3 (P = 0.75, n = 4) are compared to the preoperative night. During wakefulness, respiratory variations, including a large number of events that might be misinterpreted as relief of obstruction if scored blind to sleep/wake state, were common and an example of this is shown in figure 9.1. Note particularly the airflow trace in which these variations occur without any associated typical change in oxygen saturation other than artifact, also a common event during wakefulness. There were few, if any, identifiable significant oxygen desaturations below average levels during wakefulness that might not be attributed to movement artifact.

162 FIGURE 9.1. Subject 1, awake, preoperative night (legend continues over).

163 Figure 9.1 legend continued – an epoch of wakefulness from the preoperative night of subject 1. The epoch is readily identified as wakefulness by the moderate amplitude, high frequency EEG, blinking and high amplitude EMG. Note in the airflow trace periods that might be interpreted as obstructive hypopnoeas if examined in isolation.

Subject Preop Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7

1 96 95 95 2 90 91 4 95 96 95 96 94 95 5 98 98 96 97 7 93 92 8 95 94* 9 97 96 97 10 99* 97 13 95 14 91 15 94 16 91 17 98 98 18 96 96 19 96 96

TABLE 9.1. Average awake oxygen saturations (per cent) for all subjects with postoperative studies. Preop = preoperative night; Postop = postoperative night; * denotes oxygen therapy used.

164 Subject Preop Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7

1 15.4 34.9 35.8 2 4.0 23.8 4 28.8 8.0 26.0 13.1 18.0 24.5 5 1.5 3.0 29.3 14.9 7 9.4 22.6 8 42.9 66.9 9 27.0 41.0 37.3 10 65.6 82.1 13 65.6 14 46.5 15 165.4 16 78.0 17 38.7 63.9 18 69.6 19.4 19 69.3 48.6

TABLE 9.2. Total respiratory disturbance index for all sleep (sum of apnoeas and hypopnoeas per hour of sleep) for all subjects with postoperative studies. Preop = preoperative night; Postop = postoperative night.

Table 9.2 displays the total respiratory disturbances indices (RDI – the number of apnoeas and hypopnoeas, summed, per hour of sleep) on each night for subjects studied postoperatively. As nearly half the subjects, all having no preoperative study, were recruited on the basis of airway difficulties under anaesthesia, any attempt to summarise or compare the night-by-night data overall would be futile. A very notable and obvious comparison can be made, however, between those subjects recruited at random and those suspected of having obstructive sleep apnoea (OSA) as a result of intraoperative airway/intubation difficulties. Assuming the greatest postoperative risk occurs on the night with the maximum number of events, it seems logical to use the worst night for each subject for the purposes of this comparison. These worst-night RDIs are shown for each group in figure 9.2. Note that subject 10, with previously diagnosed severe OSA, is included in the randomly selected group as her diagnosis was not known at the time of recruitment. Despite this, the subjects recruited with suspected OSA had significantly higher RDIs (P = 0.03) on their worst postoperative night. Figure 9.3 shows the overall RDIs for all postoperative nights studied, plotted against the percentage of REM sleep for that night. It can readily be seen that there is no relationship (r2 = 0.13) between

165 these variables, suggesting that, overall, reappearance or rebound in REM sleep does not result in a clinically significant increase in RDI. On the contrary, the nights with the greatest amount of REM sleep were associated with fewer events, despite the high representation of individuals with OSA in this sample.

200

100

Random Suspected

FIGURE 9.2. Worst-night maximum postoperative respiratory disturbances index (RDI – sum of apnoeas and hypopnoeas per hour of sleep) for subjects recruited at random preoperatively versus those recruited with suspected obstructive sleep apnoea as a result of intraoperative airway/intubation difficulty.

166 200

100

0 0 10 20 30 REM%

FIGURE 9.3. The total respiratory disturbance index (events per hour) for sleep overall plotted against the percentage of REM sleep (REM%) for each postoperative night studied.

The separate RDIs for non-REM (NREM) and REM sleep are shown in tables 9.3 and 9.4 respectively. As with the total sleep RDIs, overall summaries and between-night comparisons are inappropriate. No clear pattern is visible in the data. To examine the possible effect of REM re-emergence I will consider subjects with preoperative studies individually, on the night with maximum postoperative REM. Subject one, on his third postoperative night when his REM recovered to near-preoperative values, had double the number of events compared with his preoperative study but this was entirely due to

167 an increase in the NREM RDI, his REM RDI actually falling. Subject 2, similarly increased his RDI overall but this was due more to an increase in NREM RDI than REM. While subject 4’s REM/NREM ratio of events increased considerably on his maximum REM night (a “rebound” night), the overall index of respiratory events was actually less than preoperatively. Subject 5 had a significant increase in his overall RDI on the night of REM re-emergence and he certainly had a much higher RDI in REM on that night. However, his total RDI was lower than that on the night with the least REM and the increase in the absolute number of events during NREM, making up more than 91% of his sleep, was far more important than the number of events occurring during REM. The situation with subject 8 was almost identical to that of 5. Subject 7, despite an increase in the amount of REM on his postoperative night, increased his RDI almost entirely as a result of additional NREM events. Subject 9 had greater RDIs in NREM on both postoperative nights studied. Overall, therefore, the re-emergence (or even possible rebound of REM in the case of subject 4) had little or no observable effect on the number of respiratory events during postoperative nights.

Subject Preop Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7

1 16.3 33.0 43.2 2 2.2 23.1 4 24.9 8.2 22.9 4.7 11.1 20.6 5 0.4 2.9 29.5 12.2 7 8.8 24.3 8 42.1 66.5 9 27.0 41.4 39.0 10 65.7 81.5 13 70.5 14 46.6 15 166.2 16 78.0 17 35.8 64.6 18 69.6 17.5 19 69.3 52.4

TABLE 9.3. Respiratory disturbance index for non-rapid eye movement sleep (events per hour) for all subjects with postoperative studies. Preop = preoperative night; Postop = postoperative night.

168 Subject Preop Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7

1 12.7 42.1 9.7 2 26.1 32.7 4 46.3 0.0 39.7 37.0 39.5 36.3 5 10.4 6.7 n/r 44.0 7 12.4 15.5 8 47.4 72.0 9 27.3 36.9 10.0 10 64.6 84.6 13 45.1 14 46.2 15 n/r 16 n/r 17 65.5 58.8 18 n/r 31.0 19 n/r 19.7

TABLE 9.4. Respiratory disturbance index for rapid eye movement sleep (events per hour) for all subjects with postoperative studies. Preop = preoperative night; Postop = postoperative night; n/r = no/inadequate REM for assessment.

It can also be seen from tables 9.2, 9.3 and 9.4 that subjects with the most significant increases in overall RDI over consecutive nights where opioid medication would not have been the cause (2, 10 and 17), the RDI increased as a result of a greater or equal increase in NREM RDI compared to REM for that night.

Tables 9.5 and 9.6 display the RDIs of central or mixed origin during NREM and REM sleep respectively. For the majority of subjects these were trivial or non-existent and there is no discernible effect of any specific perioperative event. The vast majority of central apnoeas occurred during sleep onset or in the phase immediately following an arousal, often brought on by another respiratory event. Subject 17, on both nights studied, in addition to a significant number of central events almost all of which occurred during stage 1, had a considerable amount of wake/sleep “transition” time (as defined in chapter 6) and this was characterised by many central events not included in his RDIs. The single night recorded for subject 15 is so extraordinary that it warrants individual consideration in detail. Forty-nine years of age, with a body mass index of 27.8kg/m2 and no history of anything other than melanoma, the only thing suggesting he might be prone to upper airway obstruction was his somewhat difficult endotracheal

169 intubation (Cormack & Lehane grade 3 laryngoscopy (Cormack 1984)). He was approached for that reason after an ilio-inguinal lymph node dissection (an almost vertical incision of approximately 30cm centred on the inguinal ligament), based on the finding of Hiremath et al. (Hiremath 1998) that difficult intubation was a predictor of OSA. He took 10mg of oxycodone about 30 minutes prior to sleep onset. A representative sample of his entire night’s sleep is shown in figure 9.4. Shown a recording of his oxygen saturation trace for the night, both he and his wife denied ever noticing anything during sleep at home and he was totally unaware that his sleep during this study was interrupted by so many events. Despite this, he refused any further involvement in the investigation, and declined the offer of free out-of-hospital polysomnographic follow-up. While it is therefore impossible to know if this pattern of respiratory instability is typical of his normal sleep, it would seem unlikely, given his wife’s denials.

Subject Preop Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7

1 4.5 6.9 2.8 2 0.0 0.4 4 0.2 0.5 0.2 0.0 0.5 0.3 5 0.0 0.0 0.0 0.0 7 0.5 0.0 8 6.3 0.1 9 0.0 0.0 0.0 10 0.5 1.6 13 9.7 14 0.0 15 80.7 16 1.4 17 4.3 13.2 18 0.0 0.0 19 0.7 0.2

TABLE 9.5. Central/mixed apnoea index for non-rapid eye movement sleep (events per hour) for all subjects with postoperative studies. Preop = preoperative night; Postop = postoperative night.

170 Subject Preop Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7

1 1.9 4.2 0.0 2 0.0 0.0 4 0.0 0.0 0.0 0.0 0.0 0.0 5 0.0 0.0 n/r 0.0 7 0.0 0.0 8 7.1 0.0 9 1.8 0.0 0.0 10 0.0 0.9 13 1.0 14 0.0 15 n/r 16 n/r 17 0.0 2.4 18 n/r 0.0 19 n/r 0.0

TABLE 9.6. Central/mixed apnoea index for rapid eye movement sleep (events per hour) for all subjects with postoperative studies. Preop = preoperative night; Postop = postoperative night; n/r = no/inadequate REM for assessment.

171 FIGURE 9.4. Subject 15, postoperative night 2 (legend continues over).

172 Figure 9.4 legend continued – an epoch from the second postoperative night of subject 15 showing two arousals resulting from respiratory events. Note the staging difficulty, using standard criteria, arising as a result of the portion of the epoch taken up by the arousals (see text) and the marked oscillatory respiratory instability associated with sleep state.

The minimum oxygen saturations for NREM and REM sleep and the average saturation nadirs for NREM and REM for all subjects with postoperative studies are shown in tables 9.7 – 9.10 respectively. For the majority of subjects and for the majority of postoperative nights there was very little difference (less than a few per cent) between these variables in NREM and REM, the direction of the difference varying both within and between subjects with no clear pattern. While three subjects (4, 10 and 19) had one or more postoperative nights where REM was associated with a 4% or more greater fall in saturation than NREM, there were four (1, 5, 8 and 14) who had a minimum saturation 4% or more lower in NREM than in REM.

173 Subject Preop Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7

1 89 89 85 2 87 87 4 89 89 85 89 85 89 5 95 95 87 90 7 89 87 8 82 85* 9 90 92 89 10 92* 89 13 84 14 82 15 87 16 85 17 89 89 18 85 87 19 90 89

TABLE 9.7. Minimum oxygen saturations in non-rapid eye movement sleep (per cent) for all subjects with postoperative studies. Preop = preoperative night; Postop = postoperative night; * denotes oxygen therapy used.

Subject Preop Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7

1 89 87 89 2 87 87 4 89 92 81 87 87 82 5 95 95 95 90 7 89 84 8 81 89* 9 90 92 87 10 98* 82 13 85 14 87 15 87 16 n/r 17 87 87 18 n/r 87 19 n/r 85

TABLE 9.8. Minimum oxygen saturations in rapid eye movement sleep (per cent) for all subjects with postoperative studies. Preop = preoperative night; Postop = postoperative night; * denotes oxygen therapy used; n/r = no REM.

174 Subject Preop Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7

1 95 94 95 2 90 90 4 94 95 93 95 93 95 5 98 98 95 97 7 92 91 8 94 98* 9 96 95 95 10 99* 96 13 93 14 90 15 92 16 90 17 97 97 18 95 95 19 95 95

TABLE 9.9. Average oxygen saturation nadirs in non-rapid eye movement sleep (per cent) for all subjects with postoperative studies. Preop = preoperative night; Postop = postoperative night; * denotes oxygen therapy used.

Subject Preop Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7

1 95 94 95 2 90 89 4 94 95 93 95 93 94 5 97 98 96 96 7 91 91 8 92 97* 9 96 94 95 10 99* 93 13 92 14 91 15 89 16 n/r 17 95 93 18 n/r 95 19 n/r 93

TABLE 9.10. Average oxygen saturation nadirs in rapid eye movement sleep (per cent) for all subjects with postoperative studies. Preop = preoperative night; Postop = postoperative night; * denotes oxygen therapy used; n/r = no REM.

175 Table 9.11 shows the average oxygen desaturation associated with respiratory events overall for all subjects studied postoperatively. It is readily seen from the similarity of results across all nights that recovery or rebound of REM is, at the most in this study, associated with an average desaturation 1% worse than preoperative nights or nights with less REM, excluding subjects with whom oxygen confuses the situation.

Subject Preop Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7

1 3 3 4 2 2 2 4 3 3 3 4 4 4 5 3 3 4 4 7 2 2 8 5 2* 9 3 3 4 10 1* 4 13 5 14 3 15 2 16 3 17 3 3 18 3 3 19 2 2

TABLE 9.11. Average oxygen desaturation (per cent) for all apnoeas/hypopnoeas throughout sleep for all subjects with postoperative studies. Preop = preoperative night; Postop = postoperative night; * denotes oxygen therapy used.

Discussion

The initial aim of this study to collect enough data to assess the effect of any REM rebound, if it occurred, on postoperative episodic hypoxaemia, was not achieved to an adequate degree as a consequence of recruitment difficulty and poor subject compliance. The majority of subjects, however, had some degree of REM in their postoperative study or studies, enabling some assessment of the role of REM sleep in this group of subjects with a high incidence of significant OSA, either suspected or not.

176 Overall, considering all of the RDIs, minimum oxygen saturations and average nadirs, REM sleep in the postoperative period in this study had very little effect that might be considered clinically significant, despite the high incidence of respiratory events. This is consistent with the findings earlier in this thesis of an overall lack of REM predominance for OSA and also with the lack of circadian variation in unexpected postoperative deaths. The original hypothesis that REM rebound might be contributing to such mortality was not substantiated.

Part of the observation shown in figure 9.3, that the number of events overall was unrelated to the percentage of REM, may result from a reduction in respiratory events caused by opioids, the reduced need for which may coincide with re-emergence of REM. Opioid usage by the subjects in this study, however, was minimal and fairly consistent across the nights studied so this should only have been a minor factor if it contributed at all.

While the software used for this study does not numerically separate events occurring during slow wave sleep (SWS) from those occurring in stages 1 and 2, giving rather a total number of events for NREM, it was seen during staging that the observation of Sériès (Series 2002) and Ayappa and Rapoport (Ayappa 2003), that obstructive events were rare in SWS, was indeed true in the postoperative period as well. The early postoperative suppression of SWS, replaced by stages 1 and 2 in which obstruction is common, is almost certainly one reason for an increase in sleep-related respiratory events prior to the recovery of REM. As SWS and REM tend to recover more or less at the same time, this may also be one reason why REM does not usually result in an overall increase in events when and if it rebounds. In this study, however, NREM was still associated overall with a higher number of events than REM for many subjects despite the recovery of SWS. The importance of SWS in the generation or prevention of postoperative respiratory events, related perhaps to alteration in sleep architecture, has not been specifically investigated.

A very important observation in the present study was the higher incidence of more severe degrees of postoperative OSA in patients identified as being prone to upper airway obstruction, based on the behaviour of their airways under anaesthesia, when compared with subjects selected at random. While the clinical importance of this

177 finding remains to be determined, it has to be assumed that patients with more significant sleep-related respiratory disturbance are likely to be at greater postoperative risk of morbidity. This study has indicated one possible means of identifying patients who might benefit from postoperative intervention, if the risk is found to be significant.

The observation of a high incidence of central and obstructive respiratory events during sleep onset and the lighter stages of sleep is similarly important. This is a recognised phenomenon (Dunai 1996; Dunai 1999) and has been discussed in previous chapters. The fact that the majority of sleep in the early postoperative period is made up of this type of lighter and “transitional” sleep is almost certainly one reason for the higher incidence of respiratory compromise during that phase of surgical recovery. The unheralded and extraordinary degree to which the sleep of subject 15 was interrupted by such sleep-onset respiratory events, every 20 seconds or so, was surprising. Assuming the observed phenomenon was not typical of his normal sleep, and it is indeed hard to see how it could be given the denial of any witnessed events by his wife and his own denial of symptoms, some postoperative factor was presumably the cause. The surgical procedure is not usually very painful, despite the length of the incision, and this is reflected in the relatively frugal use of opioid analgesia. Nevertheless, with a degree of painful stimulation increasing his arousability and/or a small amount of opioid depressing his respiratory drive and upper airway control during sleep, his state-related fluctuations in ventilation must have been amplified to the point of uncontrolled oscillation.

Peripheral chemosensitivity is known to vary widely in the normal population and it has been shown that individuals with high peripheral chemoreceptor drive are more likely to experience greater sleep-related ventilatory instability, particularly at sleep onset (Dunai 1999). It is possible that subject 15 was one such individual. If this is the case, patients with high peripheral chemosensitivity may be at particular risk of sleep-related postoperative morbidity as such frequent fluctuations in ventilation, oxygen saturation, arousal, heart rate and presumably other sympathetically mediated responses must surely impose a significant cardio-vascular burden.

The night of subject 15 illustrates another very important consideration in the assessment of both this study and those of other researchers, past and future. The epoch

178 shown in figure 9.4 contains two arousals, both arising from easily identifiable sleep- related respiratory events. If the epoch were to be staged strictly according to the criteria of Rechtschaffen and Kales (Rechtschaffen 1968), as all other postoperative sleep researchers appear to have done, this epoch should be staged as wakefulness as it contains more than 50% of arousal-type EEG. Had these criteria been strictly adhered to, this subject would have had no sleep staged, as his entire night was made up of epochs such as the one illustated, and the night would have been completely misrepresented, a patently absurd situation. Such epochs in this study were therefore staged in keeping with the guidelines suggested by Carskadon and Rechtschaffen for the staging of sleep in the presence of sleep pathology (Carskadon 1989). Rahman et al. (Rahman 2001) have previously commented on the possible role of the original arbitrary classification system in their own results. It is very likely that it was a factor in their finding of a higher number of respiratory events during epochs scored as wakefulness than sleep. Both for reasons given in the previous chapters regarding postoperative sleep staging, and for the reasons given above, that conclusion is possibly spurious.

Catley et al. (Catley 1985) stated that pronounced episodes of oxygen desaturation and obstructive apnoeas never (their emphasis) occurred while their patients were awake. Rahman et al. (Rahman 2001) were therefore surprised at their own finding of a higher number of respiratory events during epochs scored as wakefulness. The apparent conflict is actually quite readily explained. Catley’s definition of a “pronounced” respiratory event was an oxygen desaturation to less than 80%, much worse than the criteria used both by Rahman and in this present study. Rahman used much more sensitive measures of respiratory variation, presumably identifying a far greater number of events, and including, perhaps erroneously, a number occurring during wakefulness. In the present study, respiratory variations, including a large number of events that might be misinterpreted as relief of obstruction if scored blind to sleep/wake state, were common during wakefulness (figure 9.1). This again highlights the utility of staging sleep and respiratory parameters together. It also highlights the importance of the need for agreement upon what should be considered significant respiratory events in postoperative studies. The criteria used for diagnostic sleep studies and sleep research have been defined with the short and long-term sequelae of sleep related breathing

179 disorders in mind (AASM 1999). These definitions may not be appropriate for perioperative investigations.

Rahman’s group also suggested Catley’s use of 5 minute epochs as a possible cause for the differences found in their respective studies (Catley 1985; Rahman 2001). While Catley did score periods of paradoxic breathing and slow respiratory rate in epochs of 5 minutes, this does not seem to be the case for either obstructive events or sleep, the latter stated to have been scored using the standard criteria of Rechtschaffen and Kales (Rechtschaffen 1968). Notwithstanding the potential limitations of those criteria identified in the present investigation, this suggests the scoring of sleep should not have differed greatly between those two studies, and that Catley’s finding that significant obstructive events and desaturations occurred only during sleep should be reasonably accurate. The present investigation has replicated Catley’s finding, episodic hypoxaemic events occurring only during sleep or sleep onset with the possible exception of rare episodes where spasmodic pain might have induced breath-holding.

Limitations of the study

The equipment used for this research included a thermistor for oro-nasal airflow measurement. An American Academy of Sleep Medicine taskforce (AASM 1999) recently questioned the validity of the use of thermistors for research purposes, although they were probably referring mainly to their use in quantitative measurement, for which thermistors are clearly inadequate. Thermistors are relatively insensitive to small reductions in airflow and, theoretically, some hypopnoeic events may have gone undetected in this investigation. To some extent this would be offset by other limitations as indicated below. The combined use of thermistor flow detection and thoracic and abdominal plethysmography should have resulted in reasonably accurate recognition of events for the purposes of this study.

The classification of respiratory events in the postoperative setting is sometimes difficult. When oxygen therapy is used, respiratory events are associated with a lesser degree of oxygen desaturation, often less than the arbitrary criteria for the classification of an apnoea or hypopnoea if an arousal does not occur. Some apparent reductions in respiratory effort and airflow do not result in desaturation at all in the presence of

180 oxygen therapy. When an arousal occurs associated with such reductions, this can produce some ambiguity in terms of whether or not to classify the arousal as one associated with a respiratory event, and thus whether or not to classify the respiratory event as an hypopnoea.

There is, almost invariably, some quite normal reduction in respiratory effort and airflow at sleep onset. In the presence of a large number of arousals, such as might occur in the early postoperative period, reductions in ventilation often fit the arbitrary criteria for classification as an hypopnoea, because of the arousal, and this may lead to an artificially high RDI.

The extent to which these possible ambiguities and inaccuracies occurred within this study or affected the results is impossible to assess. It raises further questions about the applicability of standard sleep and respiratory event classification criteria in the postoperative setting as commented upon by Drummond et al. more than once (Rahman 2001; Drummond 2002). Until these classifications are validated in the postoperative setting, or improved upon, there is no alternative but to apply them in current research while bearing in mind the potential limitations.

Conclusions

Notwithstanding the ultimate inability of this study to completely assess the effect of possible REM rebound on postoperative nocturnal episodic hypoxaemia, the effect of REM on ventilation during postoperative sleep would appear to be minimal for the vast majority of patients, including those with significant obstructive sleep apnoea. Upper airway obstruction under anaesthesia and difficulty with intubation identify patients likely to have a greater degree of postoperative sleep-related respiratory compromise.

Some patients, possibly with a high peripheral chemoreceptor drive, may be at risk of uncontrolled sleep-onset-related oscillations in ventilation and arousal brought on by a combination of postoperative factors.

181 This investigation supports the finding of Catley’s original study (Catley 1985) that essentially all significant episodic hypoxaemic events in the postoperative period are sleep-related, with or without a contribution from opioid analgesia.

Scoring sleep state and respiratory parameters in a blinded fashion would seem to be an appropriate approach for the purposes of eliminating bias. Based on the findings of the present study, however, I believe it compromises the ability to accurately identify real and spurious events in both sleep and ventilation, particularly in postoperative studies.

182 CHAPTER 10

AROUSAL FROM POSTOPERATIVE SLEEP

The relationship between arousal from sleep following abdominal surgery and relief from airway obstruction was recently examined by Wu and Drummond (Wu 2003) who found that only 30% of postoperative arousals were associated with relief of upper airway obstruction.

This study was carried out to examine in greater detail the role of respiratory events in the generation of postoperative arousals.

Methods

Subject recruitment and the methods for staging sleep have been described and discussed in the preceding chapters. Standard definitions and criteria (ASDA 1992; AASM 1999) were used for the scoring of arousals and respiratory events. The possible implications and limitations of these definitions in the postoperative setting have been discussed as well, along with the differences between the approach taken by Wu and Drummond (Wu 2003) and that used in this investigation. Specifically, it is important to note that sleep and respiratory events were not scored in a blinded fashion for the purposes of the present study, as they were by Wu and Drummond.

Respiratory event-related arousals were recorded in two categories, those occurring before or within one breath of the completion of the respiratory event as indicated by airflow and plethysmography, and those occurring between one and five breaths after the respiratory event recovery. All other arousals were considered to have occurred without any clear association to respiratory events. When recovery from a respiratory event occurred without any evidence of arousal at all this was also noted.

183 All episodes of saturation artifact were identified and excluded manually during sleep staging.

Statistical Analysis

As discussed in previous chapters, small numbers of heterogeneous subjects and missing data render the use of inferential statistics difficult and unreliable in this study. Analysis of the data is therefore primarily descriptive. Summary statistics are presented as mean ± standard deviation unless otherwise indicated. Comparisons of paired data, where possible, are performed using two-tailed Wilcoxon signed rank test (Prism 3.0a for Macintosh, Graphpad Software Inc., San Diego, California, USA). Linear regression was also performed using the Prism software. P < 0.05 is considered significant.

Results

For reasons indicated in chapter 7, the postoperative studies of subjects 11 and 12 were technical failures and, along with subjects 3 and 6, are not considered any further in this study. The demographics, medical histories, surgical details, postoperative sleep and respiratory results of the subjects have already been described in chapters 6, 7 and 9. Subject 16 removed the nasal flow transducer 95 minutes into his single night 2 study,

184 leaving the rest of his electrodes intact for the remainder of the night. The results of his respiratory analysis must therefore be considered with some caution as the majority of it had to be carried out using oxygen saturation and abdominal and thoracic plethysmography alone, doubtless introducing some inaccuracy.

The total arousal index (arousals per hour of sleep) for all nights studied is shown in table 10.1. Overall (unmatched data) there appears to be an increase in the arousal index on the first postoperative night (49.3 ± 22.8) compared with the preoperative night (25.1 ± 6.5) but when matched pairs of studies only are compared (n=5) this does not reach statistical significance (P = 0.44).

Subject Preop Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7

1 20.0 30.8 35.7 2 27.6 41.0 4 33.7 15.8 24.9 18.9 22.4 24.2 5 20.7 33.7 42.2 18.7 7 15.5 32.1 8 28.2 71.0 9 30.2 47.2 38.4 10 67.4 60.8 13 62.7 14 45.2 15 161.9 16 80.1 17 43.4 62.4 18 69.6 22.4 19 79.7 50.7

TABLE 10.1. Total arousal index (arousals per hour of sleep) for all subjects with postoperative studies. Preop = preoperative night; postop = postoperative night.

185 The index of arousals with no clear association to respiratory events is given in table 10.2. The number of respiratory event-related arousals per hour of sleep is recorded in table 10.3 while table 10.4 gives the same respiratory-related arousal data as a percentage of all arousals. Arousals associated with respiratory events are by far the most common type on all postoperative nights, even amongst subjects not suspected of having obstruction-prone airways. Subjects 4 and 5 on the first postoperative night are the only exceptions. This finding is hard to reconcile, given one (subject 4) had significant obstructive sleep apnoea on his preoperative study while the other had almost none. There is no statistical difference (P = 0.44, n=5) between the percentage of respiratory-related arousals on the pre- and first postoperative nights when matched nights are compared. When all subjects are considered, 75 ± 31% of arousals on the first postoperative night were associated with respiratory events. When only randomly selected subjects are included, the proportion of respiratory-related arousals on the first postoperative night is still 67 ± 36%. Subjects with known or suspected obstructive sleep apnoea had a very high proportion of respiratory-related arousals, more than 80% of all arousals on all nights studied, irrespective of proximity to surgery.

Subject Preop Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7

1 6.3 2.5 5.0 2 23.2 20.3 4 9.3 9.8 5.2 9.6 9.2 6.7 5 19.8 30.7 18.4 6.6 7 6.6 9.2 8 2.7 5.0 9 4.3 7.7 2.9 10 1.4 1.2 13 1.8 14 3.9 15 1.6 16 6.4 17 5.7 2.6 18 5.4 4.4 19 7.2 3.4

TABLE 10.2. Index of arousals (number per hour of sleep) not associated with respiratory events for all subjects with postoperative studies. Preop = preoperative night; postop = postoperative night.

186 Subject Preop Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7

1 13.8 28.3 30.7 2 4.4 20.7 4 24.4 6.0 19.6 9.3 13.2 17.6 5 0.9 3.0 23.8 12.1 7 8.8 23.0 8 25.5 65.9 9 25.9 39.5 35.5 10 66.0 59.5 13 60.9 14 38.6 15 160.3 16 73.7 17 37.6 59.8 18 64.6 18.0 19 72.4 47.4

TABLE 10.3. Index of respiratory event-related arousals (number per hour of sleep) for all subjects with postoperative studies. Preop = preoperative night; postop = postoperative night.

Subject Preop Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7

1 69 92 86 2 16 50 4 72 38 79 49 59 73 5 4 9 56 65 7 31 72 8 90 93 9 86 84 92 10 98 98 13 97 14 85 15 99 16 92 17 87 96 18 93 80 19 91 93

TABLE 10.4. Respiratory event-related arousals as a proportion (per cent) of the total number of arousals for all subjects with postoperative studies. Preop = preoperative night; postop = postoperative night.

187 Tables 10.5 and 10.6 record, respectively, the percentage of arousals that were respiratory event-related for non-rapid eye movement (NREM) and rapid eye movement (REM) sleep. The proportion of arousals that are associated with respiratory events during NREM, with rare exception and on all nights both pre- and postoperative, is very similar to the proportion overall, while that in REM is more variable. This is further illustrated in figures 10.1 and 10.2, using the data from the preoperative and first postoperative nights respectively. Both figures plot the NREM and REM proportions of respiratory related arousals (Y-axis) against the overall proportion (X-axis) for each subject studied on the night concerned. On both nights, there is a close association between the NREM proportion and the overall proportion, preoperatively y = 0.97x + 3.5 (r2 = 0.90, P = 0.001) and postoperatively y = 1.00x – 0.49 (r2 = 1.00, P < 0.0001). Preoperatively, the REM proportion bears little resemblance to the overall proportion (y = 0.46x + 47, r2 = 0.60, P = 0.04). Postoperatively, while the association between the REM and overall percentages is still not as strong as during NREM, it is much more so than preoperatively (y = 0.98x + 1.9, r2 = 0.75, P = 0.01).

Table 10.7 contains the number of arousals per hour for each night studied that were classified as occurring within one and five breaths of the recovery from a respiratory event. In all cases such arousals made up a small fraction of both the respiratory event- related arousals and all arousals considered together.

Table 10.8 displays the number of respiratory events per hour of sleep that did not result in an arousal, for each night studied. These are then converted to a percentage of total respiratory disturbance index for the night concerned in table 10.9. While there is no clear pattern to these data, three subjects (5, 8 and 10) had markedly fewer events of this type on the first postoperative night, perhaps due to use of significant amounts of opioid, than on other nights devoid of opioids. Bearing in mind the problem of missing nights and unmatched subjects, overall the proportion of these events appear to decrease from the preoperative night (17 ± 17%) to the first postoperative night (6 ± 9%) and then increases back towards preoperative levels on the second (11 ± 9%) and third (17 ± 9%) nights.

188 Subject Preop Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7

1 72 92 89 2 10 49 4 68 39 78 27 49 72 5 0 8 57 59 7 57 70 8 91 93 9 84 84 93 10 98 97 13 98 14 91 15 99 16 92 17 87 96 18 92 77 19 91 94

TABLE 10.5. Respiratory event-related arousals during non-rapid eye movement sleep (NREM) as a proportion (per cent) of the total number of arousals in NREM. Preop = preoperative night; postop = postoperative night.

Subject Preop Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7

1 52 92 47 2 56 83 4 91 0 84 79 85 74 5 50 34 n/r 100 7 63 83 8 86 97 9 100 80 0 10 100 100 13 94 14 100 15 n/r 16 n/r 17 89 95 18 n/r 96 19 n/r 86

TABLE 10.6. Respiratory event-related arousals during rapid eye movement sleep (REM) as a proportion (per cent) of the total number of arousals in REM. Preop = preoperative night; postop = postoperative night; n/r = no/inadequate REM for assessment.

189 Preoperative Respiratory-Related Arousals in NREM and REM

100

0 0 25 50 75 100 Proportion of events overall

FIGURE 10.1. The percentage of arousals associated with respiratory events in non- rapid eye movement (NREM – squares) and rapid eye movement (REM – triangles) sleep plotted against the overall percentage of respiratory event-related arousals for each subject on the preoperative night. The regression line for NREM is y = 0.97x + 3.5 (r2 = 0.90, P = 0.001) and for REM y = 0.46x + 47 (r2 = 0.60, P = 0.04).

190 Postoperative Respiratory-Related Arousals in NREM and REM

100

0 0 25 50 75 100 Proportion of events overall

FIGURE 10.2. The percentage of arousals associated with respiratory events in non-rapid eye movement (NREM – squares) and rapid eye movement (REM – triangles) sleep plotted against the overall percentage of respiratory event-related arousals for each subject on the first postoperative night. The regression line for NREM is y = 1.00x – 0.49 (r2 = 1.00, P < 0.0001) and for REM y = 0.98x + 1.9 (r2 = 0.75, P = 0.01).

191 Subject Preop Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7

1 0.3 0.6 0.7 2 0.4 1.9 4 2.1 0.5 0.5 0.6 0.2 1.4 5 0.0 1.0 0.4 0.0 7 0.9 1.6 8 2.6 1.6 9 0.2 0.3 0.3 10 0.6 0.9 13 1.2 14 3.0 15 0.4 16 1.8 17 2.6 4.0 18 0.8 2.4 19 0.2 0.0

TABLE 10.7. The number of arousals per hour of sleep occurring between 1 and 5 breaths after recovery from a respiratory event. Preop = preoperative night; postop = postoperative night.

Subject Preop Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7

1 1.6 6.6 5.1 2 0.0 3.1 4 4.4 2.0 6.4 3.8 4.8 6.9 5 0.6 0.0 5.5 2.8 7 0.6 0.0 8 17.4 1.0 9 1.1 1.5 1.8 10 0.0 20.7 13 4.7 14 7.9 15 5.1 16 4.3 17 0.9 4.1 18 5.0 0.7 19 0.2 1.6

TABLE 10.8. The number of respiratory events per hour of sleep not associated with an arousal at all. Preop = preoperative night; postop = postoperative night.

192 Subject Preop Postop 1 Postop 2 Postop 3 Postop 4 Postop 5 Postop 7

1 10 19 14 2 0 13 4 15 25 25 29 27 28 5 40 0 19 19 7 6 0 8 41 1 9 4 4 5 10 0 25 13 7 14 17 15 3 16 6 17 2 6 18 7 4 19 0 3

TABLE 10.9. The proportion of respiratory events (per cent) not associated with an arousal at all. Preop = preoperative night; postop = postoperative night.

Discussion

Overall, arousals from sleep were very common, one every few minutes at the very least, on all nights studied, both pre- and postoperatively. For the subjects in this present study, at least two-thirds of arousals on the first post-operative night were related to apnoeic or hypopnoeic respiratory events. This proportion was higher still (more than 80%) for subjects with known or suspected sleep-disordered breathing, as might be expected. The overwhelming majority of respiratory event-related arousals occurred prior to or simultaneously with (within one breath of) the relief or recovery from the respiratory event.

The close association between the proportion of respiratory-related arousals in NREM and the overall proportion is not unexpected as NREM makes up around 80% of even normal sleep. With a reduction in REM postoperatively leading to a greater proportion of NREM sleep, this association would naturally become stronger, as it did in this study. The finding that the corresponding REM association also became stronger postoperatively is hard to explain. Just as respiratory events were observed to be uncommon during slow wave sleep (SWS – sleep stages 3 and 4) in this investigation, it

193 is my impression that arousals were similarly less common than in other sleep stages. Specific data on this could not be extracted from the recordings due to limitations in the Replay software. It is therefore possible that differences in the amount of SWS might have contributed to these observations.

The apparent reduction in the number of respiratory events that did not result in arousal on the first postoperative night would be consistent with greater arousability as a result, perhaps, of pain. This is in turn consistent with the alteration in the postoperative electroencephalogram (EEG) as discussed in the preceding chapters. On the other hand, auditory stimuli have been shown to produce respiratory responses without clear association to EEG arousal (Carley 1997). A variety of perioperative stimuli, including sound and pain, might alter respiratory drive without producing arousal. Depending on the balance of this effect on the individual’s respiratory pump and upper airway function, this could either transiently increase ventilation, or possibly make obstruction worse, and contribute to postoperative breathing difficulties.

Arousals from sleep may occur for a variety of reasons, especially in the postoperative setting. Pain, noise and nursing observations are but a few examples of the potential non-respiratory causes of arousal. While an increase in the number of respiratory event- related arousals might be expected on postoperative nights where opioids are required, the finding by Wu and Drummond (Wu 2003) of a high proportion of non-respiratory arousals is not entirely unexpected, if all the other possible reasons for arousal are considered. This notwithstanding, the findings of this investigation differ in many respects from those of Wu and Drummond. A much higher proportion of respiratory event-related arousals was found in the present study. Although statistically not significant, there was also a higher number of arousals after respiratory recovery compared with before in their study (figure 10.3). In this present investigation, the vast majority of arousals occurred either immediately before or coincident with the termination of the respiratory event. There are a number of possible explanations for these differences.

The methodologies of the two studies are quite different. Arousals and respiratory events were not scored blind to each other in this study for reasons that have been discussed in previous chapters. There is clearly a possibility that the results of this study

194 were affected by bias that would produce a closer association between respiratory events and arousals as was found. On the other hand, while blinding eliminates the risk of bias, it may introduce a different problem, that of potential failure to recognise real events, again for reasons already discussed. This possibility appears to be illustrated in figure 3 of Wu’s paper (reproduced in figure 10.4 below). The legend of the figure states that there is no change in breathing pattern when I believe there is. Both the respiratory rate and amplitude increase at a time corresponding to an obvious arousal in the EEG, indicating a likely respiratory event-related arousal that appears to have gone unrecognised. There are, therefore, potential deficiencies in both methodologies and these issues need to be addressed in future studies.

Another possible reason is the age, sex and surgical intervention of the subjects. All subjects in Wu’s study were females having lower abdominal surgery. Those in the present study were mainly males, older and therefore more likely to have both higher numbers of arousals and more respiratory events (Ancoli-Israel 1994; Mathur 1995). While it is difficult to compare the surgical procedures and the degree of pain involved, the subjects in Wu’s study used a greater amount of opioid analgesia and this may also have been a factor. In the present investigation, however, higher opioid usage appeared to result in a closer association between respiratory events and arousals that did not occur in their study.

The use of patient controlled analgesia (PCA) in Wu’s study could be important. Patients using PCA frequently report that they drift off to sleep shortly after a demand, only to awaken soon thereafter requiring further analgesia as the opioid bolus is relatively small, possibly resulting in only short-lived analgesia. Oral and subcutaneous opioid regimes may avoid this by creating a depot of opioid, resulting in more continuous analgesia and thus more stable conditions for sleep. This hypothesis should be subjected to further research, one possibility being to investigate the relationship between PCA demands and arousals.

195 Reproduction of Figure 7 from Wu and Drummond

FIGURE 10.3. Reproduction of figure 7 from Wu and Drummond (Wu 2003) (used with permission). The figure indicates their finding of the greatest number of arousals (top section) unrelated to a respiratory event (the bottom section deals with full transitions to wakefulness). Arousal associated with respiratory events appeared to occur more often after relief of respiratory obstruction although this was not statistically different from those occurring before. In the present study, the majority of arousals were associated with respiratory events and occurred before or coincident with recovery from the event, rather than after.

196 Reproduction of Figure 3 from Wu and Drummond

FIGURE 10.4a. An annotated reproduction of figure 3 from Wu and Drummond (Wu 2003) (used with permission, legend continues over).

197 Figure 10.4 legend continued (see also figures 7.8a & b). The original legend read, “Fig. 3. An example of a short period of arousal from sleep, indicated by the solid bar, with no change in breathing pattern. The traces are two electroencephalogram channels, a submental electromyogram (EMG), and nasal flow.” The section I have marked with the hollow bar is where a clear arousal occurs. The arrows above the nasal pressure trace have also been added to the original figure. The breath marked “A” is clearly flattened compared with the one preceding it, a feature indicating obstruction, as stated by the authors and illustrated in another figure of the manuscript, also reproduced below (10.4b). The breath marked “B” is further reduced in amplitude and this coincides with the onset of what I believe to be the true arousal, following which there is a markedly larger breath, labeled “C”.

FIGURE 10.4b. See legend above (reproduced with permission). The original legend read, “Fig. 1. Nasal cannular pressure. An example of recovery from partial respiratory obstruction: At the arrow, there is an abrupt change in signal amplitude, associated with an increase in frequency and a change in waveform. The preceding waveform is flattened.”

198 Conclusion

In this investigation, postoperative arousals were more commonly associated with respiratory events than not, especially in the presence of known or suspected sleep- disordered breathing. These findings are somewhat different to the results of previous studies although methodological issues are almost certainly a factor in this difference.

Future investigations might examine the role of age, pain, opioids and mode of analgesia delivery in the occurrence of postoperative arousals.

199 CHAPTER 11

POTENTIAL CONTRIBUTION OF THE ANAESTHETIST TO THE DIAGNOSIS OF SLEEP-DISORDERED BREATHING

Anaesthesia management for patients with known obstructive sleep apnoea (OSA) is relatively common and it is a topic of interest in published articles, at meetings and as the subject of examination questions. Few anaesthetists, if any, are unaware of the possible implications the diagnosis has for airway management and postoperative analgesia in particular, although much of the discussion on the consequences of OSA has been speculative (Loadsman 2001). In contrast, almost nothing has been said about the potential contribution anaesthetists can make to the diagnosis and management of patients presenting for surgery with undiagnosed OSA, despite the likelihood that this latter group is almost certainly far more common than diagnosed patients (Young 2002).

In 1998, Hiremath et al. (Hiremath 1998) reported a series of sleep studies performed on patients found to have difficult intubations as defined by Cormack & Lehane (grade 4 laryngoscopy) (Cormack 1984). 8/15 (53%) had OSA and 5 of these (33% overall) were severe by their definition. In contrast, 2/15 controls (13%) had OSA with none being severe, figures consistent with epidemiological data (Loadsman 2001; Young 2002). While snoring has been investigated as one possible predictor of difficult mask ventilation (Langeron 2000), there are few data on the incidence of OSA among patients presenting difficulty with spontaneous ventilation via a natural airway under general anaesthesia. There is only one recent report indicating that anaesthetised patients requiring continuous positive airway pressure (CPAP) to prevent obstruction are more likely to suffer from OSA than those who did not need CPAP (Eastwood 2002). CPAP, used in that study as an investigative tool, is not commonly employed during anaesthesia and is therefore of limited value in the anaesthetic setting as a means to predict OSA. Anaesthesia providers tend to use alternative methods to deal with

200 obstruction prone airways and the predictive value of these techniques for OSA has not been reported to date. Nevertheless, the findings of Eastwood et al. (Eastwood 2002) suggest that patients with airways prone to collapse under anaesthesia are also prone to have OSA and should therefore be referred for investigation. The present study, a prospective audit of the author’s clinical practice, was carried out to confirm whether or not this hypothesis was true and to identify patients most likely to benefit from such referral.

Methods

A single anaesthetist was involved so the general anaesthetic technique was essentially the same for all patients. Induction was commenced with 1 – 1.5mg/kg of thiopentone intravenously (a few patients required more) with or without a small amount of morphine (0.02 – 0.03mg/kg) and completed with the patient spontaneously breathing sevoflurane in oxygen. When adequately anaesthetised, a laryngeal mask airway (LMA) was inserted, without interruption of the spontaneous ventilation, into those patients not requiring neuromuscular blockade for the surgery. For patients requiring neuromuscular blockade, this was induced and intubation was carried out after a period of manual mask ventilation. In either case, if difficulty with airway maintenance occurred prior to LMA insertion or neuromuscular blockade with manual ventilation, the necessity for intervention at this time was noted as defined below. Patients with complete or partial dentures had those routinely removed prior to their arrival in the operating suite.

Patients without prior suspicion of OSA requiring a pharyngeal airway or jaw-thrust (group One) or both of these manoeuvres (group Both) to prevent severe upper airway obstruction under general anaesthesia were approached and advised of the possibility that this might indicate the presence of OSA. They were offered a standard diagnostic sleep study with polysomnography (PSG). PSG includes electroencephalography, electro-oculography and electromyography for assessment of sleep, nasal airflow transducer and thoracic and abdominal strain gauges for respiration as well as oximetry. If they declined PSG an alternative study with overnight oximetry (OOS) only was then offered. All such patients, irrespective of their decision to undergo a study or not, were informed of the potential for OSA and offered referral to a sleep physician. Some who

201 have taken up the offer of referral, including one patient with an oximetry study suggesting severe apnoea, then went on to have sleep studies organised by the physician and their results, where available, were included with the patients’ permission. In the case where both types of study were carried out, only the data from the full PSG were included.

No attempt was made to collect data on a control group as this was a preliminary investigation seeking to support the findings of Eastwood et al. (Eastwood 2002) while identifying clinically relevant airway maintenance manoeuvres under anaesthesia that might warrant further study for their likelihood of indicating the presence of OSA. Data were collected, however, on a series of 48 consecutive patients unrelated to the main group studied herein to ascertain the approximate incidence of difficult airways as defined below. This would allow a very limited comparison of the data obtained from this study with population estimates of OSA.

Definitions

Severe airway obstruction was defined as complete or almost complete obstruction despite neck extension and chin lift, requiring jaw-thrust, pharyngeal airway or both to overcome. For the purposes of this investigation, standard definitions of apnoeas and hypopnoeas (complete or partial airway obstruction of 10 seconds or more associated with arousal or desaturation of 3% or more) and OSA severity (mild = apnoea/hypopnoea index (RDI) 5 – 14 per hour of sleep, moderate = RDI 15 – 29/h, severe = RDI ≥ 30/h) were used (AASM 1999). In the case of oximetry studies, measures of airflow, respiratory effort and sleep were, by definition, unavailable. A desaturation of 3% or more was therefore used with limitations as discussed below.

Statistical Analysis

Comparison of groups One and Both for the severity of their OSA, body mass index and age was performed using the two-tailed Mann-Whitney U test. P < 0.05 was considered significant.

202 Results

Twenty-five patients (22 men, 3 women) were investigated. One male having an OOS noted that he slept very poorly on the study night as a result of noise from another patient in his room and this was confirmed by severe artifactual interruption of the oximetry trace throughout the night and the study was deemed too unreliable for analysis. He declined further investigation. Twenty-four are therefore included in the analysis, fifteen undergoing full PSG and nine having OOS. All patients exhibited some degree of OSA ranging from trivial to severe. Of the PSG group, two had mild OSA, four fell into the moderate range and nine were severe sufferers. In the OOS group, five studies suggested mild OSA, two moderate and two severe. Overall therefore, 7/24 (29%) had mild OSA, 6/24 (25%) were moderate and 11/24 (46%) severe. All three women had mild OSA.

Of those patients in group One (n=12), seven had mild OSA, three were moderate and two were severe. Of those requiring both manoeuvres (n=12), none had mild OSA, three were moderate, and nine were severe. Individual RDIs for these groups are shown in figure 11.1 and it can readily be seen that group Both had significantly more severe OSA than group One. Statistical comparison of the groups’ RDIs confirms this (P = 0.0005).

Four patients (one in group One and three in group Both) did not have both their height and weight measured/recorded. Using the data available there appears to be a tendency for those in group Both to be more obese (figure 11.2) however this did not reach statistical significance (P = 0.26). Figure 11.3 shows the ages of patients in the groups and these are clearly not different (P = 0.62).

203 Severity of OSA One Airway Manoeuvre versus Both

45 severe

30 moderate 15 mild 5 0 One Both Airway Manoeuvres Used

FIGURE 11.1. Respiratory Disturbance Indices (RDI – events per hour) for the group requiring either jaw-thrust or a pharyngeal airway (One) and for those requiring both manoeuvres (Both).

204 Body Mass Indices One Airway Manoeuvre versus Both

FIGURE 11.2. Body Mass Indices (BMI – kg/m2) for the group requiring either jaw-thrust or a pharyngeal airway (One) and for those requiring both manoeuvres (Both). Four patients (one in group One and three in group Both) did not have both their height and weight measured/recorded and are therefore not included in this graph.

205 Age One Airway Manoeuvre versus Both

FIGURE 11.3. Ages for the group requiring either jaw-thrust or a pharyngeal airway (One) and for those requiring both manoeuvres (Both).

206 To estimate the approximate incidence of difficult airways, data were collected on 48 consecutive patients anaesthetised using spontaneous breathing techniques. None of these had a prior diagnosis of OSA. 5 (10%) were found to have difficult airways as defined in this study.

Discussion

Limitations of the Study

Overnight oximetry is limited in its diagnostic value for sleep disordered breathing. Oximetry artifact could potentially be interpreted as desaturation, artificially increasing the RDI as apnoeas and hypopnoeas cannot be confirmed by measurement of airflow/effort. To some extent this error may be limited by the recognition of the characteristic sawtooth pattern of desaturation due to recurrent obstruction when the OSA is moderate to severe. Conversely, total sleep time cannot be measured either so the total study period must be used as the denominator for the RDI. As some of this time will be unrecognised wakefulness, the RDI will be underestimated. It is clearly impossible to know how much these limitations contributed to the results of this study but it is likely that most patients undergoing OOS had their RDI underestimated, particularly the more severe cases as their oximetry patterns are easier to interpret. Milder cases, on the other hand, are perhaps more likely to have tended towards overestimation. It is worth noting that, in the case of the one patient who underwent both types of study, the results of the oximetry night agreed quite closely with that of the full PSG, an RDI in the forties identified by both studies.

The investigation is also subject to a number of possible biases. The airway difficulties were identified and assessed by a single anaesthetist who also carried out and scored a number of the subsequent sleep studies. While therefore unblinded, the studies were scored prior to any attempt to stratify the severity of the airway obstruction under anaesthesia i.e. the division of the subjects into groups One and Both. Also, ten of the fifteen polysomnographic studies (four of group One and six of group Both) were performed and reported by independent sleep laboratories, one of which quite closely confirmed the original OOS findings for that patient.

207 All the OOS studies and the five PSG studies not performed by independent sleep laboratories were carried out during the subjects’ surgical admissions. As far as possible (all but one subject) these were performed on nights when no potentially confounding medications such as opioids were administered. Nevertheless, it is clearly possible that altered postoperative sleep architecture and residual drug effects may have influenced the studies. The sleep architecture of the PSG studies was, however, relatively normal and all patients underwent relatively minor surgical procedures, after which sleep disruption is not likely to be severely affected (Rosenberg-Adamsen 1996b). Again, it is worth noting that the OOS performed within days of a groin lymph node dissection for melanoma, agreed almost perfectly in terms of apnoea severity with a PSG performed weeks later in an independent laboratory.

The lack of a control group in this study precludes the calculation of the sensitivity and specificity of difficult airways as defined above to predict OSA and further trials will be necessary for this purpose. This limitation aside, positive predictive values can be measured and as OSA has a high community prevalence (around 20% of adults for at least mild OSA and more than 5% for moderate to severe OSA) these are potentially useful data. Ten per cent of the ‘prevalence’ group studied were identified as having difficult airways. This is consistent with population data concerning OSA and while it in no way compensates for the lack of a control group, it does provide a suggestion that the sensitivity and specificity of these airway maintenance manoeuvres for the detection of OSA might be quite good. This is further impetus for larger, more rigorous studies.

The presence or lack of teeth was not recorded and this was an oversight. Edentulous patients may be more prone to difficult airway maintenance under anaesthesia. Most people sleep without their dentures, however, so it is unlikely that this will have had an impact on the findings. Future studies, where possible, might identify the usual nocturnal habits of the patient in this respect and continue those for the purposes of airway assessment.

208 Implications

Based on the data of Hiremath et al. (Hiremath 1998), Eastwood et al. (Eastwood 2002) and the findings of this investigation it should be recognised that both difficult intubation and airways that are difficult to maintain during spontaneous breathing under general anaesthesia strongly suggest OSA, although it is possible there is a difference between the sexes.

The most important findings of this study, while bearing the limitations in mind, may be summarised as follows:

1. All patients identified as requiring either jaw-thrust or a pharyngeal airway to maintain patency of their airway while breathing spontaneously under general anaesthesia had OSA of at least mild severity (positive predictive value = 100%), although a few had trivial disorders which probably do not warrant therapy.

2. All patients requiring both manoeuvres to maintain their airway under anaesthesia had at least moderate OSA (positive predictive value 100%) and three quarters had severe OSA (positive predictive value 75%).

The small number of females identified for this study, their lower likelihood of having the most difficult to maintain airways (jaw-thrust or pharyngeal airway required rather than both manoeuvres) and their lack of subsequently identified severe OSA, probably reflect the lower prevalence of OSA amongst women in community prevalence studies (Young 1993c). Given the small numbers involved, it is very difficult to say whether or not the results of this study apply to women in general. Until larger studies of that subgroup can be performed some caution is required when suggesting the possibility of OSA to women who present with difficult airways. Nevertheless, it would be logical to assume that, as with men, those women with very obstruction-prone airways under anaesthesia are at risk of having OSA and ought to be referred for investigation.

While the lack of controls makes it impossible to say that the absence of airway maintenance difficulty under anaesthesia predicts a low risk of OSA, this limitation must be put into the perspective of current clinical practice. At the present time, very

209 few anaesthetists make any attempt to inform patients of the possibility or otherwise that they might have OSA based on the behaviour of the airway perioperatively. Moreover, the collapsibility of the airway under anaesthesia is an observation we make everyday as part of the routine clinical practice of anaesthesia, not a test specifically performed to confirm or rule out the diagnosis of sleep apnoea. The issues of sensitivity and specificity are therefore of limited clinical relevance. The data herein very strongly suggest that patients with airways prone to collapse under anaesthesia are similarly prone to obstruct during sleep. Moreover, those with the worst airways, requiring both jaw-thrust and a pharyngeal airway, are highly likely to suffer moderate to severe sleep apnoea with all its potential consequences. By using this information to inform our patients appropriately, something which rarely happens at present, there is everything to be gained and almost nothing to lose, apart perhaps from performing the occasional negative diagnostic sleep study. Even this might be avoided frequently by further clinical assessment and appropriate selection by sleep physicians.

There are important perioperative implications of this finding as well. It is possible that patients with moderate to severe OSA are more likely to suffer complications and difficulties in the postoperative setting, although the evidence for this is largely limited to speculation based on case reports. One of the main problems to date has been the preoperative identification of individuals most at risk and this has still not been resolved. The findings of this study, however, point to a group of patients who may well fall into this category, allowing us to take preventative measures such as high dependency care after major surgery for patients who would otherwise have not been identified as moderate to severe OSA sufferers.

There is still much information yet to be acquired, including more detailed investigation of the manoeuvres examined herein. There is, as yet, no information at all available on the ability or otherwise of airway maintenance difficulties under anaesthesia with neuromuscular blockade to predict OSA. No attempt has been made to systematically study patients who are found to be prone to develop airway obstruction in the recovery room or post-anaesthesia care unit. Difficult deployment of various other airway devices such as the laryngeal mask may also indicate OSA.

210 OSA has potentially serious neuropsychological, vascular and cardio-pulmonary sequelae as well as other endocrine and gastrointestinal consequences (Loadsman 2001). Of main concern is the risk of accident due to sleepiness and the long term risk of hypertension, cardiac and cerebral ischaemia or infarction, and respiratory failure. It is also very treatable with a number of effective therapeutic options available. Anaesthetists have some responsibility, therefore, to inform patients falling into these categories of the likelihood that they suffer from OSA and, where possible, to offer them referral to appropriate practitioners. Failing this, we should at least suggest to them that they consider follow-up with their GP who may also be informed via the usual hospital discharge summary.

In developed nations, around 10% of the adult population have a general anaesthetic every year and anaesthetists have, for well over a century, been dealing with difficult airways in the unconscious individual. This places anaesthetists in the position of having unrivalled skill and opportunity to identify large numbers of people that might otherwise have no idea of the possibility they suffer from OSA. It is an opportunity and a responsibility we must not ignore, both for the immediate and long term care of our patients.

At the time of writing, at least seven of the patients with data presented herein have gone on to have a trial of CPAP therapy for their OSA and one other is awaiting a CPAP titration study. Most of these are now settled on CPAP with a notable improvement in their subjective quality of life.

211 CHAPTER 12

CONCLUSIONS

Well? What indeed of Hypnos, god of sleep and patron of anaesthesia? John Severinghaus may not have realised the implications of his question. Is Hypnos a villain, deceiving us in his patronage of our art by trying to blow out the respiring flame Severinghaus set before him (Nunn 1969)? Thus it would seem, but perhaps his breath is not as strong as some have suggested.

Unless direct evidence is found, an important role of rapid eye movement (REM) sleep rebound in the causation of late postoperative morbidity and mortality must be questioned. Examination of the relevant literature and the results of the limited assessment of postoperative sleep presented here both suggest postoperative REM rebound to be an inconsistent phenomenon. If and when rebound of REM sleep does occur, the additional few percent of total sleep time it encompasses is unlikely, for the vast majority of patients at least, to result in a clinically significant increase in the number and severity of obstructive or centrally mediated hypoxaemic events. The finding presented in chapter 3 supports this. Significant REM-predominance of obstructive sleep apnoea (OSA) is uncommon. It is possible that a few individuals, morbidly obese with severe pre-existing OSA most probably, might be exposed to additional risk of sleep-related respiratory compromise as a consequence of REM sleep rebound, and this subgroup warrants further examination. Even that risk, however, is likely to be almost inconsequential compared to the risk to which those same patients might be exposed by the use of heavy opioid analgesia in the early postoperative period. This view is further supported by the findings of chapter 4 that most unexpected deaths occur in the day or two after surgery, with no circadian variation.

The role of opioid analgesia in the disturbance of sleep architecture, particularly the reputed suppression of REM, similarly warrants further examination. The observation in chapter 6, that more REM occurred after the administration of opioid in one subject,

212 runs counter to previous reports, although other recent data, that REM suppression occurs postoperatively without opioid analgesia, may be more in agreement (Cronin 2001). Detailed examination of other factors, endotoxins for example (Trachsel 1994), that might influence the macro-architecture of postoperative sleep is therefore required.

Also observed in the preoperative studies (chapter 6) was the higher amount of slow wave sleep and delta electroencephalogram activity in subjects with chronic respiratory disease. The implications are unclear and as it appears this observation has not been reported before it seems logical that it should be investigated further.

The high prevalence of sleep apnoea, some of which was severe, amongst a group of non-obese but somewhat elderly patients selected at random on presentation for surgery, was initially surprising. While this is consistent with prevalence studies for the general aged population, it is almost certainly a revelation to anaesthetists. The implication is that we are anaesthetising large numbers of patients with undiagnosed and often severe OSA, many with co-morbidity, without any adverse consequences and without even realising in most instances. This suggests the overall perioperative risk of morbidity for OSA sufferers is lower than might be expected. Nevertheless, there may be subgroups of patients, such as the obese, for which the risk is substantial. It is important that we identify these subgroups, both to alleviate the additional risk and to avoid the expenditure of scarce resources unnecessarily.

Postoperative studies in the present investigation confirm previous findings that sleep macro-architecture is altered after surgery in a variable fashion, mostly dependent on the type of surgical procedure and degree of postoperative discomfort although many other factors clearly play a role. Major surgery invariably results in poor postoperative sleep, even in the presence of adequate analgesia. More importantly, however, there are changes in the micro-architecture of the postoperative sleep electroencephalogram (EEG) that may not only distinguish it from normal sleep but also result in misinterpretion of postoperative studies. These changes, which seem mainly to be an increase in alpha and/or beta activity, appear to be relatively consistent after a significant surgical insult and may be the result of pain and/or opioid administration. The ambiguity these changes introduce make the staging of postoperative sleep studies using standard criteria difficult and potentially unreliable. It should not be assumed that

213 the EEG and electromyographic characteristics previously defined for normal sleep adequately define sleep in the postoperative setting, a limitation potentially affecting the results of both this and other studies. Further definition of the behavioural and electrophysiological characteristics of sleep after surgical stress is therefore required. However, the behavioural correlates of both sleep and deep sedation resulting from opioid analgesia are likely to be so similar that investigation of this issue may be difficult and contentious.

The standard but arbitrary division of polysomnograms into 30-second epochs is also a problem for postoperative studies in which arousal is common. There is little doubt that this has hampered the ability to adequately describe postoperative sleep, both in this study and others. Indeed this may be the main factor underlying the findings of some investigations. There is no obvious and simple solution to this problem but experiments using shorter epochs might identify more satisfactory methods.

The importance of daytime sleep periods after surgery, both for their effect on postoperative sleep architecture in general and for their potential role in the generation of episodic hypoxaemic events, has not been adequately examined. It was not possible to address this issue in the present study. Future studies with less intrusive forms of monitoring may meet with more success.

As with the electrophysiologic definition of sleep and its stages and the use of arbitrarily defined epoch lengths, the use of standard sleep medicine classifications for respiratory events in the postoperative setting presents potential difficulty, both in terms of their recognition and in assessment of their significance. This is another limitation that may have affected the results of this present investigation as well as those previously reported. While an oxygen desaturation of a few percent, resulting in arousal, may have long-term health implications if repeated often enough for many years, the importance of relatively minor episodic events after surgery is currently unknown. Such events are very common, much more so than the profound events originally described by Catley et al. (Catley 1985). It is likely that Catley’s observation regarding these profound events, that they all occurred during sleep, similarly applies to most episodic postoperative respiratory events, profound or otherwise (chapter 9). The severity criteria for sleep apnoea syndromes have been defined according to their likelihood of

214 producing morbidity. Similarly, we need to examine the number and degree of physiological disturbances from respiratory events occurring after surgery, perhaps with respect to what is ‘normal’ for each patient, and classify severity in terms of the likelihood of causing postoperative morbidity.

An important observation in this investigation was the number of respiratory events associated with sleep onset (chapters 6, 7 and 9). Both in this study and others an increase in the amount of stage 1 sleep has been observed postoperatively. Moreover, major surgery is associated with a substantial postoperative increase in the amount of time spent in the state of drowsiness seen prior to the onset of stage 1 sleep as classified by standard criteria, in which ventilatory instability is common. The degree to which one subject exhibited this phenomenon of sleep-onset respiratory instability, leading itself to an almost incredible degree of sleep disruption, was extraordinary and totally unexpected. It is not clear which, if any, postoperative factors contributed to this remarkable observation, but it is possible that pain and/or opioid consumption increased the gain of his ventilatory feedback loop to the point of uncontrolled sleep onset-related oscillation. The possibility that some patients, perhaps with normally high peripheral chemoreceptor drive, might be susceptible to such postoperative ventilatory perturbations begs further examination. The potential for preoperative identification of such patients, who may be at significant postoperative risk of morbidity, also warrants investigation.

The majority of arousals from sleep on the first postoperative night in this investigation were related to respiratory events – more than two thirds overall and more than 80% where the subject was suspected of having OSA. While this differed from the findings of Wu and Drummond (Wu 2003), there are aspects of methodology in both studies that are potentially problematic and this may have contributed to the difference. The present study benefited, I believe, from the contemporaneous examination of sleep and respiratory variables thus allowing the confirmation of sleep by surrogate behavioural markers and vice-versa but this introduced significant potential bias. Wu and Drummond eliminated bias but in doing so also eliminated respiration as a surrogate behavioural marker, thus introducing potential inaccuracy into the staging of postoperative sleep where ambiguity is a significant problem. These potential methodological deficiencies need to be addressed in future studies. Setting such issues

215 aside, these studies may have highlighted other factors that may influence the number of arousals and thus the amount of sleep disruption after surgery. It is possible, for example, that patient controlled analgesia (PCA) might actually increase the degree of sleep disturbance as a consequence of the need for frequent activation of the PCA device to achieve analgesia. Future investigations, therefore, might examine the role of factors such as age, pain, opioids and mode of analgesia delivery in the occurrence of postoperative arousals.

Moderate to severe OSA, with all its consequent morbidity, appears to be almost universal amongst patients with marked upper airway obstruction to spontaneous respiration under general anaesthesia, as demonstrated in chapter 11, although more rigorous studies are required to confirm this and other potential associations. Is this an opportunity for anaesthetists to atone for the treachery of Hypnos? This is but one observation, one which most anaesthetists manage so easily it is ignored, that will almost certainly make a substantial difference to the long-term wellbeing of many of our patients if acted upon through referral to sleep physicians. Upper airway obstruction under anaesthesia and difficulty with intubation also identifies patients likely to have a greater degree of postoperative sleep-related respiratory compromise (chapter 9).

Have we misjudged Hypnos? Just as respiration is necessary for life, so is sleep. The findings of this research suggest that, in the future, both clinical and research resources should be directed more towards the resolution of issues regarding sleep and sleep- disordered breathing in the early postoperative period. Thus we have and will improve further our knowledge of the process. Anaesthetists need to be aware that undiagnosed sleep apnoea may be identified with perioperative observations, and, while this also requires further research, we may better our art immediately by making use of this information to help our patients. The disharmony in Hypnos’ patronage and deity has been alleviated, but the balance remains somewhat precarious. Severinghaus’ exhortations have not been in vain and his flame still burns before the god of sleep.

216 REFERENCES

AASM (1999). AASM taskforce report: Sleep-related breathing disorders in adults: recommendations for syndrome definition and measurement techniques in clinical research. Sleep 22: 667-689. ABOM (1999). Australian Bureau of Meteorology archive of daily minimum and maximum temperatures recorded at Sydney weather station. ftp://ftp.bom.gov.au/anon/home/bmrc/perm/climate/temperature/daily/qc/sydney . Accessed 2004. Albert PS (1999). Longitudinal data analysis (repeated measures) in clinical trials. Statistics in Medicine 18: 1707-32. Altman DG (1999). Practical Statistics for Medical Research. Boca Raton, Chapman & Hall. Ambrogetti A, Olson LG, Saunders NA (1991). Differences in the symptoms of men and women with obstructive sleep apnoea. Australian & New Zealand Journal of Medicine 21(6): 863-6. Ancoli-Israel S, Bliwise D, Mant A (1994). Sleep and breathing in the elderly. In: Sleep and Breathing. N. A. Saunders and C. E. Sullivan, Eds. New York, Marcel Dekker, Inc.: 673-694. Ancoli-Israel S, Klauber MR, Stepnowsky C, Estline E, Chinn A, Fell R (1995). Sleep- disordered breathing in African-American elderly. American Journal of Respiratory & Critical Care Medicine 152(6 Pt 1): 1946-9. Aronow WS & Ahn C (2003). Circadian variation of death from congestive heart failure after prior myocardial infarction in patients >60 years of age. American Journal of Cardiology 92(11): 1354-5. ASDA (1992). ASDA report: EEG arousals: scoring rules and examples. Sleep 15(2): 173-184. ASDA (1994). Practice parameters for the use of laser-assisted uvulopalatoplasty. Standards of Practice Committee of the American Sleep Disorders Association. Sleep 17(8): 744-8. ASDA (1995). Practice parameters for the treatment of snoring and obstructive sleep apnea with oral appliances. American Sleep Disorders Association. Sleep 18(6): 511-3. ASDA (1996). Practice parameters for the treatment of obstructive sleep apnea in adults: the efficacy of surgical modifications of the upper airway. Report of the American Sleep Disorders Association. Sleep 19(2): 152-5. Aserinsky E & Kleitman N (1953). Regularly occurring periods of eye motility and concomitant phenomena during sleep. Science 118: 273. Aspinall RL & Mayor A (2001). A prospective randomized controlled study of the efficacy of ketamine for postoperative pain relief in children after adenotonsillectomy. Paediatric Anaesthesia 11(3): 333-6. Audit B & Ouzounis CA (2003). From genes to genomes: universal scale-invariant properties of microbial chromosome organisation. Journal of Molecular Biology 332(3): 617-33. Aurell J & Elmqvist D (1985). Sleep in the surgical intensive care unit: continuous polygraphic recording of sleep in nine patients receiving postoperative care. British Medical Journal Clinical Research Ed. 290(6474): 1029-32.

217 Ayappa I & Rapoport DM (2003). The upper airway in sleep: physiology of the pharynx. Sleep Medicine Reviews 7(1): 9-33. Bacchi A & Pelizzi G (1999). Conformational variety for the ansa chain of rifamycins: comparison of observed crystal structures and molecular dynamics simulations. Journal of Computer-Aided Molecular Design 13(4): 385-96. Badr MS, Skatrud JB, Dempsey JA, Begle RL (1990). Effect of mechanical loading on expiratory and inspiratory muscle activity during NREM sleep. Journal of Applied Physiology 68(3): 1195-202. Ballard RD, Saathoff MC, Patel DK, Kelly PL, Martin RJ (1989). Effect of sleep on nocturnal bronchoconstriction and ventilatory patterns in asthmatics. Journal of Applied Physiology 67(1): 243-9. Bauer G (1993). EEG, drug effects, and central nervous system poisoning. In: Electroencephalography: basic principles, clinical applications, and related fields. E. Niedermeyer and F. L. da Silva, Eds. Baltimore, Maryland, Williams & Wilkins: 631-642. Bay J, Nunn JF, Prys-Roberts C (1968). Factors influencing arterial PO2 during recovery from anaesthesia. British Journal of Anaesthesia 40(6): 398-407. Bearpark H, Elliott L, Grunstein R, Cullen S, Schneider H, Althaus W, Sullivan C (1995). Snoring and sleep apnea. A population study in Australian men. American Journal of Respiratory & Critical Care Medicine 151(5): 1459-65. Bellamy N, Sothern RB, Campbell J, Buchanan WW (2002). Rhythmic variations in pain, stiffness, and manual dexterity in hand osteoarthritis. Annals of the Rheumatic Diseases 61(12): 1075-80. Benumof JL (2001). Obstructive sleep apnea in the adult obese patient: implications for airway management. Journal of Clinical Anesthesia 13(2): 144-56. Berry RB & Gleeson K (1997). Respiratory arousal from sleep: mechanisms and significance. Sleep 20(8): 654-75. Berssenbrugge A, Dempsey J, Iber C, Skatrud J, Wilson P (1983). Mechanisms of hypoxia-induced periodic breathing during sleep in humans. Journal of Physiology 343: 507-26. Bhalla A, Singh R, Sachdev A, D'Cruz S, Duseja A (2002). Circadian pattern in cerebro vascular disorders. Neurology India 50(4): 526-7. Bhalla A, Sood A, Mahapatra M, D'Cruz S, Singh R (2001). Circadian pattern of cardiovascular and cerebrovascular diseases in geriatric population. Journal of the Association of Physicians of India 49: 1066-9. Bilora F, Manfredini R, Petrobelli F, Vettore G, Boccioletti V, Pomerri F (2001). Chronobiology of non fatal pulmonary thromboembolism. Panminerva Medica 43(1): 7-10. Bisgaard T, Klarskov B, Kehlet H, Rosenberg J (2002). Recovery after uncomplicated laparoscopic cholecystectomy. Surgery 132(5): 817-25. Bixler EO, Vgontzas AN, Ten Have T, Tyson K, Kales A (1998). Effects of age on sleep apnea in men: I. Prevalence and severity. American Journal of Respiratory & Critical Care Medicine 157(1): 144-8. Black S, Mahla ME, Cucchiara RF (1994). Neurologic Monitoring. In: Anesthesia. R. D. Miller, Ed. New York, Churchill Livingstone, Inc. 1: 1322-1325. Bliwise DL (1989). Normal Aging. In: Principles and Practice of Sleep Medicine. M. Kryger, R. Roth and W. C. Dement, Eds. Philadelphia, WB Saunders: 24-29. Boushra NN (1996). Anaesthetic management of patients with sleep apnoea syndrome. Canadian Journal of Anaesthesia 43(6): 599-616.

218 Bradley TD (1992a). Right and left ventricular functional impairment and sleep apnea. Clinics in Chest Medicine 13(3): 459-79. Bradley TD & Phillipson EA (1992b). Central sleep apnea. Clinics in Chest Medicine 13(3): 493-505. Brooks B, Cistulli PA, Borkman M, Ross G, McGhee S, Grunstein RR, Sullivan CE, Yue DK (1994). Obstructive sleep apnea in obese noninsulin-dependent diabetic patients: effect of continuous positive airway pressure treatment on insulin responsiveness. Journal of Clinical Endocrinology & Metabolism 79(6): 1681-5. Burgess KR (1997). Central sleep apnoea and heart failure (Part I). Respirology 2(4): 243-53. Burgess LP, Derderian SS, Morin GV, Gonzalez C, Zajtchuk JT (1992). Postoperative risk following uvulopalatopharyngoplasty for obstructive sleep apnea. Otolaryngology - Head & Neck Surgery 106(1): 81-6. Byrne M, Dacke M, Nordstrom P, Scholtz C, Warrant E (2003). Visual cues used by ball-rolling dung beetles for orientation. Journal of Comparative Physiology A- Sensory Neural & Behavioral Physiology 189(6): 411-8. Canham PB, Finlay HM, Boughner DR (1997). Contrasting structure of the saphenous vein and internal mammary artery used as coronary bypass vessels. Cardiovascular Research 34(3): 557-67. Canham PB, Finlay HM, Dixon JG, Ferguson SE (1991a). Layered collagen fabric of cerebral aneurysms quantitatively assessed by the universal stage and polarized light microscopy. Anatomical Record 231(4): 579-92. Canham PB, Talman EA, Finlay HM, Dixon JG (1991b). Medial collagen organization in human arteries of the heart and brain by polarized light microscopy. Connective Tissue Research 26(1-2): 121-34. Carley DW, Applebaum R, Basner RC, Onal E, Lopata M (1997). Respiratory and arousal responses to acoustic stimulation. Chest 112(6): 1567-71. Carlson JT, Hedner J, Fagerberg B, Ejnell H, Magnusson B, Fyhrquist F (1992). Secondary polycythaemia associated with nocturnal apnoea - a relationship not mediated by erythropoietin? Journal of Internal Medicine 231(4): 381-7. Carlson JT, Hedner JA, Ejnell H, Peterson LE (1994). High prevalence of hypertension in sleep apnea patients independent of obesity. American Journal of Respiratory & Critical Care Medicine 150(1): 72-7. Carskadon MA & Rechtschaffen A (1989). Monitoring and staging human sleep. In: Principles and Practice of Sleep Medicine. M. Kryger, R. Roth and W. C. Dement, Eds. Philadelphia, WB Saunders: 665-683. Cartwright RD, Diaz F, Lloyd S (1991). The effects of sleep posture and sleep stage on apnea frequency. Sleep 14(4): 351-3. Casetta I, Granieri E, Fallica E, la Cecilia O, Paolino E, Manfredini R (2002). Patient demographic and clinical features and circadian variation in onset of ischemic stroke. Archives of Neurology 59(1): 48-53. Catley DM, Thornton C, Jordan C, Lehane JR, Royston D, Jones JG (1985). Pronounced, episodic oxygen desaturation in the postoperative period: its association with ventilatory pattern and analgesic regimen. Anesthesiology 63(1): 20-8. Chan CS, Grunstein RR, Bye PT, Woolcock AJ, Sullivan CE (1989). Obstructive sleep apnea with severe chronic airflow limitation. Comparison of hypercapnic and eucapnic patients. American Review of Respiratory Disease 140(5): 1274-8. Cheung RT, Mak W, Chan KH (2001). Circadian variation of stroke onset in Hong Kong Chinese: a hospital-based study. Cerebrovascular Diseases 12(1): 1-6.

219 Cistulli PA (1996). Craniofacial abnormalities in obstructive sleep apnoea: implications for treatment. Respirology 1(3): 167-74. Cistulli PA & Sullivan CE (1993). Sleep-disordered breathing in Marfan's syndrome. American Review of Respiratory Disease 147(3): 645-8. Cnaan A, Laird NM, Slasor P (1997). Using the general linear mixed model to analyse unbalanced repeated measures and longitudinal data. Statistics in Medicine 16: 2349–2380. Colmenero C, Esteban R, Albarino AR, Colmenero B (1991). Sleep apnoea syndrome associated with maxillofacial abnormalities. Journal of Laryngology & Otology 105(2): 94-100. Connolly LA (1991). Anesthetic management of obstructive sleep apnea patients. Journal of Clinical Anesthesia 3(6): 461-9. Connolly RA & Baker AB (1995). Excision of bilateral carotid body tumours. Anaesthesia & Intensive Care 23(3): 342-5. Conradt R, Hochban W, Heitmann J, Brandenburg U, Cassel W, Penzel T, Peter JH (1998). Sleep fragmentation and daytime vigilance in patients with OSA treated by surgical maxillomandibular advancement compared to CPAP therapy. Journal of Sleep Research 7(3): 217-23. Cormack RS & Lehane J (1984). Difficult tracheal intubation in obstetrics. Anaesthesia 39(11): 1105-11. Cronin A, Keifer JC, Baghdoyan HA, Lydic R (1995). Opioid inhibition of rapid eye movement sleep by a specific mu receptor agonist. British Journal of Anaesthesia 74(2): 188-92. Cronin AJ, Keifer JC, Davies MF, King TS, Bixler EO (2001). Postoperative sleep disturbance: influences of opioids and pain in humans. Sleep 24(1): 39-44. Cullen DJ (2001). Obstructive sleep apnea and postoperative analgesia--a potentially dangerous combination. Journal of Clinical Anesthesia 13(2): 83-5. Dahan A & Teppema LJ (2003). Influence of anaesthesia and analgesia on the control of breathing. British Journal of Anaesthesia 91(1): 40-9. Dahan A, van den Elsen MJ, Berkenbosch A, DeGoede J, Olievier IC, Burm AG, van Kleef JW (1994a). Influence of a subanesthetic concentration of halothane on the ventilatory response to step changes into and out of sustained isocapnic hypoxia in healthy volunteers. Anesthesiology 81(4): 850-9. Dahan A, van den Elsen MJ, Berkenbosch A, DeGoede J, Olievier IC, van Kleef JW (1995). Halothane affects ventilatory afterdischarge in humans. British Journal of Anaesthesia 74(5): 544-8. Dahan A, van den Elsen MJ, Berkenbosch A, DeGoede J, Olievier IC, van Kleef JW, Bovill JG (1994b). Effects of subanesthetic halothane on the ventilatory responses to hypercapnia and acute hypoxia in healthy volunteers. Anesthesiology 80(4): 727-38. Dahan A & Ward DS (1991). Effect of i.v. midazolam on the ventilatory response to sustained hypoxia in man. British Journal of Anaesthesia 66(4): 454-7. Dahan A & Ward DS (1994c). Effects of 20% nitrous oxide on the ventilatory response to hypercapnia and sustained isocapnic hypoxia in man. British Journal of Anaesthesia 72(1): 17-20. Deegan PC & McNicholas WT (1995). Pathophysiology of obstructive sleep apnoea. European Respiratory Journal 8(7): 1161-78. Delle Karth G, Reinelt P, Buberl A, Geppert A, Huelsmann M, Berger R, Heinz G (2003). Circadian variation in ventricular tachycardia and atrial fibrillation in a medical-cardiological ICU. Intensive Care Medicine 29(6): 963-8.

220 Dement WC & Kleitman N (1957). Cyclic variations in EEG during sleep and their relation to eye movements, body motility, and dreaming. Electroencephalography & Clinical Neurophysiology 9: 673. Dempsey JA & Skatrud JB (1986). A sleep-induced apneic threshold and its consequences. American Review of Respiratory Disease 133(6): 1163-70. Dempsey JA, Skatrud JB, Badr MS, Henke KG (1991). Effects of Sleep on the Regulation of Breathing and Respiratory Muscle Function. In: The Lung: Scientific Foundations. R. G. Crystal, J. B. West and et al., Eds. New York, Raven Press, Ltd.: 1615-1629. Douglas NJ & Polo O (1994). Pathogenesis of obstructive sleep apnoea/hypopnoea syndrome. Lancet 344(8923): 653-5. Drew T & Doucet S (1991). Application of circular statistics to the study of neuronal discharge during locomotion. Journal of Neuroscience Methods 38(2-3): 171-81. Drewes AM, Nielsen KD, Arendt-Nielsen L, Birket-Smith L, Hansen LM (1997). The effect of cutaneous and deep pain on the electroencephalogram during sleep--an experimental study. Sleep 20(8): 632-40. Drummond GB (1989). Influence of thiopentone on upper airway muscles. British Journal of Anaesthesia 63(1): 12-21. Drummond GB (1996). Comparison of sedation with midazolam and ketamine: effects on airway muscle activity. British Journal of Anaesthesia 76(5): 663-7. Drummond GB, Stedul K, Kingshott R, Rees K, Nimmo AF, Wraith P, Douglas NJ (2002). Automatic CPAP compared with conventional treatment for episodic hypoxemia and sleep disturbance after major abdominal surgery. Anesthesiology 96(4): 817-26. Dunai J, Kleiman J, Trinder J (1999). Ventilatory instability during sleep onset in individuals with high peripheral chemosensitivity. Journal of Applied Physiology 87(2): 661-72. Dunai J, Wilkinson M, Trinder J (1996). Interaction of chemical and state effects on ventilation during sleep onset. Journal of Applied Physiology 81(5): 2235-43. Eastwood PR, Szollosi I, Platt PR, Hillman DR (2002). Comparison of upper airway collapse during general anaesthesia and sleep. Lancet 359(9313): 1207-9. Edell-Gustafsson UM, Hetta JE, Aren CB (1999). Sleep and quality of life assessment in patients undergoing coronary artery bypass grafting. Journal of Advanced Nursing 29(5): 1213-20. Ellis BW & Dudley HA (1976). Some aspects of sleep research in surgical stress. Journal of Psychosomatic Research 20(4): 303-8. Entwistle MD, Roe PG, Sapsford DJ, Berrisford RG, Jones JG (1991). Patterns of oxygenation after thoracotomy. British Journal of Anaesthesia 67(6): 704-11. Etheredge JA, Perez SM, Taylor OR, Jander R (1999). Monarch butterflies (Danaus plexippus L.) use a magnetic compass for navigation. Proceedings of the National Academy of Sciences of the United States of America 96(24): 13845-6. Feinberg I (1974). Changes in sleep cycle patterns with age. Journal of Psychiatric Research 10: 283-306. Ferber R, Millman R, Coppola M, Fleetham J, Murray CF, Iber C, McCall V, Nino- Murcia G, Pressman M, Sanders M, et al. (1994). Portable recording in the assessment of obstructive sleep apnea. ASDA standards of practice. Sleep 17(4): 378-92. Findley LJ, Zwillich CW, Ancoli-Israel S, Kripke D, Tisi G, Moser KM (1985). Cheyne-Stokes breathing during sleep in patients with left ventricular heart failure. Southern Medical Journal 78(1): 11-5.

221 Fisher NI (1993). Statistical Analysis of Circular Data. Cambridge, Cambridge University Press. Fitzpatrick DC, Kuwada S, Batra R (2002). Transformations in processing interaural time differences between the superior olivary complex and inferior colliculus: beyond the Jeffress model. Hearing Research 168(1-2): 79-89. Flavell H, Marshall R, Thornton AT, Clements PL, Antic R, McEvoy RD (1992). Hypoxia episodes during sleep in high tetraplegia. Archives of Physical Medicine & Rehabilitation 73(7): 623-7. Fleetham JA (1992). Upper airway imaging in relation to obstructive sleep apnea. Clinics in Chest Medicine 13(3): 399-416. Fries R, Konig J, Schonecke O, Schafers HJ, Bohm M (2001). [Daily activities and circadian variation of ventricular tachyarrhythmias in patients with implanted defibrillator]. Deutsche Medizinische Wochenschrift 126(49): 1385-90. Galatius-Jensen S, Hansen J, Rasmussen V, Bildsoe J, Therboe M, Rosenberg J (1994). Nocturnal hypoxaemia after myocardial infarction: association with nocturnal myocardial ischaemia and arrhythmias. British Heart Journal 72(1): 23-30. Gallerani M, Manfredini R, Dal Monte D, Calo G, Brunaldi V, Simonato M (2001). Circadian differences in the individual sensitivity to opiate overdose. Critical Care Medicine 29(1): 96-101. Garpestad E, Ringler J, Parker JA, Remsburg S, Weiss JW (1995). Sleep stage influences the hemodynamic response to obstructive apneas. American Journal of Respiratory & Critical Care Medicine 152(1): 199-203. Gastaut H, Tassinari CA, Duron B (1966). Polygraphic study of the episodic diurnal and nocturnal (hypnic and respiratory) manifestations of the Pickwick syndrome. Brain Research 1(2): 167-86. Gavriely N & Jensen O (1993). Theory and measurements of snores. Journal of Applied Physiology 74(6): 2828-37. Gay PC, Westbrook PR, Daube JR, Litchy WJ, Windebank AJ, Iverson R (1991). Effects of alterations in pulmonary function and sleep variables on survival in patients with amyotrophic lateral sclerosis. Mayo Clinic Proceedings 66(7): 686- 94. Gelb AW & Knill RL (1978). Subanaesthetic halothane: its effect on regulation of ventilation and relevance to the recovery room. Canadian Anaesthetists Society Journal 25(6): 488-94. George CF, Millar TW, Kryger MH (1988). Sleep apnea and body position during sleep. Sleep 11(1): 90-9. Ghiandoni G, Secli R, Rocchi MB, Ugolini G (1998). Does lunar position influence the time of delivery? A statistical analysis. European Journal of Obstetrics, Gynecology, & Reproductive Biology 77(1): 47-50. Gill NP, Wright B, Reilly CS (1992). Relationship between hypoxaemic and cardiac ischaemic events in the perioperative period. British Journal of Anaesthesia 68(5): 471-3. Gillis AM, Connolly SJ, Dubuc M, Yee R, Lacomb P, Philippon F, Kerr CR, Kimber S, Gardner MJ, Tang AS, et al. (2001). Circadian variation of paroxysmal atrial fibrillation. PA3 Investigators. Atrial Pacing Peri-ablation for Prevention of Atrial Fibrillation Trial. American Journal of Cardiology 87(6): 794-8 A8. Glotzbach SP & Heller HC (1989). Thermoregulation. In: Principles and Practice of Sleep Medicine. M. Kryger, R. Roth and W. C. Dement, Eds. Philadelphia, WB Saunders: 300-9.

222 Gogenur I, Rosenberg-Adamsen S, Carstensen M, Lie C, Rasmussen V, Rosenberg J (2004). Relationship between nocturnal hypoxaemia, tachycardia and myocardial ischaemia after major abdominal surgery. British Journal of Anaesthesia 93(3): 333-8. Gogenur I, Rosenberg-Adamsen S, Lie C, Rasmussen V, Rosenberg J (2002). Lack of circadian variation in the activity of the autonomic nervous system after major abdominal operations. European Journal of Surgery 168(4): 242-6. Gold AR & Schwartz AR (1996). The pharyngeal critical pressure. The whys and hows of using nasal continuous positive airway pressure diagnostically. Chest 110(4): 1077-88. Goldman MD, Reeder MK, Muir AD, Loh L, Young JD, Gitlin DA, Casey KR, Smart D, Fry JM (1993). Repetitive nocturnal arterial oxygen desaturation and silent myocardial ischemia in patients presenting for vascular surgery. Journal of the American Geriatrics Society 41(7): 703-9. Grunstein RR (1995). Sleep-related breathing disorders. 5. Nasal continuous positive airway pressure treatment for obstructive sleep apnoea. Thorax 50(10): 1106-13. Grunstein RR, Ellis E, Hillman D, McEvoy RD, Robertson CF, Saunders NA (1991a). Treatment of sleep disordered breathing. Thoracic Society of Australia and New Zealand. Medical Journal of Australia 154(5): 355-9. Grunstein RR, Handelsman DJ, Lawrence SJ, Blackwell C, Caterson ID, Sullivan CE (1989). Neuroendocrine dysfunction in sleep apnea: reversal by continuous positive airways pressure therapy. Journal of Clinical Endocrinology & Metabolism 68(2): 352-8. Grunstein RR, Ho KY, Sullivan CE (1991b). Sleep apnea in acromegaly. Annals of Internal Medicine 115(7): 527-32. Grunstein RR & Sullivan CE (1988). Sleep apnea and hypothyroidism: mechanisms and management. American Journal of Medicine 85(6): 775-9. Grunstein RR & Wilcox I (1994). Sleep-disordered breathing and obesity. Baillieres Clinical Endocrinology & Metabolism 8(3): 601-28. Guilleminault C, Cummiskey J, Motta J, Lynne-Davies P (1978). Respiratory and hemodynamic study during wakefulness and sleep in myotonic dystrophy. Sleep 1(1): 19-31. Guilleminault C & Dement WC (1991). General Physiology of Sleep. In: The Lung: Scientific Foundations. R. G. Crystal, J. B. West and et al., Eds. New York, Raven Press, Ltd.: 1609-1614. Guilleminault C, Lehrman AT, Forno L, Dement WC (1977). Sleep apnoea syndrome: states of sleep and autonomic dysfunction. Journal of Neurology, Neurosurgery & Psychiatry 40(7): 718-25. Guilleminault C, Stoohs R, Clerk A, Cetel M, Maistros P (1993). A cause of excessive daytime sleepiness. The upper airway resistance syndrome. Chest 104(3): 781-7. Hanning CD (1989). Obstructive sleep apnoea. British Journal of Anaesthesia 63(4): 477-88. Harris JC & Allen RP (1996). Is excessive daytime sleepiness characteristic of Prader- Willi syndrome? The effects of weight change. Archives of Pediatrics & Adolescent Medicine 150(12): 1288-93. Haskell EH, Palca JW, Walker JM, Berger RJ, Heller HC (1981). The effects of high and low ambient temperatures on human sleep stages. Electroencephalography & Clinical Neurophysiology 51(5): 494-501.

223 Hawley DJ, Wolfe F, Lue FA, Moldofsky H (2001). Seasonal symptom severity in patients with rheumatic diseases: a study of 1,424 patients. Journal of Rheumatology 28(8): 1900-9. Helfaer MA, McColley SA, Pyzik PL, Tunkel DE, Nichols DG, Baroody FM, April MM, Maxwell LG, Loughlin GM (1996). Polysomnography after adenotonsillectomy in mild pediatric obstructive sleep apnea. Critical Care Medicine 24(8): 1323-7. Helfaer MA & Wilson MD (1994). Obstructive sleep apnea, control of ventilation, and anesthesia in children. Pediatric Clinics of North America 41(1): 131-51. Henke KG, Dempsey JA, Badr MS, Kowitz JM, Skatrud JB (1991). Effect of sleep- induced increases in upper airway resistance on respiratory muscle activity. Journal of Applied Physiology 70(1): 158-68. Henriques JP, Haasdijk AP, Zijlstra F, Zwolle Myocardial Infarction Study G (2003). Outcome of primary angioplasty for acute myocardial infarction during routine duty hours versus during off-hours. Journal of the American College of Cardiology 41(12): 2138-42. Hill NS (1993). Noninvasive ventilation. Does it work, for whom, and how? American Review of Respiratory Disease 147(4): 1050-5. Hill NS, Redline S, Carskadon MA, Curran FJ, Millman RP (1992). Sleep-disordered breathing in patients with Duchenne muscular dystrophy using negative pressure ventilators. Chest 102(6): 1656-62. Hillman DR, Platt PR, Eastwood PR (2003). The upper airway during anaesthesia. British Journal of Anaesthesia 91(1): 31-9. Hiremath AS, Hillman DR, James AL, Noffsinger WJ, Platt PR, Singer SL (1998). Relationship between difficult tracheal intubation and obstructive sleep apnoea. British Journal of Anaesthesia 80(5): 606-11. Hirsch HV, Leventhal AG, McCall MA, Tieman DG (1983). Effects of exposure to lines of one or two orientations on different cell types in striate cortex of cat. Journal of Physiology 337: 241-55. Hoffstein V, Herridge M, Mateika S, Redline S, Strohl KP (1994). Hematocrit levels in sleep apnea. Chest 106(3): 787-91. Hoffstein V & Szalai JP (1993). Predictive value of clinical features in diagnosing obstructive sleep apnea. Sleep 16(2): 118-22. Horner RL (1996). Motor control of the pharyngeal musculature and implications for the pathogenesis of obstructive sleep apnea. Sleep 19(10): 827-53. Horner RL (2001). The neuropharmacology of upper airway motor control in the awake and asleep states: implications for obstructive sleep apnoea. Respiratory Research 2(5): 286-94. Hsu AA & Staats BA (1998). "Postpolio" sequelae and sleep-related disordered breathing. Mayo Clinic Proceedings 73(3): 216-24. Hu FB, Willett WC, Colditz GA, Ascherio A, Speizer FE, Rosner B, Hennekens CH, Stampfer MJ (1999). Prospective study of snoring and risk of hypertension in women. American Journal of Epidemiology 150(8): 806-16. Hudgel DW & Devadatta P (1984a). Decrease in functional residual capacity during sleep in normal humans. Journal of Applied Physiology: Respiratory, Environmental & Exercise Physiology 57(5): 1319-22. Hudgel DW, Martin RJ, Johnson B, Hill P (1984b). Mechanics of the respiratory system and breathing pattern during sleep in normal humans. Journal of Applied Physiology: Respiratory, Environmental & Exercise Physiology 56(1): 133-7.

224 Huerta S, DeShields S, Shpiner R, Li Z, Liu C, Sawicki M, Arteaga J, Livingston EH (2002). Safety and efficacy of postoperative continuous positive airway pressure to prevent pulmonary complications after Roux-en-Y gastric bypass. Journal of Gastrointestinal Surgery 6(3): 354-8. Hung J, Whitford EG, Parsons RW, Hillman DR (1990). Association of sleep apnoea with myocardial infarction in men. Lancet 336(8710): 261-4. Iber C, Berssenbrugge A, Skatrud JB, Dempsey JA (1982). Ventilatory adaptations to resistive loading during wakefulness and non-REM sleep. Journal of Applied Physiology: Respiratory, Environmental & Exercise Physiology 52(3): 607-14. Isono S, Remmers JE, Tanaka A, Sho Y, Sato J, Nishino T (1997). Anatomy of pharynx in patients with obstructive sleep apnea and in normal subjects. Journal of Applied Physiology 82(4): 1319-26. Isono S, Tanaka A, Nishino T (2002). Lateral position decreases collapsibility of the passive pharynx in patients with obstructive sleep apnea. Anesthesiology 97(4): 780-5. Javaheri S, Parker TJ, Liming JD, Corbett WS, Nishiyama H, Wexler L, Roselle GA (1998). Sleep apnea in 81 ambulatory male patients with stable heart failure. Types and their prevalences, consequences, and presentations. Circulation 97(21): 2154-9. Javaheri S, Parker TJ, Wexler L, Michaels SE, Stanberry E, Nishyama H, Roselle GA (1995). Occult sleep-disordered breathing in stable congestive heart failure. Annals of Internal Medicine 122(7): 487-92. Jennum P & Borgesen SE (1989). Intracranial pressure and obstructive sleep apnea. Chest 95(2): 279-83. Johns MW, Large AA, Masterton JP, Dudley HA (1974). Sleep and delirium after open heart surgery. British Journal of Surgery 61(5): 377-81. Johnson JT & Sanders MH (1986). Breathing during sleep immediately after uvulopalatopharyngoplasty. Laryngoscope 96(11): 1236-8. Jones JG, Sapsford DJ, Wheatley RG (1990). Postoperative hypoxaemia: mechanisms and time course. Anaesthesia 45(7): 566-73. Kain ZN & Caldwell-Andrews AA (2003). Sleeping Characteristics of Adults Undergoing Outpatient Elective Surgery: A Cohort Study. Journal of Clinical Anesthesia 15: 505-509. Kain ZN, Mayes LC, Caldwell-Andrews AA, Alexander GM, Krivutza D, Teague BA, Wang SM (2002). Sleeping characteristics of children undergoing outpatient elective surgery. Anesthesiology 97(5): 1093-101. Kales A, Caldwell AB, Cadieux RJ, Vela-Bueno A, Ruch LG, Mayes SD (1985). Severe obstructive sleep apnea--II: Associated psychopathology and psychosocial consequences. Journal of Chronic Diseases 38(5): 427-34. Kales A, Jacobson A, Kales JD, Kun T, Weissbuch R (1967). All-night EEG sleep measurements in young adults. Psychonomic Science 7: 67. Karacan I, Thornby JI, Anch AM, Williams RL, Perkins HM (1978). Effects of high ambient temperature on sleep in young men. Aviation Space & Environmental Medicine 49(7): 855-60. Kavey NB & Ahshuler KZ (1979a). Sleep in herniorrhaphy patients. American Journal of Surgery 138(5): 683-7. Kavey NB & Altshuler KZ (1979b). Sleep in herniorrhaphy patients. American Journal of Surgery 138(5): 683-7. Kavey NB & Altshuler KZ (1983). Flurazepam and the sleep of herniorrhaphy patients. Journal of Clinical Pharmacology 23(5-6): 199-208.

225 Kay DC, Pickworth WB, Neidert GL, Falcone D, Fishman PM, Othmer E (1979). Opioid effects on computer-derived sleep and EEG parameters in nondependent human addicts. Sleep 2(2): 175-91. Kehlet H (1997). Multimodal approach to control postoperative pathophysiology and rehabilitation. British Journal of Anaesthesia 78(5): 606-17. Kerr P, Shoenut JP, Millar T, Buckle P, Kryger MH (1992). Nasal CPAP reduces gastroesophageal reflux in obstructive sleep apnea syndrome. Chest 101(6): 1539-44. Khan MS & Ahmad SI (2003). Circadian variation--increased morning incidence of acute myocardial infarction in patients with coronary artery disease. JPMA - Journal of the Pakistan Medical Association 53(2): 84-7. Khoo MC, Shin JJ, Asyali MH, Kim TS, Berry RB (1998). Ventilatory dynamics of transient arousal in patients with obstructive sleep apnea. Respiration Physiology 112(3): 291-303. Kimmel PL, Miller G, Mendelson WB (1989). Sleep apnea syndrome in chronic renal disease. American Journal of Medicine 86(3): 308-14. Kinjo K, Sato H, Shiotani I, Kurotobi T, Ohnishi Y, Hishida E, Nakatani D, Ito H, Koretsune Y, Hirayama A, et al. (2001). Circadian variation of the onset of acute myocardial infarction in the Osaka area, 1998-1999: characterization of morning and nighttime peaks. Japanese Circulation Journal 65(7): 617-20. Knill RL, Bright S, Manninen P (1978a). Hypoxic ventilatory responses during thiopentone sedation and anaesthesia in man. Canadian Anaesthetists Society Journal 25(5): 366-72. Knill RL & Clement JL (1982). Variable effects of anaesthetics on the ventilatory response to hypoxaemia in man. Canadian Anaesthetists Society Journal 29(2): 93-9. Knill RL & Clement JL (1984). Site of selective action of halothane on the peripheral chemoreflex pathway in humans. Anesthesiology 61(2): 121-6. Knill RL, Clement JL, Thompson WR (1981). Epidural morphine causes delayed and prolonged ventilatory depression. Canadian Anaesthetists Society Journal 28(6): 537-43. Knill RL & Gelb AW (1978b). Ventilatory responses to hypoxia and hypercapnia during halothane sedation and anesthesia in man. Anesthesiology 49(4): 244-51. Knill RL, Kieraszewicz HT, Dodgson BG, Clement JL (1983). Chemical regulation of ventilation during isoflurane sedation and anaesthesia in humans. Canadian Anaesthetists Society Journal 30(6): 607-14. Knill RL, Manninen PH, Clement JL (1979). Ventilation and chemoreflexes during enflurane sedation and anaesthesia in man. Canadian Anaesthetists Society Journal 26(5): 353-60. Knill RL, Moote CA, Rose EA, Skinner MI (1987). Marked hypoxemia after gastroplasty due to disorders of breathing in REM sleep. Anesthesiology 67(3A): A552. Knill RL, Moote CA, Skinner MI, Rose EA (1990). Anesthesia with abdominal surgery leads to intense REM sleep during the first postoperative week. Anesthesiology 73(1): 52-61. Knudsen J (1970). Duration of hypoxaemia after uncomplicated upper abdominal and thoraco-abdominal operations. Anaesthesia 25(3): 372-7. Koenig SM (2001). Pulmonary complications of obesity. American Journal of the Medical Sciences 321(4): 249-79.

226 Kontrobarsky Y & Love J (1997). Gluteal compartment syndrome following epidural analgesic infusion with motor blockage. Anaesthesia & Intensive Care 25(6): 696-8. Koskenvuo M, Kaprio J, Telakivi T, Partinen M, Heikkila K, Sarna S (1987). Snoring as a risk factor for ischaemic heart disease and stroke in men. British Medical Journal Clinical Research Ed. 294(6563): 16-9. Koskenvuo M, Partinen M, Kaprio J, Vuorinen H, Telakivi T, Kajaste S, Salmi T, Heikkila K (1994). Snoring and cardiovascular risk factors. Annals of Medicine 26(5): 371-6. Kravath RE, Pollak CP, Borowiecki B, Weitzman ED (1980). Obstructive sleep apnea and death associated with surgical correction of velopharyngeal incompetence. Journal of Pediatrics 96(4): 645-8. Kryger MH (1992). Management of obstructive sleep apnea. Clinics in Chest Medicine 13(3): 481-92. Kudrow L, McGinty DJ, Phillips ER, Stevenson M (1984). Sleep apnea in cluster headache. Cephalalgia 4(1): 33-8. Kuna ST & Sant'Ambrogio G (1991). Pathophysiology of upper airway closure during sleep. Jama 266(10): 1384-9. Laks L, Lehrhaft B, Grunstein RR, Sullivan CE (1995). Pulmonary hypertension in obstructive sleep apnoea. European Respiratory Journal 8(4): 537-41. Langeron O, Masso E, Huraux C, Guggiari M, Bianchi A, Coriat P, Riou B (2000). Prediction of difficult mask ventilation. Anesthesiology 92(5): 1229-36. Langevin B, Fouque D, Leger P, Robert D (1993). Sleep apnea syndrome and end-stage renal disease. Cure after renal transplantation. Chest 103(5): 1330-5. Lavie P, Herer P, Peled R, Berger I, Yoffe N, Zomer J, Rubin AH (1995). Mortality in sleep apnea patients: a multivariate analysis of risk factors. Sleep 18(3): 149-57. Lebedev MA & Nelson RJ (1996). High-frequency vibratory sensitive neurons in monkey primary somatosensory cortex: entrained and nonentrained responses to vibration during the performance of vibratory-cued hand movements. Experimental Brain Research 111(3): 313-25. Lee SK, Lee DS, Andrews WL, Baboolal R, Pendray M, Stewart S, Canadian Neonatal N (2003). Higher mortality rates among inborn infants admitted to neonatal intensive care units at night. Journal of Pediatrics 143(5): 592-7. Lehmkuhl P, Prass D, Pichlmayr I (1987). General anesthesia and postnarcotic sleep disorders. Neuropsychobiology 18(1): 37-42. Levine JD, Funes P, Dowse HB, Hall JC (2002). Signal analysis of behavioral and molecular cycles. BMC Neuroscience 3(1): 1. Loadsman JA & Hillman DR (2001). Anaesthesia and sleep apnoea. British Journal of Anaesthesia 86(2): 254-66. Loomis AL, Harvey EN, Hobart GA (1937). Cerebral states during sleep as studied by human brain potentials. Journal of Experimental Psychology 21: 127. Lugaresi E, Mondini S, Zucconi M, Montagna P, Cirignotta F (1983). Staging of heavy snorers' disease. A proposal. Bulletin Europeen de Physiopathologie Respiratoire 19(6): 590-4. Malhotra A & White DP (2002). Obstructive sleep apnoea. Lancet 360(9328): 237-45. Manfredini R, La Cecilia O, Boari B, Steliu J, Michelinidagger V, Carlidagger P, Zanotti C, Bigoni M, Gallerani M (2002). Circadian pattern of emergency calls: implications for ED organization. American Journal of Emergency Medicine 20(4): 282-6.

227 Mann KA, Gupta S, Race A, Miller MA, Cleary RJ (2003). Application of circular statistics in the study of crack distribution around cemented femoral components. Journal of Biomechanics 36(8): 1231-4. Marcus CL, Keens TG, Bautista DB, von Pechmann WS, Ward SL (1991). Obstructive sleep apnea in children with Down syndrome. Pediatrics 88(1): 132-9. Marler JR, Price TR, Clark GL, Muller JE, Robertson T, Mohr JP, Hier DB, Wolf PA, Caplan LR, Foulkes MA (1989). Morning increase in onset of ischemic stroke. Stroke 20(4): 473-6. Mathew OP, Abu-Osba YK, Thach BT (1982). Influence of upper airway pressure changes on genioglossus muscle respiratory activity. Journal of Applied Physiology: Respiratory, Environmental & Exercise Physiology 52(2): 438-44. Mathur R & Douglas NJ (1995). Frequency of EEG arousals from nocturnal sleep in normal subjects. Sleep 18(5): 330-3. McArdle P (1999). Intravenous analgesia. Critical Care Clinics 15(1): 89-104. McColley SA, April MM, Carroll JL, Naclerio RM, Loughlin GM (1992). Respiratory compromise after adenotonsillectomy in children with obstructive sleep apnea. Archives of Otolaryngology - Head & Neck Surgery 118(9): 940-3. McGinty DJ, Drucker-Colin R, Morrison A, Parmeggiani PL (1985). Brain Mechanisms of Sleep. New York, Raven Press. McNamara SG, Grunstein RR, Sullivan CE (1993). Obstructive sleep apnoea. Thorax 48(7): 754-64. Mehta RH, Manfredini R, Hassan F, Sechtem U, Bossone E, Oh JK, Cooper JV, Smith DE, Portaluppi F, Penn M, et al. (2002). Chronobiological patterns of acute aortic dissection. Circulation 106(9): 1110-5. Mehta Y, Manikappa S, Juneja R, Trehan N (2000). Obstructive sleep apnea syndrome: anesthetic implications in the cardiac surgical patient. Journal of Cardiothoracic & Vascular Anesthesia 14(4): 449-53. Meoli AL, Casey KR, Clark RW, Coleman JA, Jr., Fayle RW, Troell RJ, Iber C (2001). Hypopnea in sleep-disordered breathing in adults: Report of the Clinical Practice Review Committee, American Academy of Sleep Medicine. Sleep 24(4): 469- 70. Meoli AL, Rosen CL, Kristo D, Kohrman M, Gooneratne N, Aguillard RN, Fayle R, Troell R, Kramer R, Casey KR, et al. (2003). Upper airway management of the adult patient with obstructive sleep apnea in the perioperative period - avoiding complications: Report of the Clinical Practice Review Committee, American Academy of Sleep Medicine. Sleep 26(8): 1060-5. Mercado-Hernandez R, Fernandez-Salas I, Villarreal-Martinez H (2003). Spatial distribution of the larval indices of Aedes aegypti in Guadalupe, Nuevo Leon, Mexico, with circular distribution analysis. Journal of the American Mosquito Control Association 19(1): 15-8. Miljeteig H, Hoffstein V, Cole P (1992). The effect of unilateral and bilateral nasal obstruction on snoring and sleep apnea. Laryngoscope 102(10): 1150-2. Millar-Craig MW, Bishop CN, Raftery EB (1978). Circadian variation of blood- pressure. Lancet 1(8068): 795-7. Mohsenin V & Valor R (1995). Sleep apnea in patients with hemispheric stroke. Archives of Physical Medicine & Rehabilitation 76(1): 71-6. Moote CA & Knill RL (1988). Isoflurane anesthesia causes a transient alteration in nocturnal sleep. Anesthesiology 69(3): 327-31.

228 Moote CA, Knill RL, Clement J (1986). Ventilatory compensation for continuous inspiratory resistive and elastic loads during halothane anesthesia in humans. Anesthesiology 64(5): 582-9. Muller JE, Ludmer PL, Willich SN, Tofler GH, Aylmer G, Klangos I, Stone PH (1987). Circadian variation in the frequency of sudden cardiac death. Circulation 75(1): 131-8. Muller JE, Stone PH, Turi ZG, Rutherford JD, Czeisler CA, Parker C, Poole WK, Passamani E, Roberts R, Robertson T, et al. (1985). Circadian variation in the frequency of onset of acute myocardial infarction. New England Journal of Medicine 313(21): 1315-22. Munoz X, Marti S, Sumalla J, Bosch J, Sampol G (1998). Acute delirium as a manifestation of obstructive sleep apnea syndrome. American Journal of Respiratory & Critical Care Medicine 158(4): 1306-7. Muzet A, Ehrhart J, Candas V, Libert JP, Vogt JJ (1983). REM sleep and ambient temperature in man. International Journal of Neuroscience 18(1-2): 117-26. Nagyova B, Dorrington KL, Robbins PA (1993). Effects of midazolam and flumazenil on ventilation during sustained hypoxia in humans. Respiration Physiology 94(1): 51-9. Nagyova B, Dorrington KL, Robbins PA (1994). Effect of low-dose enflurane on the ventilatory response to hypoxia in humans. British Journal of Anaesthesia 72(5): 509-14. Nieuwenhuijs D, Coleman EL, Douglas NJ, Drummond GB, Dahan A (2002). Bispectral index values and spectral edge frequency at different stages of physiologic sleep. Anesthesia & Analgesia 94(1): 125-9. Nishino T, Shirahata M, Yonezawa T, Honda Y (1984). Comparison of changes in the hypoglossal and the phrenic nerve activity in response to increasing depth of anesthesia in cats. Anesthesiology 60(1): 19-24. Nunn JF (1969). Applied Respiratory Physiology. London, Butterworths. Nunn JF & Payne JP (1962). Hypoxaemia after general anaesthesia. Lancet 2: 631-2. O'Sullivan RA, Hillman DR, Mateljan R, Pantin C, Finucane KE (1995). Mandibular advancement splint: an appliance to treat snoring and obstructive sleep apnea. American Journal of Respiratory & Critical Care Medicine 151(1): 194-8. Ochiai R, Guthrie RD, Motoyama EK (1989). Effects of varying concentrations of halothane on the activity of the genioglossus, intercostals, and diaphragm in cats: an electromyographic study. Anesthesiology 70(5): 812-6. Oksenberg A, Silverberg DS, Arons E, Radwan H (1997). Positional vs nonpositional obstructive sleep apnea patients: anthropomorphic, nocturnal polysomnographic, and multiple sleep latency test data. Chest 112(3): 629-39. Orr WC & Stahl ML (1977). Sleep disturbances after open heart surgery. American Journal of Cardiology 39(2): 196-201. Ostermeier AM, Roizen MF, Hautkappe M, Klock PA, Klafta JM (1997). Three sudden postoperative respiratory arrests associated with epidural opioids in patients with sleep apnea. Anesthesia & Analgesia 85(2): 452-60. Pantin CC, Hillman DR, Tennant M (1999). Dental side effects of an oral device to treat snoring and obstructive sleep apnea. Sleep 22(2): 237-40. Papper EM (1971). Some causes of postanesthetic hypoxia. Anales de la Real Academia Nacional de Medicina 88(4): 493-506. Parikh SN, Stuchin SA, Maca C, Fallar E, Steiger D (2002). Sleep apnea syndrome in patients undergoing total joint arthroplasty. Journal of Arthroplasty 17(5): 635- 42.

229 Pevernagie DA & Shepard JW, Jr. (1992). Relations between sleep stage, posture and effective nasal CPAP levels in OSA. Sleep 15(2): 162-7. Piper AJ & Sullivan CE (1994). Effects of short-term NIPPV in the treatment of patients with severe obstructive sleep apnea and hypercapnia. Chest 105(2): 434- 40. Pivik RT & Harman K (1995). A reconceptualization of EEG alpha activity as an index of arousal during sleep: all alpha activity is not equal. Journal of Sleep Research 4: 131-7. Pochron ST (2000). Sun avoidance in the yellow baboons (Papio cynocephalus cynocephalus) of Ruaha National Park, Tanzania. Variations with season, behavior and weather. International Journal of Biometeorology 44(3): 141-7. Popper RA, Leidinger MJ, Williams AJ (1986). Endoscopic observations of the pharyngeal airway during treatment of obstructive sleep apnea with nasal continuous positive airway pressure - a pneumatic splint. Western Journal of Medicine 144(1): 83-5. Rahman MQ, Kingshott RN, Wraith P, Adams WH, Drummond GB (2001). Association of airway obstruction, sleep, and phasic abdominal muscle activity after upper abdominal surgery. British Journal of Anaesthesia 87(2): 198-203. Rains JC & Penzien DB (2003). Sleep and chronic pain: challenges to the alpha-EEG sleep pattern as a pain specific sleep anomaly. Journal of Psychosomatic Research 54(1): 77-83. Rampil IJ (1998). A primer for EEG signal processing in anesthesia. Anesthesiology 89(4): 980-1002. Rana JS, Mukamal KJ, Morgan JP, Muller JE, Mittleman MA (2003). Circadian variation in the onset of myocardial infarction: effect of duration of diabetes. Diabetes 52(6): 1464-8. Rechtschaffen A & Kales A (1968). A Manual of Standardised Terminology, Techniques, and Scoring System for Sleep Stages of Human Subjects. Bethesda, Maryland, US Department of Health, Education and Welfare. Redline S, Tosteson T, Tishler PV, Carskadon MA, Millman RP, Millman RP (1992). Studies in the genetics of obstructive sleep apnea. Familial aggregation of symptoms associated with sleep-related breathing disturbances. American Review of Respiratory Disease 145(2 Pt 1): 440-4. Reeder MK, Goldman MD, Loh L, Muir AD, Casey KR, Gitlin DA (1991a). Postoperative obstructive sleep apnoea. Haemodynamic effects of treatment with nasal CPAP. Anaesthesia 46(10): 849-53. Reeder MK, Goldman MD, Loh L, Muir AD, Casey KR, Lehane JR (1992a). Late postoperative nocturnal dips in oxygen saturation in patients undergoing major abdominal vascular surgery. Predictive value of pre-operative overnight pulse oximetry. Anaesthesia 47(2): 110-5. Reeder MK, Goldman MD, Loh L, Muir AD, Foex P, Casey KR, McKenzie PJ (1992b). Postoperative hypoxaemia after major abdominal vascular surgery. British Journal of Anaesthesia 68(1): 23-6. Reeder MK, Muir AD, Foex P, Goldman MD, Loh L, Smart D (1991b). Postoperative myocardial ischaemia: temporal association with nocturnal hypoxaemia. British Journal of Anaesthesia 67(5): 626-31. Rennotte MT, Baele P, Aubert G, Rodenstein DO (1995). Nasal continuous positive airway pressure in the perioperative management of patients with obstructive sleep apnea submitted to surgery. Chest 107(2): 367-74.

230 Rizzuto DS, Madsen JR, Bromfield EB, Schulze-Bonhage A, Seelig D, Aschenbrenner- Scheibe R, Kahana MJ (2003). Reset of human neocortical oscillations during a working memory task. Proceedings of the National Academy of Sciences of the United States of America 100(13): 7931-6. Roa NL & Moss KS (1984). Treacher-Collins syndrome with sleep apnea: anesthetic considerations. Anesthesiology 60(1): 71-3. Robinson RW & Zwillich CW (1989). The Effects of Drugs on Breathing During Sleep. In: Principles and Practice of Sleep Medicine. M. Kryger, R. Roth and W. C. Dement, Eds. Philadelphia, WB Saunders: 501-13. Rocchi MB & Perlini C (2002). Is the time of suicide a random choice? A new statistical perspective. Crisis: Journal of Crisis Intervention & Suicide 23(4): 161-6. Rocchi MB, Perlini C, Ciatti R, Burroni M (2001). Is the birthdate a risk factor for congenital hypothyroidism? A statistical answer based on personal experience. Minerva Pediatrica 53(6): 531-6. Rocco MB, Barry J, Campbell S, Nabel E, Cook EF, Goldman L, Selwyn AP (1987). Circadian variation of transient myocardial ischemia in patients with coronary artery disease. Circulation 75(2): 395-400. Rosen CL (1996). Obstructive sleep apnea syndrome (OSAS) in children: diagnostic challenges. Sleep 19(10 Suppl): S274-7. Rosen CL, D'Andrea L, Haddad GG (1992). Adult criteria for obstructive sleep apnea do not identify children with serious obstruction. American Review of Respiratory Disease 146(5 Pt 1): 1231-4. Rosenberg J, Dirkes WE, Kehlet H (1989). Episodic arterial oxygen desaturation and heart rate variations following major abdominal surgery. British Journal of Anaesthesia 63(6): 651-4. Rosenberg J & Kehlet H (1993). Postoperative mental confusion--association with postoperative hypoxemia. Surgery 114(1): 76-81. Rosenberg J, Pedersen MH, Ramsing T, Kehlet H (1992). Circadian variation in unexpected postoperative death. British Journal of Surgery 79(12): 1300-2. Rosenberg J, Rasmussen GI, Wojdemann KR, Kirkeby LT, Jorgensen LN, Kehlet H (1999). Ventilatory pattern and associated episodic hypoxaemia in the late postoperative period in the general surgical ward. Anaesthesia 54(4): 323-8. Rosenberg J, Rasmussen V, von Jessen F, Ullstad T, Kehlet H (1990). Late postoperative episodic and constant hypoxaemia and associated ECG abnormalities. British Journal of Anaesthesia 65(5): 684-91. Rosenberg J, Rosenberg-Adamsen S, Kehlet H (1995). Post-operative sleep disturbance: causes, factors and effects on outcome. European Journal of Anaesthesiology - Supplement 10: 28-30. Rosenberg J, Wildschiodtz G, Pedersen MH, von Jessen F, Kehlet H (1994). Late postoperative nocturnal episodic hypoxaemia and associated sleep pattern. British Journal of Anaesthesia 72(2): 145-50. Rosenberg-Adamsen S, Kehlet H, Dodds C, Rosenberg J (1996a). Postoperative sleep disturbances: mechanisms and clinical implications. British Journal of Anaesthesia 76(4): 552-9. Rosenberg-Adamsen S, Skarbye M, Wildschiodtz G, Kehlet H, Rosenberg J (1996b). Sleep after laparoscopic cholecystectomy. British Journal of Anaesthesia 77(5): 572-5.

231 Rosenberg-Adamsen S, Stausholm K, Edvardsen L, Zwarts M, Kehlet H, Rosenberg J (1997). Body position and late postoperative nocturnal hypoxaemia. Anaesthesia 52(6): 589-92. Ross AJ, Beagley HA, Sayers BM (1980). Signal statistics in objective auditory evoked potential (AEP) detection by the phase spectral method. Journal of Biomedical Engineering 2(4): 310-4. Sadeh A & Acebo C (2002). The role of actigraphy in sleep medicine. Sleep Medicine Reviews 6(2): 113-24. Sajkov D, Cowie RJ, Thornton AT, Espinoza HA, McEvoy RD (1994). Pulmonary hypertension and hypoxemia in obstructive sleep apnea syndrome. American Journal of Respiratory & Critical Care Medicine 149(2 Pt 1): 416-22. Santamaria J & Chiappa KH (1987). The EEG of drowsiness in normal adults. Journal of Clinical Neurophysiology 4(4): 327-82. Sarton E, Dahan A, Teppema L, van den Elsen M, Olofsen E, Berkenbosch A, van Kleef J (1996). Acute pain and central nervous system arousal do not restore impaired hypoxic ventilatory response during sevoflurane sedation. Anesthesiology 85(2): 295-303. Sauerland EK & Harper RM (1976). The human tongue during sleep: electromyographic activity of the genioglossus muscle. Experimental Neurology 51(1): 160-70. Sauret Valet J (1999). [Pickwick or Pliny?]. Archivos de Bronconeumologia 35(2): 91- 3. Scrima L, Broudy M, Nay KN, Cohn MA (1982). Increased severity of obstructive sleep apnea after bedtime alcohol ingestion: diagnostic potential and proposed mechanism of action. Sleep 5(4): 318-28. Seigel JM (1989). Brain stem mechanisms generating REM sleep. In: Principles and Practice of Sleep Medicine. M. Kryger, R. Roth and W. C. Dement, Eds. Philadelphia, WB Saunders: 104-120. Semenza GL & Pyeritz RE (1988). Respiratory complications of mucopolysaccharide storage disorders. Medicine 67(4): 209-19. Series F (2002). Upper airway muscles awake and asleep. Sleep Medicine Reviews 6(3): 229-42. Sher AE, Schechtman KB, Piccirillo JF (1996). The efficacy of surgical modifications of the upper airway in adults with obstructive sleep apnea syndrome. Sleep 19(2): 156-77. Short DJ, Stradling JR, Williams SJ (1992). Prevalence of sleep apnoea in patients over 40 years of age with spinal cord lesions. Journal of Neurology, Neurosurgery & Psychiatry 55(11): 1032-6. Simantirakis EN, Chrysostomakis SI, Marketou ME, Kochiadakis GE, Vardakis KE, Mavrakis HE, Vardas P (2001). Atrial and ventricular refractoriness in paced patients; circadian variation and its relationship to autonomic nervous system activity. European Heart Journal 22(23): 2192-200. Sjogren D, Sollevi A, Ebberyd A, Lindahl SG (1994). Poikilocapnic hypoxic ventilatory response in humans during 0.85 MAC isoflurane anesthesia. Acta Anaesthesiologica Scandinavica 38(2): 149-55. Sjogren D, Sollevi A, Ebberyd A, Lindahl SG (1995). Isoflurane anaesthesia (0.6 MAC) and hypoxic ventilatory responses in humans. Acta Anaesthesiologica Scandinavica 39(1): 17-22.

232 Skatrud J, Iber C, McHugh W, Rasmussen H, Nichols D (1980). Determinants of hypoventilation during wakefulness and sleep in diaphragmatic paralysis. American Review of Respiratory Disease 121(3): 587-93. Skatrud JB & Dempsey JA (1985). Airway resistance and respiratory muscle function in snorers during NREM sleep. Journal of Applied Physiology 59(2): 328-35. Skatrud JB, Dempsey JA, Badr S, Begle RL (1988). Effect of airway impedance on CO2 retention and respiratory muscle activity during NREM sleep. Journal of Applied Physiology 65(4): 1676-85. Sleigh JW, Andrzejowski J, Steyn-Ross A, Steyn-Ross M (1999). The bispectral index: a measure of depth of sleep? Anesthesia & Analgesia 88(3): 659-61. Solin P, Roebuck T, Swieca J, Walters EH, Naughton MT (1998). Effects of cardiac dysfunction on non-hypercapnic central sleep apnea. Chest 113(1): 104-10. Sollevi A & Lindahl SG (1995). Hypoxic and hypercapnic ventilatory responses during isoflurane sedation and anaesthesia in women. Acta Anaesthesiologica Scandinavica 39(7): 931-8. Spence AA & Alexander JI (1972). Mechanisms of postoperative hypoxaemia. Proceedings of the Royal Society of Medicine 65(1): 12-4. Stephenson EA, Collins KK, Dubin AM, Epstein MR, Hamilton RM, Kertesz NJ, Alexander ME, Cecchin F, Triedman JK, Walsh EP, et al. (2002). Circadian and seasonal variation of malignant arrhythmias in a pediatric and congenital heart disease population. Journal of Cardiovascular Electrophysiology 13(10): 1009- 14. Stergiou GS, Vemmos KN, Pliarchopoulou KM, Synetos AG, Roussias LG, Mountokalakis TD (2002). Parallel morning and evening surge in stroke onset, blood pressure, and physical activity. Stroke 33(6): 1480-6. Stern E, Parmelee AH, Akiyama Y, Schultz MA, Wenner WH (1969). Sleep cycle characteristics in infants. Pediatrics 43(1): 65-70. Stokes DC, Phillips JA, Leonard CO, Dorst JP, Kopits SE, Trojak JE, Brown DL (1983). Respiratory complications of achondroplasia. Journal of Pediatrics 102(4): 534-41. Stoohs R & Guilleminault C (1992). MESAM 4: an ambulatory device for the detection of patients at risk for obstructive sleep apnea syndrome (OSAS). Chest 101(5): 1221-7. Stradling JR & Crosby JH (1991). Predictors and prevalence of obstructive sleep apnoea and snoring in 1001 middle aged men. Thorax 46(2): 85-90. Sullivan CE, Issa FG, Berthon-Jones M, Eves L (1981). Reversal of obstructive sleep apnoea by continuous positive airway pressure applied through the nares. Lancet 1(8225): 862-5. Taasan V, Wynne JW, Cassisi N, Block AJ (1981a). The effect of nasal packing on sleep-disordered breathing and nocturnal oxygen desaturation. Laryngoscope 91(7): 1163-72. Taasan VC, Block AJ, Boysen PG, Wynne JW (1981b). Alcohol increases sleep apnea and oxygen desaturation in asymptomatic men. American Journal of Medicine 71(2): 240-5. Tabachnik E, Muller NL, Bryan AC, Levison H (1981). Changes in ventilation and chest wall mechanics during sleep in normal adolescents. Journal of Applied Physiology: Respiratory, Environmental & Exercise Physiology 51(3): 557-64. Temp JA, Henson LC, Ward DS (1992). Does a subanesthetic concentration of isoflurane blunt the ventilatory response to hypoxia? Anesthesiology 77(6): 1116-24.

233 Temp JA, Henson LC, Ward DS (1994). Effect of a subanesthetic minimum alveolar concentration of isoflurane on two tests of the hypoxic ventilatory response. Anesthesiology 80(4): 739-50. Terris DJ, Clerk AA, Norbash AM, Troell RJ (1996). Characterization of postoperative edema following laser-assisted uvulopalatoplasty using MRI and polysomnography: implications for the outpatient treatment of obstructive sleep apnea syndrome. Laryngoscope 106(2 Pt 1): 124-8. Thompson DR, Sutton TW, Jowett NI, Pohl JE (1991). Circadian variation in the frequency of onset of chest pain in acute myocardial infarction. British Heart Journal 65(4): 177-8. Tirosh E & Borochowitz Z (1992). Sleep apnea in fragile X syndrome. American Journal of Medical Genetics 43(1-2): 124-7. Trachsel L, Schreiber W, Holsboer F, Pollmacher T (1994). Endotoxin enhances EEG alpha and beta power in human sleep. Sleep 17(2): 132-9. Tremel F, Pepin JL, Veale D, Wuyam B, Siche JP, Mallion JM, Levy P (1999). High prevalence and persistence of sleep apnoea in patients referred for acute left ventricular failure and medically treated over 2 months. European Heart Journal 20(16): 1201-9. Tsai WH, Flemons WW, Whitelaw WA, Remmers JE (1999). A comparison of apnea- hypopnea indices derived from different definitions of hypopnea. American Journal of Respiratory & Critical Care Medicine 159(1): 43-8. Tusiewicz K, Bryan AC, Froese AB (1977a). Contributions of changing rib cage-- diaphragm interactions to the ventilatory depression of halothane anesthesia. Anesthesiology 47(4): 327-37. Tusiewicz K, Moldofsky H, Bryan AC, Bryan MH (1977b). Mechanics of the rib cage and diaphragm during sleep. Journal of Applied Physiology: Respiratory, Environmental & Exercise Physiology 43(4): 600-2. Valnicek SM, Zuker RM, Halpern LM, Roy WL (1994). Perioperative complications of superior pharyngeal flap surgery in children. Plastic & Reconstructive Surgery 93(5): 954-8. van den Elsen M, Sarton E, Teppema L, Berkenbosch A, Dahan A (1998). Influence of 0.1 minimum alveolar concentration of sevoflurane, desflurane and isoflurane on dynamic ventilatory response to hypercapnia in humans. British Journal of Anaesthesia 80(2): 174-82. van den Elsen MJ, Dahan A, Berkenbosch A, DeGoede J, van Kleef JW, Olievier IC (1994). Does subanesthetic isoflurane affect the ventilatory response to acute isocapnic hypoxia in healthy volunteers? Anesthesiology 81(4): 860-7; discussion 26A. van Lunteren E (1991). Upper-Airway Effects on Breathing. In: The Lung: Scientific Foundations. R. G. Crystal, J. B. West and et al., Eds. New York, Raven Press, Ltd.: 1631-44. van Lunteren E, Van de Graaff WB, Parker DM, Mitra J, Haxhiu MA, Strohl KP, Cherniack NS (1984). Nasal and laryngeal reflex responses to negative upper airway pressure. Journal of Applied Physiology: Respiratory, Environmental & Exercise Physiology 56(3): 746-52. VanDercar DH, Martinez AP, De Lisser EA (1991). Sleep apnea syndromes: a potential contraindication for patient-controlled analgesia. Anesthesiology 74(3): 623-4. Vieito Amor M, Hernandez Iniesta J, Santiveri X, Garcia C, Maestre P, Villalonga A, Ruiz B (2002). [Morbidity and mortality related to anesthesia and surgery in 60

234 patients treated with bariatric surgery]. Revista Espanola de Anestesiologia y Reanimacion 49(7): 365-72. Wallace DG, Hines DJ, Whishaw IQ (2002). Quantification of a single exploratory trip reveals hippocampal formation mediated dead reckoning. Journal of Neuroscience Methods 113(2): 131-45. Warren PM, Taylor JH, Nicholson KE, Wraith PK, Drummond GB (2000). Influence of tramadol on the ventilatory response to hypoxia in humans. British Journal of Anaesthesia 85(2): 211-6. Warwick JP & Mason DG (1998). Obstructive sleep apnoea syndrome in children. Anaesthesia 53(6): 571-9. Watanabe M, Nakagawa M, Nobe S, Ohie T, Takahashi N, Hara M, Yonemochi H, Ito M, Saikawa T (2002). Circadian variation of short-lasting asymptomatic paroxysmal supraventricular tachycardia. Journal of Electrocardiology 35(2): 135-8. Welch PD (1967). The use of the fast Fourier transform for the estimation of power spectra: a method based on time averaging over short, modified periodograms. IEEE Transactions on Audio and Electroacoustics 15: 70-3. Whittaker P & Canham PB (1991). Demonstration of quantitative fabric analysis of tendon collagen using two-dimensional polarized light microscopy. Matrix 11(1): 56-62. Wiegand L, Zwillich CW, White DP (1988). Sleep and the ventilatory response to resistive loading in normal men. Journal of Applied Physiology 64(3): 1186-95. Wilcox I, Grunstein RR, Hedner JA, Doyle J, Collins FL, Fletcher PJ, Kelly DT, Sullivan CE (1993). Effect of nasal continuous positive airway pressure during sleep on 24-hour blood pressure in obstructive sleep apnea. Sleep 16(6): 539-44. Wilcox I, McNamara SG, Wessendorf T, Willson GN, Piper AJ, Sullivan CE (1998). Prognosis and sleep disordered breathing in heart failure. Thorax 53 (Suppl 3): S33-6. Williams RL, Agnew HW, Jr., Webb WB (1964). Sleep patterns in young adults: An EEG study. Electroencephalography & Clinical Neurophysiology 17: 376. Willich SN, Levy D, Rocco MB, Tofler GH, Stone PH, Muller JE (1987). Circadian variation in the incidence of sudden cardiac death in the Framingham Heart Study population. American Journal of Cardiology 60(10): 801-6. Wilson PA, Skatrud JB, Dempsey JA (1984). Effects of slow wave sleep on ventilatory compensation to inspiratory elastic loading. Respiration Physiology 55(1): 103- 20. Wolpert C, Jung W, Herwig S, Schimpf R, Luderitz B (2001). [A new dual-chamber defibrillation system: duration, frequency and circadian variation of recurrent supraventricular tachyarrhythmias]. Herz 26(1): 55-63. Worsnop C, Kay A, Kim Y, Trinder J, Pierce R (2000). Effect of age on sleep onset- related changes in respiratory pump and upper airway muscle function. Journal of Applied Physiology 88(5): 1831-9. Worsnop C, Kay A, Pierce R, Kim Y, Trinder J (1998). Activity of respiratory pump and upper airway muscles during sleep onset. Journal of Applied Physiology 85(3): 908-20. Wu A & Drummond GB (2003). Sleep arousal after lower abdominal surgery and relation to recovery from respiratory obstruction. Anesthesiology 99(6): 1295- 302. Yamasaki F, Seo H, Furuno T, Hamashige N, Kawai K, Owaki T, Hujinaga H, Group KS, Kochi Acute Myocardial I (2002). Effect of age on chronological variation

235 of acute myocardial infarction onset: study in Japan. Clinical & Experimental Hypertension (New York) 24(1-2): 1-9. Younes M (2004). Role of arousals in the pathogenesis of obstructive sleep apnea. American Journal of Respiratory & Critical Care Medicine 169(5): 623-33. Young CH, Drummond GB, Warren PM (1993a). Effect of a sub-anaesthetic concentration of halothane on the ventilatory response to sustained hypoxia in healthy humans. British Journal of Anaesthesia 71(5): 642-7. Young ML & Hanson CW (1993b). An alternative to tracheostomy following transsphenoidal hypophysectomy in a patient with acromegaly and sleep apnea. Anesthesia & Analgesia 76(2): 446-9. Young T, Palta M, Dempsey J, Skatrud J, Weber S, Badr S (1993c). The occurrence of sleep-disordered breathing among middle-aged adults. New England Journal of Medicine 328(17): 1230-5. Young T, Peppard PE, Gottlieb DJ (2002). Epidemiology of obstructive sleep apnea: a population health perspective. American Journal of Respiratory & Critical Care Medicine 165(9): 1217-39.

236 APPENDIX 1 - Research Study Report John Loadsman

Department of Anaesthetics Patient: Subject 1 UR: 0000000 Study date:

Sex: M Date of birth: Weight: 77 Height: BMI:

SLEEP STATISTICS

Report time from 22:00:00 to 05:33:29 = 453.5 min Time available for sleep (lights out) = 453.5 min Sleep latency = 19.0 min REM latency = 44.5 min Sleep period from 22:19:00 to 05:32:59 = 434.0 min Total time awake during sleep period = 52.5 min Stage 1 = 18.5 min 4.9% Total Sleep = 374.5 min Stage 2 = 179.0 min 47.8% NREM Sleep = 280.0 min 74.8% Stage 3 = 80.0 min 21.4% REM Sleep = 94.5 min 25.2% Stage 4 = 2.5 min 0.7% Movement time = 7.0 min Unsure time = 5.5 min Sleep Efficiency = 82.6%

RESPIRATORY / SLEEP STATISTICS

NREM REM Back Other All Back Other All

SaO2% min average 96 95 95 95 95 95 SaO2% lowest 93 89 89 92 89 89 Total duration (min) Unsure 0.0 0.0 0.0 0.0 0.0 0.0 Central Apnea 0.0 2.1 2.1 0.0 0.5 0.5 Obstructive Apnea 0.0 0.7 0.7 0.0 0.4 0.4 Mixed Apnea 0.0 3.8 3.8 0.0 0.5 0.5 Hypopnea 5.1 21.1 26.2 0.2 8.8 9.0 Apnea+Hypopnea 5.1 27.6 32.7 0.2 10.2 10.4 RDI Unsure 0.0 0.0 0.0 0.0 0.0 0.0 Central Apnea 0.0 2.0 1.7 0.0 1.5 1.3 Obstructive Apnea 0.0 0.5 0.4 0.0 0.8 0.6 Mixed Apnea 0.0 2.5 2.1 0.0 0.8 0.6 Hypopnea 15.7 11.3 12.0 4.0 11.3 10.2 Apnea+Hypopnea 15.7 16.4 16.3 4.0 14.3 12.7 16.3 12.7

Total RDI = 15.4 SaO2 awake average = 96 % Average SaO2 desaturation = 3 % Mean Apnea / Hypopnea duration = 27.0 sec Longest Apnea = 37 sec Longest Hypopnea = 92 sec

237 APPENDIX 1 - Research Study Report John Loadsman

Department of Anaesthetics Patient: Subject 1 UR: 0000000 Study date: REM MOV AWK 1 2 3 4

FL BR

100 SaO2

+5 Cn.A +5 Ob.A +5 Mx.A +5 Hyp +5 Uns

120

Time (Hrs) 96% 3 SaO2 % Time 91-100 6.27 81-90 0.02

1.5 71-80 0 61-70 0 51-60 0 <= 50 0.01 0<= 50 60 70 80 90 100 SaO2 %

238 APPENDIX 1 - Research Study Report John Loadsman

Department of Anaesthetics Patient: Subject 1 UR: 0000000 Study date:

Desaturation % Number of

>= 2 235 >= 3 76 >= 4 42 >= 5 16

Time (%) BPM % 25 120-111 0 110-101 0.29 100-91 2.38 90-81 3.2 12.5 80-71 9.28 70-61 71.5 60-51 8.15 0 50-41 1.25 20 40 60 80 100 120 40-31 2.35 30-21 0.91 BPM

AROUSAL STATISTICS (Sleep time)

REM NREM Total Number of Other 10 28 38 FLA 1 0 1 Apnea/Hyp 12 72 84 Post-resp 0 2 2 0 0 0 125 Per hour Other 6.3 6.0 6.1 FLA 0.6 0.0 0.2 Apnea/Hyp 7.6 15.4 13.5 Post-resp 0.0 0.4 0.3 0.0 0.0 0.0 20.0

239 APPENDIX 2

Royal Prince Alfred Hospital

Missenden Road, Camperdown, N. S. W. 2050, Australia.

DEPARTMENT OF ANAESTHETICS

Fax: (02) 9519 2455 Telephone: (02) 9515 6111 Extension No. 58564 or 58507

RESEARCH STUDY INTO SLEEP AND BREATHING BEFORE AND AFTER SURGERY INFORMATION FOR PARTICIPANTS

You are invited to take part in a research study into sleep and breathing before and after surgery. The objective is to investigate the changes that occur in sleep patterns after operations and whether or not these changes lead to problems with breathing. It is now known that after you have an operation your sleep patterns may be disturbed for several days, not just in ways that are obvious to you like waking up when you have pain for example, but also in ways that can only be detected by measuring your brain’s electrical waves while you sleep. We think that these disturbances in sleep might be causing some people to have trouble with their breathing and possibly also with their heart and this is what we want to find out. The study is being conducted by Dr Loadsman and Professor Baker from the Department of Anaesthetics and Professor Sullivan from the Royal Prince Alfred Hospital Sleep Unit.

If you agree to participate in this study we will monitor you on the night before your operation (to check your preoperative sleep patterns), for as many nights after the operation as we can (to a maximum of seven), and, whenever possible, while you are napping during the day as well. The things we need to monitor are your brain waves, the movements of your eyes and chin muscles, the amount of oxygen in your blood, your breathing movements and your heart’s electrical rhythm or ECG. The monitor uses small electrodes or “dots” that stick on your skin, a “peg” that goes on your finger or toe, and a small sensor under your nose. The leads from the electrodes will be connected to a small box that records the information. They can be disconnected whenever you need to go to the bathroom or for whatever reason if they are going to get in the way.

240 We will arrange them in a way that makes sure you are inconvenienced by them as little as possible. Apart from the inconvenience of having these leads attached to you, there will be no extra side effects or risks, and your normal treatment will not be affected in any way.

All aspects of the study, including results, will be strictly confidential and only the investigators named above will have access to information on participants. A report of the study may be submitted for publication, but individual participants will not be identifiable in such a report.

While we intend that this research study furthers medical knowledge and may improve our management of patients after surgery in the future, it may not be of direct benefit to you.

Participation in this study is entirely voluntary: you are not obliged to participate and - if you do participate - you can withdraw at any time. Whatever your decision it will not affect your medical treatment or your relationship with medical staff.

When you have read this information, Dr Loadsman will discuss it with you further and answer any questions you may have. If you would like to know more at any stage, please feel free to contact Dr Loadsman via the hospital switchboard (02) 9515 6111. This information sheet is for you to keep.

Any person with concerns or complaints about the conduct of a research study can contact the Secretary of the Ethics Review Committee of the Central Sydney Area Health Service on (02) 9515 6766.

241