The Visual Scoring of Sleep and Arousal in Infants and Children

The Visual Scoring of Sleep and Arousal in Infants and Children Madeleine Grigg-Damberger, M.D. 1 ; David Gozal, M.D. 2 ; Carole L. Marcus, M.B.B.Ch. 3 ; Stuart F. Quan, M.D. 4 ; Carol L. Rosen, M.D. 5 ; Ronald D. Chervin, M.D. 6 ; Merrill Wise, M.D. 7 ; Daniel L. Picchietti, M.D. 8 ; Stephan H. Sheldon, D.O. 9 ; Conrad Iber, M.D. 10 1 University of New Mexico School of Medicine, Albuquerque, NM; 2 University of Louisville School of Medicine, Louisville, KY; 3 Children s Hospital of Philadelphia, University of Pennsylvania, Philadelphia, PA; 4 Sleep Disorders Center, University of Arizona, Tucson, AZ ; 5 Case Western Reserve University School of Medicine, Cleveland, OH; 6 University of Michigan School of Medicine, Ann Arbor, MI; 7 Methodist Healthcare Sleep Disorders Center, Memphis, TN; 8 University of Illinois and Carle Clinic, Urbana, IL; 9 Northwestern University, Feinberg School of Medicine, Chicago, IL; 10 University of Minnesota, Minneapolis, MN Abstract: Age is probably the single most crucial factor determining how humans sleep. Age and level of vigilance significantly influence the electroencephalogram (EEG) and the polysomnogram (PSG). The Pediatric Task Force provide an evidence-based review of the age-related development of the polysomnographic features of sleep in neonates, infants, and children, assessing the reliability and validity of these features, and assessing alternative methods of measurement. We used this annotated supporting text to develop rules for scoring sleep and arousals in infants and children. A pediatric EEG or PSG can only be determined to be normal by assessing whether the EEG patterns are appropriate for maturational age. Sleep in infants at term can be scored as NREM and REM sleep because all the polysomnographic and EEG features of REM sleep are present and quiet sleep, if not NREM sleep, is at least not REM sleep. The dominant posterior rhythm (DPR) of relaxed wakefulness increases in frequency with age: 1) 3.5-4.5 Hz in 75% of normal infants by 3-4 months post-term; 2) 5-6 Hz in most infants 5-6 months postterm; 3) 6 Hz in 70% of normal children by 2 months of age; and 3) 8 Hz (range 7.5-9.5 Hz) in 82% of normal children age 3 years, 9 Hz in 65% of 9-year-olds, and 10 Hz in 65% of 15-year-old controls. Sleep spindles in children occur independently at two different frequencies and two different scalp locations: 11.0-12.75 Hz over the frontal and 13.0-14.75 Hz over the centroparietal electrodes; these findings are most prominent in children younger than 13 years. Centroparietal spikes are often maximal over the vertex (C z ), less often maximal over the left central (C 3 ) or right Disclosure Statement This is was not an industry supported study. Dr. Grigg-Damberger has financial interests in GlaxoSmithKline and Sanofi-Aventis. Dr. Gozal is on the speakers bureau for Merck. Dr. Marcus has received research support and use of equipment from Respironics. Dr. Quan has participated in speaking engagements for Takeda and has received research support and/or use of equipment from Respironics. Dr. Rosen has received research support from Cephalon. Dr. Chervin is on the scientific advisory board of Pavad Medical and is a consultant for Alexa Pharmaceuticals. Drs. Wise, Picchietti, Sheldon, and Iber have indicated no financial conflicts of interest. Submitted for publication February 1, 2007 Accepted for publication March 15, 2007 Address correspondence to: Madeleine M. Grigg-Damberger, M.D., University of New Mexico School of Medicine, Department of Neurology, 915 Camino de Salud, NE, ACC-2, Albuquerque, NM 87131-0001, Tel: (505) 272-3342; Fax: (505) 272-6692; E-mail: mgriggd@salud.unm.edu central (C 4 ) EEG derivation. About 50% of sleep spindles within a particular infant s PSG are asynchronous before 6 months of age, 30% at 1 year. Based on this, we recommend that: 1) sleep spindles be scored as a polysomnographic signature of NREM stage 2 sleep (N2) at whatever age they are first seen in a PSG, typically present by 2 to 3 months postterm; 2) identify and score sleep spindles from the frontal and centroparietal EEG derivations, especially in infants and children younger than 13 years. NREM sleep in an infant or child can be scored if the dominant posterior rhythm occupies <50% of a 30-second epoch, and one or more of the following EEG patterns appear: 1) a diffuse lower voltage mixed frequency activity; 2) hypnagogic hypersynchrony; 3) rhythmic anterior theta of drowsiness; 4) diffuse high voltage occipital delta slowing; 5) runs or bursts of diffuse, frontal, frontocentral, or occipital maximal rhythmic 3-5 Hz slowing; 6) vertex sharp waves; and/or 7) post-arousal hypersynchrony. K complexes first appear 5 months post-term and are usually present by 6 months post-term, whereas clearly recognizable vertex sharp waves are most often seen 16 months post-term. Vertex sharp waves are best seen over the central (C z, C 3, C 4 ) and K complexes over the frontal (F z, F 3, F 4 ) electrodes. Slow wave activity (SWA) of slow wave sleep (SWS) is first seen as early as 2 to 3 months post-term and is usually present 4 to 4.5 months post-term. SWA of SWS in an infant or child often has a peak-to-peak amplitude of 100 to 400 μv. Based on consensus voting we recommended scoring N1, N2, and N3 corresponding to NREM 1, 2, and SWS whenever it was recognizable in an infant s PSG, usually by 4 to 4.5 months post-term (as early as 2-3 months post-term). Epochs of NREM sleep which contain no sleep spindles, K complexes, or SWA would be scored as N1; those which contain either K complexes or sleep spindles and <20% SWS as N2, and those in which >20% of the 30-second epoch contain 0.5 to 2 Hz >75 μv (usually 100-400 μv) activity as N3. The DPR should be scored in the EEG channel that is best observed, (typically occipital), but DPR reactive to eye opening can be seen in central electrodes. Because sleep spindles occur independently over the frontal and central regions in children, they should be scored whether they occur in the frontal or central regions. Because sleep spindles are asynchronous before age 2 years, simultaneous recording of left and right frontal and central activity may be warranted in children 1-2 years of age. Simultaneous recording of left, right, and midline central electrodes may be appropriate because of the asynchronous nature of sleep spindles before age 2 years, but reliability testing is needed. Evidence has shown that the PSG cannot reliably be used to identify neurological deficits or to predict behavior or outcome in infants because of significant diversity of results, even in normal infants. Journal of Clinical Sleep Medicine, Vol. 3, No. 2, 2007 201

M Grigg-Damberger, D Gozal, CL Marcus et al Normal sleep EEG patterns and architecture are present in the first year of life, even in infants with severe neurological compromise. Increasing evidence suggests that sleep and its disorders play critical roles in the development of healthy children and healthy adults thereafter. Reliability studies comparing head-to-head different scoring criteria, recording techniques, and derivations are needed so that future scoring recommendations can be based on evidence rather than consensus opinion. We need research comparing clinical outcomes with PSG measures to better inform clinicians and families exactly what meaning a PSG has in evaluating a child s suspected sleep disorder. Keywords: Children, EEG, infants, REM, NREM, pediatric, PSG, scoring, sleep, visual. Citation: Grigg-Damberger M; Gozal D; Marcus CL et al. The visual scoring of sleep and arousal in infants and children: development of polygraphic features, reliability, validity, and alternative methods. J Clin Sleep Med 2007:3(2);201-240 1.0 HISTORICAL PERSPECTIVE Age is probably the single most crucial factor that determines how humans sleep. Infants spend half their time sleeping, adults only about one-third. Age and level of vigilance also influence the electroencephalogram (EEG) and the polysomnogram (PSG). Development, maturation, and involution characterize the curve of our lives, and our sleep patterns change accordingly. Aserinsky (as a graduate student at the University of Chicago) was first assigned by his advisor, Kleitman, to study sleep not in adults, but infants. At that time, experimental studies of sleep in infants were based solely on behavioral observations. 1,2 Between 1949 and 1952, Aserinsky sat watching about two dozen infants sleep, studying whether eye movements in infants correlated with sleep depth. He found that sleeping infants exhibited a recurring motility cycle manifested by ocular and gross bodily activity. 3 These observations prompted study of whether these changes also occurred in adults. Kleitman then encouraged Aserinsky to learn Jacobson s technique for recording eye movements in human subjects while awake. 4 Aserinsky and Kleitman then used this approach for recording eye movements (electro-oculography, [EOG]) and coupled it with EEG to record a two-channel polygraph which confirmed that rapid eye movements occurred in regular recurring cycles across the night during sleep, and that these movements were accompanied by increased peaks of gross bodily activity, a low voltage mixed frequency EEG, and more rapid respiration. 5 Further investigation led to the landmark 1957 paper of Dement and Kleitman, in which they coined the term, REM sleep, identifying it by the presence of rapid eye movements and low voltage fast EEG activity and observing that NREM and REM sleep alternated cyclically across a night of sleep. 6 All of these seminal discoveries began by watching infants sleep. Aserinsky said that when he presented his work on the study of sleep in infants as part of his doctoral thesis defense, his examiners told him that doctorates are awarded not for obtained findings but in hope of future results. 7 How fortunate for those of us who practice sleep medicine this proved true. Beginning in late 1950s, Dreyfus-Brisac and Monod 8-10 and Parmalee et al 11-13 began to study and report the distinctive EEG patterns of sleep in infants. These investigators found that normal post-term infants demonstrated 2 distinctive sleep states which they called active sleep and quiet sleep: quiet sleep (QS) is characterized by preserved chin EMG, few body movements, regular respiration and heart rate, and no eye movements; active sleep (AS) is characterized by rapid eye movements, frequent small face and limb movements, irregular respiration and heart rate, and the absence of or minimal chin EMG activity. These observations remain valid today. By the late 1960s, the distinctive EEG patterns of sleep in neonates and infants were delineated: Dreyfus-Brisac 14,15 and Parmelee et al described how EEG patterns in premature infants evolved with age. 13 In 1968, Parmelee et al provided a detailed catalog of EEG frequencies, amplitudes, and patterns demonstrating how the characteristic EEG patterns of sleep change in premature infants with increasing conceptional age (CA). 13 Using this catalog, they were able to predict an infant s CA within 2 weeks in 85% of infant EEGs scored blindly. 13 These early EEG masters applied power spectra analysis to confirm the validity of their visual interpretations. 12 They found that quiet sleep in infants by the time they reach term is characterized by one of two EEG patterns: tracé alternant or high voltage slow (HVS) activity. Tracé alternant is an EEG pattern in which 3-8 second bursts of moderate to high voltage 0.5-3.0 Hz slow waves intermixed with 2-4 Hz sharply contoured waveforms alternate with 4- to 8-second intervals of attenuated mixed frequency EEG activity; because this pattern alternates between activity and much less activity it is considered to be discontinuous. 12 In contrast, HVS consists of continuous moderately rhythmic 50-150 μv 0.5-4 Hz slow activity, without the bursting activity of tracé alternant. HVS represents the more mature pattern of quiet sleep in infants. Two EEG patterns are observed in active sleep in infants at term: a continuous EEG pattern called activité moyenne which consists of either low voltage (<50 μv) 5-6 Hz activity or a mixture of high and low voltage activity including delta activity called mixed (M). 9,16-18 Credit for applying a developmental perspective to the study of sleep in humans belongs to Roffwarg, Muzio, and Dement who published their classic description of sleep state ontogenesis in Science in 1966. 19,20 As a psychiatrist, Roffwarg, was interested in the relationship between dreaming and REM sleep physiology. 20 Anders reported that Roffwarg initially presumed neonates would not have REM sleep because they did not dream. 20 Instead, Roffwarg and colleagues found infants spend half of their total sleep time in REM sleep, double that of young adults; findings which prompted them to theorize that REM sleep must play an important role in fostering development and maturation of the immature brain. In 1969, the Association of the Psychophysiological Study of Sleep (APSS) chartered an ad hoc committee to develop a guide for scoring sleep in infants because the sleep scoring criteria published by Rechtschaffen and Kales 21 were applicable only to the adult and had not taken into account the unique features of the developing infant. 22 This committee chaired by Anders, Emde, and Parmelee, was composed of investigators with considerable expertise in infant sleep research. Under the auspices of the UCLA Brain Information Service, they met several times in 1969-1970, and developed consensus-derived terminology, scoring criteria, and a manual containing illustrative examples for scoring sleep in normal newborn term infants. A Manual for Standardized Terminology, Techniques, and Criteria for Scoring Journal of Clinical Sleep Medicine, Vol. 3, No. 2, 2007 202

Sleep scoring in infants and children of States of Sleep and Wakefulness in Newborn Infants, first published in 1971, 22 was designed to increase comparability of research results. Like the Rechtschaffen and Kales scoring manual, it was not intended for clinical purposes. The authors hoped it would be complemented in the future by manuals dealing with premature and older infants. Since the 1971 publication of what this paper calls the Anders manual, several criteria for scoring sleep in infants and/or children have been published, but none have been widely accepted or used. Guilleminault and Souquet 23 first published in 1979 their criteria for scoring sleep in infants between 6 weeks and 12 months of age because they thought the Anders manual overlooked much information contained in the EEG of older infants. They reported they had used these rules to score sleep in over 400 nocturnal or 24-hour PSGs. Hoppenbrouwers reported in 1987 that she and her colleagues at the University of California in Los Angeles had used their sleep scoring criteria in over 400 overnight-psg studies in infants between birth and 6 months of age. 24 Crowell et al (1997) published a set of sleep, arousal, and respiratory scoring criteria which they used to score PSG recorded at home in 415 infants (35 to 64 weeks conceptional age) as part of a multi-center prospective study. 25 In 1999, Scholle and Schäfer 26 published guidelines developed by the Pediatric Task Force of the German Sleep Research Society, which they reported represented an age-appropriate adaptation of Rechtschaffen and Kales. 21 Despite this rich legacy of scientific endeavor, those who study sleep in children still have no universally accepted criteria for scoring sleep in pediatric populations. Most clinicians use Rechtschaffen and Kales rules, which were never developed for pediatric subjects. In 2004, the Board of Directors of the American Academy of Sleep Medicine (AASM) decided that a new sleep scoring manual was needed. 27 They hoped the new manual would be based on digital PSG recording techniques, incorporate the effects of age and pathology, and address not only visual sleep stage scoring, but also scoring rules for arousals, movements, and respiratory and cardiac events during sleep. Six task forces were organized, one of which was devoted to pediatric issues. The Pediatric Task Force (see page 233) initially acted as a liaison to other task forces, assisting them in addressing issues unique to pediatric populations. We reviewed position papers, participated in consensus voting, and assisted in the development of rules by the other task forces. We helped develop specific pediatric rules for respiratory and cardiac events which are summarized in the respiratory and cardiac review papers. As we reviewed the adult visual sleep stage scoring and arousal scoring papers, it became apparent to the Pediatric Task Force and the Scoring Manual Steering Committee that it was important to write specific rules and terminology for visual scoring of sleep in infants and children. Although in the end, the Pediatric Task Force provided no different rules for scoring arousals in infants and children than those recommended for adults, we performed an evidence-based review of this material, discussed it at length, and cite it here to explain the decisions we eventually reached regarding scoring arousals in infants and children. This review paper summarizes the evidence used to support the terminology and rules contained in the new scoring manual for visual scoring of sleep and arousals in infants and children. 2.0 METHODS In November 2005, the Pediatric Task Force recommended a separate pediatric review paper on visual scoring of sleep and arousals in infants and children; the Scoring Manual Steering Committee agreed with this recommendation. For the purpose of this review, the Pediatric Task Force, like other task forces was to: 1) use an evidence-based medicine process for identifying and grading evidence; 2) produce tables following methods used by the American Academy of Sleep Medicine (AASM) Standards of Practice Committee; 3) develop a referenced evidence review paper that would serve as annotated supporting text for the manual recommendations; and 4) provide a ranked list of considerations to the steering committee for developing scoring rules. The Pediatric Task Force did a systematic and comprehensive evidence-based review of the medical literature. In order to review relevant literature amenable to evidence-based analysis, the Pediatric Task Force first formulated the following questions regarding analysis and scoring of sleep and wakefulness in neonates, infants, and children. The questions we formulated are shown in Table 1. With these questions in mind, we performed our first computer-based Medline literature search using the PubMed search engine on November 29, 2005 using the following key words: REM sleep, stage 4 sleep, stage 3 sleep, stage 2 sleep, stage 1 sleep, sleep onset, alpha AND sleep, delta AND sleep, drowsiness AND normal, eye blinks, eye movements AND sleep, K complex, spindles, sleep staging, arousal, sleep disruption. We restricted the search to child (0 to 18 years) and all human studies published in English between 1995 and 2005 year-to-date. The first search yielded 1,658 citations. A second search covering the time period of 1966 to 2004, done on December 27,2005, identified 160 additional citations. These two searches found 1,818 articles; of these, 242 were deemed relevant to this topic after review of the title and the abstracts. Additional topic-focused searches were performed. The most recent on August 16, 2006, using a process known as pearling in which additional articles were identified from the bibliographies of the articles previously cited and the articles we used to write this paper. We reviewed all abstracts relevant to pediatric sleep scoring or arousal reliability and validity. We assessed reliability by evaluating test-retest reliability and inter- and intrascorer reliability. Validity evidence was evaluated by identifying physiological correlates, outcome effects, and/or comparison to a standard. Ultimately, a total of 344 references were used as evidence and as a basis to write this review. In general, included papers had to present evidence relevant to the scoring of sleep and arousals in infants and children. Exclusion criteria included abstracts, case studies, editorials, reviews, and theoretical papers. However, when relevant, we considered and sometimes used the latter types of citations as background information to help write the general introduction and discussion sections of this review paper. The Pediatric Task Force then prepared evidence tables which helped the Pediatric Task Force and Scoring Manual Steering Committee to develop terminology, rules, standards, guidelines, and recommendations for scoring sleep and arousals in infants and children. Evidence tables for this paper will be placed on the AASM website (which can be accessed on the web at www.aasmnet.org). We graded the evidence using a classification approach used by the AASM Standards of Practice Committee (Table 2), which is a modification of the classification system developed by Sackett. 28 Journal of Clinical Sleep Medicine, Vol. 3, No. 2, 2007 203

M Grigg-Damberger, D Gozal, CL Marcus et al Table 1 Questions formulated by the Pediatric Task Force: analyzing and scoring sleep/wakefulness in infants, children, and adolescents 1. How should we define the following: neonate, preterm, full term, infant, conceptional age (CA), and child? 2. Are there existing criteria for scoring sleep in infants and children? 3. Are active sleep and quiet sleep immature forms of REM and NREM sleep, respectively? 4. At what age following term birth are sleep spindles usually present? 5. Are there features of sleep spindles that evolve by age in infants, children, and adolescents? 6. At what age following term birth are K complexes and vertex sharp waves usually present? 7. At what age does slow wave activity (SWA) of slow wave sleep (SWS) appear in normal infants following term birth? 8. After what age following term birth can we identify and score NREM sleep as stages 1, 2, and SWS? 9. How does the dominant posterior rhythm (DPR) of relaxed wakefulness develop and change with age in infants, children, and adolescents? 10. How does the waking EEG background change with age in infants, children, and adolescents? 11. Are there distinctive paroxysmal EEG patterns seen in pediatric subjects in the transition between wake/sleep and stage 1 NREM sleep? 12. How does the waking EEG background change with age in infants, children, and adolescents? 13. What are the characteristic EEG backgrounds of drowsiness and stage 1 NREM sleep (stage N1) in infants and children? 14. Are there distinctive paroxysmal EEG patterns seen in pediatric subjects in the transition between wake/sleep and stage 1 NREM sleep? 15. What is known about the ontogeny of the development of the non-eeg physiological measures of REM sleep (rapid eye movements, chin EMG and limb atonia, irregular respiration, phasic muscle twitches, and gross body movements in infants? How can we use these to identify REM sleep in infants? 16. What is the evidence for the reliability of recording and scoring sleep in in-laboratory polysomnograms in infants and children? 17. What are the indications for polysomnography in infants, children, and adolescents? 18. What is the evidence for the validity of scoring sleep in infants and children? 19. Until what age following term birth should we use the Anders infant sleep scoring criteria? 20. Which epoch length should be used to score sleep in infants and is there compelling evidence that the scoring interval should be different than adults? 21. Over which EEG electrodes are dominant posterior rhythm, sleep spindles (SS), K complexes (KC), slow wave activity (SWA), sawtooth waves best seen in pediatric subjects? Is the scalp topography of these distinctive polysomnographic features different in infants, children, or adolescents, compared with adults? 22. Are significant first night effects observed when recording a single night of in-laboratory PSG in a child? 23. What technical considerations are appropriate when recording PSG in pediatric subjects? Are any of these different from those recommended for adults? When insufficient evidence was available for specific items, the Pediatric Task Force used the UCLA/Appropriateness Method 29 to develop consensus agreement. We submitted our evidence review and consensus balloting of the Pediatric Task Force to the steering committee who, with the Pediatric Task Force chair, crafted the final scoring rules, standards, guidelines, consensus recommendations, and technical specifications. Different members of the Table 2 Evidence classification used by the Pediatric Task Force Evidence Study Design Levels 1 Randomized well-designed trials with low-alpha & lowbeta errors 2 Randomized trials with high-beta errors 3 Nonrandomized controlled or concurrent cohort studies which study a reasonably well defined sample of adequate sample size, using standardized techniques 4 Nonrandomized uncontrolled historical cohort or observational studies 5 Case reports, case series, or observational studies not fulfilling the criteria of Level 4 Pediatric Task Force researched, wrote, and submitted portions of the review paper. A subgroup of the task force (listed as authors) reviewed and edited this work and wrote this review paper. All members of the Pediatric Task Force* were directors or members of sleep disorders centers which either exclusively, usually, or often record sleep in infants and children. All members of the Task Force and the steering committee completed detailed conflict-of-interest statements, none of whom reported inherent conflicts of interest related to this subject. The Pediatric Task Force met by telephone conference call 10 times between June 21, 2004, and April 9, 2006, for evidence review and RAND/ UCLA consensus voting. 3.0 THE VISUAL SCORING OF SLEEP AND AROUSAL IN INFANTS AND CHILDREN 3.1 Definitions of age, term, and conceptional age in infants and children Normalcy of an infant s EEG or PSG can only be determined by assessing whether the EEG patterns are appropriate for maturational age. 30 Normalcy of an infant EEG is determined not by chronological age (number of days or weeks following birth) but by conceptional age (CA). The American Academy of Pediatrics (AAP) published in 2004 recommendations for age terminology to more accurately define length of gestation and age in neonates, infants and children. 31 The AAP recommends that gestational age (GA) be defined as the time elapsed in completed weeks between the first day of the last normal menstrual period and the day of delivery. A fetus with a GA of 25 weeks, 5 days is still considered a 25-week fetus; rounding the GA up to 26 weeks is regarded as inconsistent with national and international norms. 32 If the pregnancy was achieved using assisted reproductive technology, GA is calculated by adding 2 weeks to the CA. Chronological age is the time in days, weeks, months, or years from birth. The AAP policy statement recommends abandonment of the term conceptional age because of its many inherent inaccuracies. 33 However, conceptional age (CA) must be defined for this paper because much of the evidence we present defines age-appropriate EEG patterns in neonates and infants in relation to conceptional age. CA is calculated by adding the estimated gestational age to the chronological age. The AAP policy now recommends we use the following terminology: 1) When a neonate is still in the hospital following birth, postmenstrual age is the preferred term defined by the gestational age plus the chronological age in weeks; 2) After this Journal of Clinical Sleep Medicine, Vol. 3, No. 2, 2007 204

Sleep scoring in infants and children perinatal period, corrected age is the preferred term, defined as the chronological age in weeks or months reduced by the number of weeks born before 40 weeks of gestation. Corrected age should be used only for children born prematurely and only until 3 years of age. In future studies, we recommend reporting data using this new age-related terminology. By convention, prematurity is conceptional age less than 38 wks, and full term is 38 to 42 weeks CA. A neonate is a child during the first 28 days after birth. An infant is a child age 1 to 12 months of age. A child is someone younger than 18 years of age. From a physiological standpoint, an adolescent is probably best defined using Tanner sexual maturity staging. 34 The importance of defining an infant s conceptional age arises because the EEG (or a PSG) of a normal infant is more dependent upon the age of the brain following conception than the number of days following birth. 35,36 Except when stressed, or in situations involving encephalopathy or medication-related factors, the EEG or PSG of a neonate reflects the actual developmental age of the brain. 37 The brain, EEG, and PSG of an infant continue to develop and mature at a similar rate, independent of whether the infant is in utero or post-delivery. An EEG or PSG of a normal premature infant born at 32 weeks gestational age whose chronological age is 8 weeks should resemble that of a normal infant born at 40 weeks gestational age two days earlier. The EEG and PSG patterns observed in infants 6 months or younger correlate most closely with the infant s CA; after that the number of months in age post-term birth usually suffices. 3.2 Definitions of electroencephalography crucial to understanding arguments in this review The frequency range for scalp-derived EEG in humans lies between 0.3 to 70 Hz (cycles per second). 38 By convention, normal EEG background is divided into 4 main frequency bands: delta (0.5 to <4 Hz); theta (4 to <8 Hz); alpha (8 to 13 Hz); and beta (>13 Hz). 39 Most EEG laboratories use the International 10/20 system to place electrodes in standardized scalp locations. This system places electrodes at 10 and 20 percent deviations from four anatomical landmarks (the nasal bridge (nasion), the inion (occipital proturberance), and the right and the left preauricular points (depression in front of each ear). Electrodes are named by their location (e.g., F, C, O for frontal, central, and occipital, respectively); those on the left side of the head are given odd numbers; those on the right are even; and the midline derivations are denoted with z. Electrode names of particular interest to those involved with PSG are: frontal (F 3, F 4 ), central (C 3, C 4 ), occipital (O 1, O 2 ), and midline (F z, C z, O z ). These standard deviations and electrode position names are used in this paper. There are 2 fundamental methods of representing EEG activity: referential and bipolar derivations (also called montages). A referential montage links an active electrode placed over a biologically active site to a reference electrode placed over a relatively inactive site such as the earlobe (A 1, A 2 ) or the mastoid (M 1, M 2 ). The signal represents a recording of the electropotential difference between the recording site of interest and the reference (comparison) site. The Rechtschaffen and Kales scoring manual recommended central EEG derivations (C 3 -A 2, C 4 -A 1 ) for scoring sleep stages; these are referential derivations. 21 A bipolar montage links two biologically active electrodes to each other, and the signal represents the net electrical negative or positive potential difference between the Concept of Surface-Negative and Surface- Positive Deflection in EEG and PSG Negative deflection of a K-complex Positive deflection of a K-complex Figure 1 Concept of surface-negative and surface-positive deflection in EEG and PSG. By convention, upward pen (or digital) deflections mean a net negative potential difference between two electrodes; a downward deflection means a positive electrical difference between two electrodes. A biphasic K complex waveform has two phases, an initial upward negative deflection followed by a positive downward deflection. Sometimes, the waveform which is predominantly upward is described as surface negative and downward as surface positive because much of the EEG is a radial electrical dipole. 2 electrodes (e.g., F z -C z which links the midline frontal [F z ] to the midline central [C z ] electrode). By convention, upward pen (or digital screen) deflections indicate a negative potential difference between 2 electrodes; a downward deflection indicates a positive electrical difference between the 2 electrodes. For example (Figure 1), a K complex is a biphasic wave (i.e., the complex crosses the EEG baseline twice, once up (negative) then down (positive). Sometimes, we call the waveform which is predominantly upward surface negative and downward surface positive because most of the normal background EEG is a radial electrical dipole. 3.3. Earlier criteria for scoring sleep and wakefulness in pediatric subjects 3.3.1 The Anders manual for scoring sleep/wake states in full-term newborns The Anders manual defined sleep scoring criteria and terminology only for normal full-term newborns. It differs from the Rechtschaffen and Kales manual in that it provides a system for coding only sleep/wake states (and their behavioral correlates), leaving specific criteria for scoring states to the investigator. 22 The Anders manual provides a system for coding respiration, eye movements, muscle tone, respiration, movements, and vocalizations to distinguish states of sleep and wakefulness. Like Rechtschaffen and Kales, the criteria proposed by the Anders manual were consensus-based, reflecting the views, experience, and medical research available to the authors. Anders and his coauthors classified sleep into 3 states: active sleep (AS), quiet sleep (QS), and indeterminate sleep (IS); they further presumed that QS and AS were, respectively, antecedents of NREM and REM sleep. 22 The authors recognized and defined 3 states of wakefulness ( crying, active awake, and quiet awake ) and 4 EEG patterns of sleep (tracé alternant and HVS in quiet sleep, low voltage irregular (LVI) or M in active sleep) in post-term infants. The Anders criteria emphasized that behavioral Journal of Clinical Sleep Medicine, Vol. 3, No. 2, 2007 205

M Grigg-Damberger, D Gozal, CL Marcus et al observations were necessary to differentiate between states of sleep and wakefulness in the infant and were critical for accurate interpretation of polysomnographic recordings (PSG) in infants. 16,19,40 The minimum behavioral notations needed to interpret each epoch of an infant PSG were: 1) eyes open or closed; 2) presence or absence of body movements; and 3) crying (when present). The Anders manual recommended scoring an entire sleep record from Lights Out to Lights On using either 20-second or 30-second epochs. The authors thought the most useful polysomnographic characteristic for scoring sleep in infants was the regularity or irregularity of respiration. They defined regularity or irregularity of respiration by measuring the variability in respiratory rate between the longest and shortest respiratory cycle within a 20- or 30-second epoch, extrapolating these rates for one minute. Regular respiration was then a period in which the respiratory rate varied <20 breaths per minute; irregular respiration when it varied by >20 breaths per minute. 10 Appendix Table 1 (which can be accessed on the web at www.aasmnet.org) summarizes the EEG and polysomnographic features of active sleep (AS), quiet sleep (QS), and indeterminate sleep (IS) of the Anders manual. 22 3.3.2 Hoppenbrouwers sleep scoring criteria for infants, birth to 6 months Hoppenbrouwers sleep scoring criteria were developed to score sleep in infants from birth to 6 months post-term. 24 Similar criteria in the Anders manual, there were 4 recognized 4 sleep/wake states: awake (AW), active sleep (AS), quiet sleep (QS), and indeterminate sleep (IS). The Hoppenbrouwers criteria emphasized the importance of behavioral correlates, especially regularity or irregularity of respiration, in distinguishing AS and QS (particularly when the EEG background is one of the continuous EEG patterns). They scored sleep using 60-second epoch lengths. To score active sleep (AS) in infants from birth to 6 months post-term using the Hoppenbrouwers criteria requires the absence of sustained EMG tonus together with three of the following criteria: 1) at least one eye movement, independent of chin and gross body movements; 2) breathing rate variation greater than 25 breaths per minute as measured by the respiratory tachometer; 3) presence of twitches and brief head movements; and/or 4) absence of EEG spindles or tracé alternant. Scoring quiet sleep (QS) requires all of the following: 1) breathing variation of no greater than 25 breaths per minute as measured by the respiratory tachometer; 2) eyes are closed with no more than one isolated eye movement; 3) sustained EMG tonus, EEG spindles, tracé alternant, or both. To score awake (AW) requires at least three of the following criteria be met: 1) Sustained EMG tonus with activity bursts; 2) Eyes open; 3) Within a given minute breathing rate variation greater than 45 breaths per minute as measured by the respiratory tachometer; 4) Vocalization; and/or 5) Sustained gross movements. Epochs of sleep in infants in which the criteria for AW, AS, and QS are not fulfilled, or minutes in which these criteria are fulfilled for less than 30 consecutive seconds (state transitions), are scored as indeterminate sleep (IS). 3.3.3 Guilleminault and Souquet sleep scoring criteria for infants, 3-12 months Guilleminault and Souquet criteria recommended scoring sleep and wake in infants 3 to 12 months post-term based upon EEG, EOG, chin EMG, and behavior correlates, using 30-second epoch lengths. 23 Their scoring criteria reflect how EEG patterns of wakefulness, sleep onset, NREM, and REM sleep change with age. Sleep onset in infants 3 to <6 months post-term was heralded by a high amplitude burst of >100 μv theta activity. Stage 1 NREM sleep in infants 3 to <6 months post-term consisted of a wide range of mixed 1-15 Hz EEG frequencies, if <20% of the 30-second epoch contained delta slow waves >150 μv. If sleep spindles were present, the epoch would be scored as stage 2. Sleep onset in infants 6 to 12 months post-term was heralded by a burst of regular high amplitude theta waves and stage 1 by mixed (1-12 Hz) EEG frequencies, but with a predominance of 3-7 Hz theta activity. They distinguished stage 1 from stage 2 by the absence of sleep spindles in stage 1 and by <20% SWS (>150 μv <2 Hz delta waves) in the 30-second epoch. Characteristic for stage 2 sleep in infants 6-12 months of age was predominant theta activity, 12-14 Hz sleep spindles, and <20% of the epoch with >150 μv <2 Hz delta waves. SWS was scored whenever >20% of the 30-second epoch contained >150 μv <2 Hz delta waves. Their criteria characterized the EEG background of REM sleep in infants 3 to 12 months post-term as a predominance of theta activity, again emphasizing the importance of relying upon behavioral correlates (EOG, chin EMG, movement) to identify REM sleep in infants. The authors recommended continuing to score epochs as stage 2 unless the spindle-to-spindle interval was longer than 5 minutes (versus the 3-minute rule of Rechtschaffen and Kales). Appendix Table 2 (which can be accessed on the web at www.aasmnet.org) summarizes the Guilleminault/Souquet criteria for scoring sleep in infants 6 weeks or older. 23 3.3.4 Crowell infant sleep scoring criteria Crowell et al scored sleep in 415 infants (34 to 64 weeks CA) using criteria they developed for recording PSG at home as part of the CHIME study (a prospective multi-center collaborative home infant monitoring study). 25 They used C 4 -A 1 as the primary EEG derivation and C 3 -A 2 as a back-up channel in case the first channel malfunctioned. To score a particular EEG pattern, it had to occupy >50% of a 30-second epoch. As others had earlier, they recognized two different continuous EEG patterns during active sleep. They used the same name as others had earlier for Mixed which they described as high and low voltage waves with little periodicity. However, they called low voltage mixed frequencies fast, characterized by low voltage <35 μv 5-8 Hz theta intermixed at times with 1-5 Hz activity. They recognized and scored three different EEG patterns during quiet sleep: high voltage (>50 μv, 0.5 to 4.0 Hz for >50% of the 30-second epoch), Mixed (as defined for active sleep), or trace alternant. Tracé alternant was scored when an alternating pattern of 3 or more runs of high voltage activity lasting 3 to 8 seconds alternating with low voltage activity lasting 4 to 12 seconds occurred, cautioning that periodicity needed to be readily observed and allowing scorers to look at the preceding or following 30-second epochs to find the 3 patterns. They defined regularity or irregularity of respiration using Anders criteria and methodology. Criteria required eye movements to be out-of-phase, similarly shaped, and produce synchronous deflections of at least 30 μv or more in both EOG channels. Journal of Clinical Sleep Medicine, Vol. 3, No. 2, 2007 206

Sleep scoring in infants and children 3.3.5 Scholle and Schäfer criteria for scoring sleep in infants and children The Pediatric Task Force found only one set of criteria devised for scoring sleep in children older than infants. Scholle and Schäfer published in Somnologie (the official journal of the German Sleep Research Society) criteria for scoring sleep in infants and children accompanying an atlas, which shows representative normal PSGs in infants and children. 26 The Pediatric Task Force of the German Research Society who developed these criteria said they were needed because EEG amplitudes in infants and children during NREM sleep are much higher than in an adults so that the Rechtschaffen and Kales >75 μv amplitude criterion is not applicable to children. They said these age-appropriate sleep scoring criteria were adapted from Rechtschaffen and Kales. 21 Significant features of the Scholle/ Schäfer criteria include: 1) noting the dominant posterior alpha rhythm of relaxed wakefulness increases in frequency with increasing age; 2) the EEG patterns of stage 1 NREM sleep include either vertex sharp waves, low voltage theta and delta activity, or hypnagogic waves; 3) the high voltage of slow wave activity to score slow wave sleep is gauged by comparing it to the mean amplitude of EEG activity observed during stage 2 NREM sleep. They provided the following age-related frequencies for the posterior alpha rhythm in children: starting at 2 to 4 months, 2-4 Hz; 6-7 Hz at 12 months; 5 to 8 Hz at 1 to 3 years; 6-8 to 7-9 Hz at 3 to 5 years; 8-10 to 9-11 Hz at 5 to 12 years; and 10 Hz (8.5-13 Hz) >12 years. The Scholle and Schäfer criteria also reported that EMG muscle tone was high during wakefulness, progressively decreased with deepening NREM sleep reaching its lowest level during REM sleep. EOG during wakefulness included eye blinks and rapid eye movements, slow eye movements during NREM 1, and rapid eye movements during REM sleep. Sleep spindles need to last 0.5 seconds or more to score them as a feature of NREM 2. Sleep spindles and broad K complexes may be seen in NREM 3; sleep spindles are observed sometimes even in NREM 4. NREM 3 sleep was identified by high voltage delta waves occupying 20% to 50% of an epoch, and greater than 50% of epoch characterized NREM 4 sleep. Of note, they recommended identifying high voltage slow wave sleep in children by the delta waves higher than the mean amplitude of NREM 2 sleep. Finally, the Scholle and Schäfer sleep scoring criteria 26 identified typical EEG patterns seen during sleep and noted how these changes with maturation which they modified from Niedermeyer and Lopes da Silva. 41 Scholle and Schafer criteria specify that: 1) 12-15 Hz pre-spindles may be seen before 10 weeks of age and note that immature sleep spindles are often rounded negative and have a sharp positive component; 2) progressive increases of inter-spindle intervals occur with increasing age; 3) from 18 to 24 months, mature sleep spindles without a rounded component are observed; 4) anterior 12 Hz and central 14 Hz sleep spindles are seen from age 2 years on; and 5) 14 Hz sleep spindles maximal at the vertex (Cz) with independent 12 Hz sleep spindles over the anterior regions after 3 years of age. High amplitude K complexes are first seen at 5 months of age, fully developed by age 2 years; vertex sharp waves are first seen at 8 weeks of age and are most prominent between ages 3 to 10 years. So-called REM storms, are rapid eye movement burst patterns, first seen after 28 weeks conceptional age and maximal at 33 to 36 weeks; rapid eye movement inter-burst intervals typically declined to less than one per second beginning between ages 2 months and 12 months. Finally, they characterized the age-related changes in the sleep/wake transition as: 1) indeterminate between ages 33 to 44 weeks conceptional age; 2) rhythmic theta activity increasing in amplitude compared to wakefulness was often noted beginning at 6 months of age; 3) Hypnagogic hypersynchrony (rhythmic 4 to 6 Hz theta activity) first observed age 1 year, progressively declines and seldom seen after age 12 years. 3.3.6 Criteria for scoring sleep in infants based solely upon behavioral observations Early studies of scoring or staging sleep in infants were based solely upon behavioral observations. 1-2 In 1964, Prechtl and Beintema developed a scale for scoring sleep/wake states in infants based solely on observable behaviors. 345 They classified sleep/wake in infants 36 to 44 weeks CA into 5 different stages based upon behavioral features: 1) eyes closed, regular respiration, no movements; 2) eyes closed, irregular respiration, no gross movements; 3) eyes open, no gross movements; 4) eyes open, gross movements, no crying; 5) eyes open or closed, crying. Even after the development of the Anders manual, which combined behavioral and polysomnographic correlates, Anders and many of his colleagues continued to study infants using only behavioral observations. One such behavioral scoring criteria used by Anders and Chalemian 270 classified wakefulness into 4 different states: 1) Fussy- Cry (FC) characterized by the presence of vigorous diffuse motor activity and varying intensities of vocalization; 2) Wakeful Activity (WA) with frequent spurts of diffuse motor activity, open eyes, and occasional grunts; 3) Alert Inactivity (AI) with occasional directed motor actions and wide open eyes that pursued targets; and 4) Drowsy (DR) with relative immobility, absence of focused attention, and opening and closing of eyes. Quiet sleep (QS) was identified by regular respiration and the absence of eye and body movements except for an occasional nonnutritive sucking and active/rem (AR) sleep when rapid eye movements, facial grimacing, writhing body movements, irregular respiration, and isolated limb twitches were observed. 3.4 Are quiet sleep and active sleep immature forms of NREM and REM sleep, respectively? The authors of the Anders manual presumed that quiet sleep and active sleep were immature forms of NREM and REM sleep, respectively. Conflicting evidence and opinions on this issue are justified in part by findings in neonatal animal models whose maturational state is equivalent to that of very immature infants. Gramsbergen 42 and Frank and Heller 43 argue that sleep in infant rats younger than 12 days of age represents a primitive undifferentiated behavioral state an amalgam of sleep components, which gradually differentiate into REM and NREM sleep later in development. They further argued that AS was best considered as an undifferentiated behavioral state (which they called presleep) from which both SWS and REM sleep eventually emerge. Other investigators have argued that the development of sleep in mammals represents a continuous elaboration of the components that are already in place and identifiable in early infancy. 44-46 Recently, Karlsson et al 47 demonstrated that all the neuronal connections which generate muscle atonia and myoclonic twitching Journal of Clinical Sleep Medicine, Vol. 3, No. 2, 2007 207

M Grigg-Damberger, D Gozal, CL Marcus et al during AS are already well established in neonatal rats equivalent in developmental age to human infants 32 weeks CA. This prompted the argument that if all the REM sleep generator mechanisms and connections are present in infant rats at birth and are similar to those seen in adult cats, then it is likely that AS in infants by 32 weeks CA is simply an immature form of REM sleep seen in older children and adults. Although it seems intuitive, we actually have no evidence that QS is an immature form of slow wave sleep. Some have suggest that quiet sleep before the appearance of sleep spindles, K complexes, or SWA is an undifferentiated sleep state and simply not AS/REM sleep. Whether the high voltage 0.75 to 1.75 Hz activity which emerges as early as 2 months post-term represents EEG activity generated by the same neuronal generators and connections as SWA seen at a later developmental age is controversial. 48,49 More research and evidence is needed. Summary: Are active and quiet sleep immature forms of REM and NREM sleep, respectively? The Pediatric Task Force discussed whether to continue using the traditional terms quiet sleep (QS) and active sleep (AS) when scoring infants, especially those younger than 3 months post-term. The Pediatric Task Force used the following reasoning to recommend that sleep in infants 2 months post-term should be scored as NREM and REM sleep: 1) all the EEG and polysomnographic features of REM sleep are present by this age; 2) convenience and simplicity; and, 3) quiet sleep, if not NREM sleep by this age, is at least not REM sleep. 3.5 Development of the non-eeg polysomnographic features of REM and NREM sleep in infants Readers less experienced in scoring sleep in young infants may at first be puzzled by the inclusion of this topic so early in the position paper. Under normal circumstances, REM sleep and its behavioral correlates would be appropriate for discussion after those of wakefulness and NREM sleep. However, sleep onset is typically REM sleep in infants <3 months post-term, and REM sleep constitutes about 50% of the total sleep time (TST) in term infants. Finally, differentiating quiet sleep and active sleep cannot be done based solely on EEG when an infant s sleep EEG is a continuous pattern of HVS (seen in QS at term) or mixed (seen in the first cycle of AS in term infants). Because active and quiet sleep cannot be differentiated in neonates and young infants solely based upon the EEG, the behavioral correlates (i.e., the non-eeg physiological correlates) of REM (and NREM) sleep are needed to differentiate QS from AS. 22,50 3.5.1 EEG and behavioral correlates of active and quiet sleep clearly recognized by 36 weeks conceptional age Sleep is undifferentiated in human infants before 32 weeks CA. By approximately 32 weeks CA, active sleep and quiet sleep can be distinguished by their behavioral correlates, but not by the stillundifferentiated EEG activity. 17,51 At approximately 32 weeks CA, rapid eye movements and phasic muscle twitches identify AS, and QS is associated with the presence of far fewer movements (except for an occasional myoclonic jerk and some buccolingual movements). 17 Recognizable EEG patterns of AS and QS first appear at approximately 34 weeks CA. By 36 weeks CA, all the EEG and behavioral correlates of wakefulness, active sleep, and quiet sleep are clearly recognizable, although large percentages of sleep are still best scored as indeterminate sleep. Indeterminate sleep represents discordant sleep where mixtures of > 2 sleep/wake states are seen within a given epoch. The percentage of indeterminate sleep falls rapidly after 36 weeks CA. Regular heart rate and respiration, elevated chin EMG tone, and absence of eye movements are observed in epochs of well-differentiated QS in infants at 40 weeks CA. 16,52,53 Normal infants at 40 weeks CA exhibit a sizeable repertoire of motor behaviors: sucking movements, fine muscle twitches, chin, body and limb tremors, grimaces, smiles, intermittent stretching, and large athetoid limb movements. 54,55 Early infant sleep researchers found many epochs of sleep are best scored as indeterminate sleep as late as 3 months post-term. Coons and Guilleminault scored 33% ± 7% of total sleep time as indeterminate sleep in 10 normal infants at age 3 weeks post-term, 32% ± 7% at 6 weeks. 56 Parmalee et al reported that 67% of total sleep time was IS in an infant 30 weeks CA, 38% at 40 weeks CA, and falls to 29% 3 months post-term, 57 still a sizeable percentage of the total sleep time. Anders and Sostek found they were able to significantly reduce the amount of indeterminate sleep they scored in a PSG by permitting scoring of AS, QS, or AW if <2 polysomnographic correlates were discordant for the sleep state. 58 DeWeerd and van den Bossche emphasize in their excellent review of the development of sleep during the first months of life that state of a child at one particular moment can only be assessed approximately 50% to 80% of the time, and the remaining periods are indeterminate. 59 They caution that while it is desirable to have a global idea of which state an infant is in at particular recording time, a procrustean approach if taken too often, leads to forced and often meaningless scoring of sleep/wake states in a young infant s PSG. Indeterminate sleep in infants is particularly challenging when trying to train an automatic computer scoring system. 3.5.2 Ontogeny of rapid eye movements Rapid eye movements (REMs) are first seen as early as 31 weeks CA or even earlier. 17,60-62 Parmelee et al (1969, 1972) reported that eye movements occurred almost continuously in infants at 28 to 30 weeks CA but then only at a rate of 1 to 4 per minute. 11,52 However, by 32 weeks CA, REMs when seen often occur in clusters: 1 to 6 eye movements per minute in 37% of 20- second epochs, 9 per minute in 13% of epochs. Premature infants will often display increased REM density. Clusters of intense REMs in premature infants have been called REM storms. 63 Kohyama et al used PSG to study REM densities in 32 normal infants 33 to 184 weeks CA, finding that REM densities were highest in infants 36 to 38 weeks CA. 64 Becker and Thoman studied REM densities in PSGs of 15 normal infants 2 to 5 weeks of age and again at 3, 6, and 12 months. They found that REM storms markedly decreased after 40 weeks CA. Their data suggested that the continued presence of REM storms after 6 months of age could indicate developmental delay or dysfunction. 11,63 Of note, the Scholle/Schäfer criteria call the increased REM density of preterm infants the burst pattern of rapid eye movements, say it is characterized by high density bursts with inter-burst intervals of less than one second, first seen after 28 weeks, is maximal 33 to 36 weeks CA, and decreases across the first year of life. 26 Journal of Clinical Sleep Medicine, Vol. 3, No. 2, 2007 208