Quantitative Electroencephalogram (EEG) in ...

3446

Current Pharmaceutical Design, 2008, 14, 3446-3455

Quantitative Electroencephalogram (EEG) in Insomnia: A New Window on Pathophysiological Mechanisms Cristina Marzano1, Michele Ferrara2, Emilia Sforza3 and Luigi De Gennaro1,* 1

Department of Psychology, “Sapienza” University of Rome, Italy; 2Department of Health Sciences, University of L’Aquila, Italy and 3Laboratoire d'Explorations Fonctionnelles du Système Nerveux Service de Neurologie CHU, Reims, France Abstract: In the last two decades quantitative electroencephalogram (EEG) analysis has been widely used to investigate the neurophysiological characteristics of insomnia. These studies provided evidence in support of the hypothesis that primary insomnia is associated with hyperarousal of central nervous system and altered sleep homeostasis. However, we have here underlined that these results have intrinsic methodological problems, mainly related to constraints of standard assessment in clinical research. We have proposed that future studies should be performed on larger samples of drug-free patients, using within-subjects designs and longitudinally recording patients adapted to sleep laboratory. All these methodological improvements will allow to partial out the contribution of individual differences, pharmacological influences and first-night effects on EEG frequencies. Moreover, we have discussed the potential relevance of recent findings from basic research concerning local changes during physiological sleep, which could be extended to the study of insomnia. We have suggested that, if normal sleep exhibits specific regional characteristics, also disorders in initiating and maintaining sleep should be characterized by local changes. The extension of this theoretical framework to the study of insomnia could provide new insights on the underlying pathophysiological mechanisms. As a first step toward the integration of knowledge from basic and clinical research focused on local sleep changes, here we showed some preliminary data from sleep onset recordings of patients with paradoxical insomnia. This approach supports the heuristic potential of our proposal, pointing to a local functional impairment in the process of synchronization in insomniac patients compared to normal subjects, the former exhibiting more beta and less delta and sigma power on anterior scalp locations than the latter.

Key Words: Insomnia; electroencephalogram (EEG); EEG power; EEG topography, local sleep; paradoxical insomnia; individual differences. INTRODUCTION Insomnia is a common and widespread complaint with a significant impact on both night-time and daytime functioning. Insomnia has been mainly explained by behavioral and neurocognitive models [e.g., 1, 2]. Especially in the last two decades, quantitative electroencephalogram (EEG) analysis has been used to investigate the neurophysiological characteristics of insomnia, starting from knowledge obtained from basic sleep research. For instance, quantitative EEG analysis showed that EEG power density in the low frequency range is an indicator of a progressively declining process during sleep whose initial value is determined by the duration of prior waking [3]. The amount of slow wave activity (SWA) in non-rapid eye movement (NREM) sleep is considered a marker of NREM sleep intensity and the electrophysiological correlate of a sleep-wake dependent ‘Process S’ underlying sleep homeostasis [4], a process influenced by different physiological [5] or experimental conditions [6]. Many investigations carried out using quantitative analysis support the hypothesis that primary insomnia is associated with hyperarousal of central nervous system (CNS), because patients with insomnia (PI) exhibited increased high *Address correspondence to this author at the Department of Psychology, “Sapienza” University of Rome, Italy; E-mail: [email protected]

1381-6128/08 $55.00+.00

frequency EEG activity during both sleep onset and all-night sleep [e.g., 7, 8]. Some studies support the hypothesis that sleep homeostasis is altered in primary insomnia, as expressed by a slow wave sleep (SWS) deficiency [e.g., 9], and that homeostatic dysregulation may represent a predisposing, precipitating and/or perpetuating factor of insomnia. Recently, it was also suggested that hyperarousal and altered sleep homeostasis (and even circadian dysregulation) may interact simultaneously in chronic insomnia [10]. Nevertheless, the body of research that provided these results has some intrinsic limits, often ascribed to standard assessment and clinical research methods. (1) Betweensubjects designs seem to be inadequate to assume generalizable characteristics in patients with sleep disorders. Evidence from basic sleep research suggest that normal sleep is characterized by large individual differences [e.g., 11, 12], which could constitute a confounding factor in the evaluation of the physiological basis in pathological sleep. A growing body of evidence points to genetic influences on normal and pathological sleep, in humans and in animals [13]. As an example, it has been shown that a stable frequency-specific (8.0-15.5 Hz) pattern of EEG topography along the antero-posterior cortical axis during NREM sleep distinguishes each individual like a “fingerprint” [11], reflecting genetic influences [12]. Therefore, further studies with larger samples of patients and within-subjects designs (i.e., longitudinal studies) © 2008 Bentham Science Publishers Ltd.

Quantitative Electroencephalogram (EEG) in Insomnia

are needed. (2) Several studies on PI used EEG data derived from the first in-laboratory study night. One single night of recording seems to be inadequate to assess the existence of stable EEG alterations and the ‘first-night effect’ [14-16] might interfere with results. (3) Some studies were performed on not completely drug-free PI, making it difficult to discriminate the specific pharmacological contribution on rapid frequency EEG bands. The aim of this review is to suggest a different point of view about the employment of quantitative EEG analysis in clinical research. First, after a brief section on the behavioral and neurocognitive models of insomnia, we outline how quantitative EEG analysis provided support to the hypothesis that insomnia is characterized by central nervous system hyperarousal and homeostatic dysfunction, considering the current methodological limitations of these studies. Following the literature review, we discuss the potential relevance of the recent findings from basic research concerning local changes during physiological sleep, which could provide new insights on the underlying mechanisms in pathological sleep. Finally, we suggest perspectives for future research, by an exemplification of integration of knowledge from basic and clinical research. THE BEHAVIORAL AND MODELS OF INSOMNIA

NEUROCOGNITIVE

Since 1980s, several theories about the etiology of insomnia have been proposed. The behavioral model [1] posits that trait (psychological, biological, and social components) and precipitating factors (acute occurrences of life stress events) result in acute insomnia, which is reinforced by maladaptive coping strategies. These strategies, in turn, result in conditioned arousal and chronic insomnia. Two constructs of the conditioned arousal have received particular attention: somatic and cognitive arousal. In fact it was shown that PI are physiologically hyperaroused prior to sleep onset and/or during sleep [17, 18] and characterized by intrusive cognitions around sleep onset [19, 20]. Perlis et al. [2] introduced a neurocognitive perspective that specifically focuses on cortical arousal. The neurocognitive model is based on the observation that beta and gamma EEG activity are typically associated with cognitive processes [e.g., 21, 22] and that these high frequency EEG activities are enhanced in insomnia at/or around sleep onset [e.g., 7, 23]. This model proposes that in chronic insomnia high frequency EEG activity occurs as a result of classical conditioning, since this activity is elicited in response to the visual and/or temporal cues usually associated with sleepiness and sleep. This high level of cortical arousal allows for increased sensory processing, information processing and memory formation that, in turn, interfere with sleep. Recently, Espie et al. [24] proposed the attention-intention-effort pathway, an explanatory model about the developmental and maintaining factors of primary insomnia. This model suggests that ‘de-arousal’ process and sleep engagement in PI may be particularly vulnerable if the process is switched out of its natural automated mode. Thus, from this point of view insomnia may be explained as the failure to inhibit wakefulness. Persistent insomnia is the result of a selective attention bias to sleep, leading to active intention

Current Pharmaceutical Design, 2008, Vol. 14, No. 32

3447

and striking effort to sleep, with pervasive sleep preoccupation and a host of mental and behavioral strategies used by the patient with the aim of inducing sleep. Summarizing, the behavioral and neurocognitive approaches provide an alternative point of view respect to physiological approaches to explain insomnia disorders. These methods are showing to be very effective and promising also for treatment of insomnia. Further investigations integrating behavioral, neurocognitive and physiological methods could provide a deeper understanding of this phenomenon. QUANTITATIVE EEG ANALYSIS IN THE STUDY OF PRIMARY INSOMNIA Sleep Onset Sleep onset is a unique window on insomnia problems, since patients with subjective and objective insomnia complain for difficulties in falling asleep and overestimate their sleep latencies. Studies investigating the wake-sleep transition focus their analyses on short intervals just around sleep onset. The definition of a common “set-point” is methodologically crucial when comparing controls and patients. In such way, we can go beyond the history of difficulties in falling asleep. Independently from the time course and length of the wake-sleep transition, the definition of an unequivocal sleep onset point makes comparable the transition periods of controls and patients with insomnia. Otherwise, the implicitly longer sleep latency of the insomnia patients would bias the between-group comparisons. Furthermore, the definition of sleep onset may also affect results. We have provided a robust evidence that the onset of stage 2 (i.e., the appearance of the first K complex or sleep spindle) represents a more reliable boundary between wakefulness and sleep [25]. Comparisons between studies and actual evidence of between-group differences could be affected also by this choice. In normal subjects, the wake-sleep transition is characterized by a decrease of the high frequency EEG activity and by a parallel increase of the low frequency EEG activity [25]. The patients with insomnia show a different sleep pattern during the sleep onset period respect to normal subjects. The first investigation on the spectral differences between PI and normal sleepers was performed by Freedman [23] on 12 primary sleep-onset PI and 12 age-matched normal sleepers. Comparisons were carried out on the first minute of each sleep stage, showing more beta activity in PI than normal sleepers during only stage 1 and REM sleep. Although limited to the first min of stage 1 and REM, the increased beta activity was interpreted as a physiological substratum of anxiety and cognitive processes that characterize the difficult sleep onset of these patients. In a study by Merica and Gaillard [7], 12 patients with sleep maintenance insomnia and 23 controls were compared during sleep onset (i.e., during a period of 25 min in the first NREM cycle). They found higher beta activity and lower delta activity in PI than normal controls. However, in that study PI were not completely drug-free, and it is a wellconsolidated evidence that the benzodiazepines (BZ) intake is associated to an increase in beta and sigma activity and a

3448 Current Pharmaceutical Design, 2008, Vol. 14, No. 32

decrease in delta activity [26, 27]. Furthermore, the analysis on the sleep onset was performed on the relative beta and delta components (beta/delta activity index) in the EEG, without a clear possibility to discriminate the individual contribution of these two specific bands and without any information about the other EEG frequency bands. More recently, La Marche and Ogilvie [28] assessed the quantitative EEG differences between 6 patients with primary insomnia, 6 patients with psychiatric insomnia and 6 normal controls. They found that patients with psychiatric insomnia showed lower relative beta power values during sleep onset period. Relative beta power during wakefulness was higher in patients with primary insomnia than patients with psychiatric insomnia and normal controls. Therefore, the few studies here reviewed seem to indicate that sleep onset in PI is characterized by an increased beta activity. However, their small number and the methodological concerns listed above strongly suggest the need for studies with larger samples of drug-free patients, to be carried out on clearly defined sleep onset intervals. All-Night Sleep The structure of sleep is well established for chronic insomnia, because polysomnographic (PSG) data and macrostructural characteristics of these patients have been extensively studied. Insomnia patients exhibit sleep maintenance difficulties and poor sleep with frequent intrusions of wakefulness. In order to determine if insomnia may be related to an hyperarousal of the CNS, many studies have examined the spectral characteristics of the whole night sleep EEG. In order to confirm the hypothesis that beta activity during sleep occurs specifically in association with primary insomnia, Perlis et al. [8] compared 9 patients with primary insomnia, 9 patients with secondary insomnia (insomnia related to major depression) and 9 normal controls, using power spectral analysis (PSA) technique. Patients with primary insomnia exhibited more beta-1 (14-20 Hz), beta-2 (20-35 Hz) and gamma (35-45 Hz) activity during NREM sleep than either patients with secondary insomnia or normal sleepers, confirming that a high frequency activity occurs predominantly in primary insomnia. A limitation of this study is that the results were not presented for each sleep stage; moreover, stage 1 was included in the NREM amount. In other words, there is no way to understand if the results are due to a higher amount of stage 1 in patients with primary insomnia (PSG data of stage 1 duration were not reported). In addition, the data for this study were derived from the first in-laboratory study night, so the ‘first-night effect’ might be a confounding factor [14-16], even though one study did not found remarkable differences between the first and second night analyzed by means of Fast Fourier Transform (FFT) [29]. In the before mentioned study [6], the authors also evaluated whether high frequency activity is associated with discrepancies between subjective and objective (PSG) measures of sleep continuity (sleep latency, number of awakenings, wake after sleep onset and total sleep time), because patients with insomnia have the tendency to underestimate their total sleep time and to overestimate their sleep problems. Correlational analysis revealed that NREM beta-1 and beta-2 activity were negatively correlated with

Marzano et al.

subjective-objective discrepancy measures for total sleep time. This correlational analysis was performed on the distribution of discrepancies of both patients and control groups. In such way, the intrinsic differences between controls, patients with primary insomnia and patients with secondary insomnia in the reported subjective/objective mismatch about total sleep time weaken this approach. It has been hypothesized that beta power density may reflect the level of a central arousal mechanism, as delta power density has been considered to reflect sleep intensity [30]. It might be expected that PI would also exhibit a lower level of delta activity: nevertheless, a clear and direct evidence supporting this hypothesis is still lacking. In order to specifically assess if sleep maintenance insomnia could be associated to a low level of ‘Process S’ [4] that might explain the frequent intrusions of wakefulness in sleep and the SWA deficiency in response to sleep deprivation in PI, Besset et al. [9] evaluated the effect of partial (21 hours) sleep deprivation, restricting sleep of 7 patients with sleep maintenance insomnia and of 7 normal sleepers to the first three hours of sleep. For each cycle, SWA values were expressed as a percentage of the average of the all-night sleep recording. Although PI showed a lower SWA than controls during the recovery night, the SWA level in PI was higher in recovery night than in baseline night indicating that the homeostatic process was regularly operating. In addition, this manipulation (21 hours of sleep deprivation) seems to be inadequate to strengthen the hypothesis of an altered homeostatic process, because patients with insomnia are always partially sleep deprived. More recently, Nofzinger et al. [31] investigated the neurobiological basis of poor sleep and daytime fatigue in insomnia using [18F]Fluorodeoxyglucose (FDG) positron emission tomography (PET) in 7 insomnia patients and 20 healthy subjects. In this study the authors also assessed EEG spectral power of all sleep NREM episodes, although only scarce information about the quantitative analysis performed was provided. The two groups did not differ in delta activity (0.5-4 Hz) and beta activity (20-32 Hz), failing to show an hyperarousal of CNS or a reduced sleep intensity in PI. To summarize, the available data seem to lend some support to the hyperarousal hypothesis, while the homeostatic dysregulation needs to be further evaluated. Future studies on these issues, however, should take into account the methodological concerns raised above. All-Night Dynamics Only few studies provided data about the temporal distribution of EEG activity throughout the night and the sleep cycles. Based on the observation of an enhanced beta activity at sleep onset period, Merica et al. [32] examined in the same sample differences across the entire night (i.e., the first four NREM/REM sleep cycles) finding that the increased beta activity in PI persisted throughout the night both in NREM and REM sleep. Procedural details cast some doubts, since there was no mention to an artefact rejection and, above all, only episodes that contained periods of wakefulness lasting 5 min or more were discarded from the analyses. The PSG showed that PI had a higher amount of waking



after sleep onset than healthy subjects, and this may affect the higher beta activity found in PI respect to normal controls. Perlis et al. [33] also evaluated the temporal and stagewise distribution of high frequency activity in the same three groups of the study mentioned above [8]. The high frequency activity, as emerged by PSA, tended to increase across NREM cycles and occurred maximally during stage 1 and during REM sleep, and that it was greater in patients with primary insomnia than in the other two groups. It was also found that patients with primary insomnia exhibited greater beta power at the first epoch following sleep onset and a steeper slope following sleep onset. In addition, the high frequency activity tended to be inversely associated with delta power across the course of the night. These data may suggest that primary insomnia reflects an increase in the homeostatic regulator associated with propensity for wakefulness. As a summary, while there are many observations on increased the beta activity in PI during all-night sleep, little is known about its temporal distribution across sleep cycles. The indication that primary insomnia may be associated with an increased pressure for wakefulness deserves further investigation. EFFECTS OF NON-PHARMACOLOGICAL PHARMACOLOGICAL INTERVENTIONS

AND

The treatments for chronic insomnia include cognitive behavior therapy (CBT) and drug therapy. Quantitative analysis of EEG activity could shed light on the mechanisms involved in the efficacy of these interventions and, indirectly, on the physiological mechanisms involved in insomnia. Jacobs et al. [34], firstly, assessed treatment efficacy of multifactor behavioral intervention comparing 12 patients with chronic sleep onset insomnia and 14 normal sleepers at pre-treatment and post-treatment. Patients with insomnia exhibited a significant pre-treatment to post-treatment reduction in sleep onset polysomnographic latency and did not differ from normal sleepers at post-treatment. In addition, beta EEG power showed a significant pre-treatment to posttreatment reduction prior to sleep onset. More recently, Cervena et al. [35] evaluated the modifications occurring in EEG power densities during sleep after 8 weeks of cognitive behavioral therapy for insomnia (CBTI) in 9 drug-free patients with psychophysiological insomnia. Patients were recorded in the laboratory for two consecutive nights before and after CBT-I, and quantitative EEG analysis was performed on the second night. The results showed that SWA increased after CBT-I, and the exponential SWA decay was characterized by a smaller time constant. Sigma activity decreased after CBT-I, but probably this effect was only an indirect consequence of the increased amount of SWA, because there is an inverse relationship between SWA and spindles [27, 36-38]. As suggested by the authors, the main limit of this study is the lack of a control group of PI, as it might be hypothesized that sleep improvement is dependent on reasons other than CBT-I. In addition, these results should be taken with caution, as the therapeutic programme involved a ‘sleep restriction’ that typically produces itself an increase of SWA in normal sleepers [39].

3449

Despite CBT has been shown to be as effective as pharmacotherapy [40], benzodiazepine-receptor agonists are the most frequently prescribed treatment for insomnia. A limited number of studies used quantitative EEG analysis to examine chronic effects of BZ in PI. Short-term BZ use produces an increase in beta and sigma activity and a decrease in delta activity both in PI and healthy subjects [26, 27]. Nevertheless, it is unknown whether these effects persist, following treatment discontinuation. Indeed there is a recognized need for long-term randomized, double-blind, parallel group, controlled trials on BZ effects [41]. In the study of Bastien et al. [42], 15 PI using BZ chronically (i.e., BZ usage for at least three nights per week for more than three months), 15 drugfree PI and 16 controls (all participants were older adults) were compared during the first four cycles of sleep. BZ users exhibited less delta and theta activity over the entire night than controls and during cycle 2 than drug-free PI. In addition, BZ users had more beta1 activity during cycle 3 than controls and during cycle 4 than drug-free insomniacs and controls. The decrease of SWA and the increase of high frequency activity might explain the poor sleep quality usually reported by chronic BZ users. Altogether, these findings partially support the idea that the positive effects of CBT on insomnia symptoms should be accompanied by a decreased EEG power in the high frequencies and by an increase in the delta band, although more studies are needed to fully evaluate this issue. As regards BZ effects, BZ users show a higher cortical activation that could explain why they still subjectively perceive their sleep as disturbed even during the periods of drug intake. However, data about long-term effects on EEG power of both behavioral and pharmacologic interventions are needed. Paradoxical Insomnia Paradoxical insomnia (or sleep state misperception or subjective insomnia) is a subtype of insomnia characterized by normal PSG measures in patients having profound and, at times, dramatic sleep complaints, and so by a mismatch between subjective and objective measures. Conventional PSG analysis seems inadequate to assess the existence of specific physiological correlates and to explain the underestimation of sleep latency and of total sleep time in this disorder. In order to assess if paradoxical insomnia could be associated with increased physiological activation, Bonnet et al. [43] compared whole body metabolic rate between 9 patients with paradoxical insomnia and 9 normal sleepers. Patients with paradoxical insomnia had a significantly increased 24-hour metabolic rate compared with normal sleepers, in both the nocturnal measurements and in the daytime observations. Although the overall 6% increase in VO2 was not as large as the 9% increase reported in another study comparing patients with primary insomnia and matched normal sleepers [44], this result seems to establish a physiological basis for paradoxical insomnia. Although this insomnia subtype has been included into the classification systems many years ago, little research has been devoted to it, and only one study specifically assessed quantitative EEG changes. Krystal et al. [45] compared 12 patients with paradoxical insomnia,18 patients with primary insomnia and 20 normal


controls. Patients with paradoxical insomnia exhibited more alpha, sigma and beta activity and lower delta activity during NREM sleep than either patients with primary insomnia or normal sleepers, suggesting that increased high frequency and diminished low frequency EEG power during NREM sleep represent the physiologic correlates of this sleep complaint and reflect heightened arousal levels during sleep in patients with paradoxical insomnia. The authors also assessed the correlation between EEG power spectra and the degree of sleep underestimation. A correlational analysis revealed that delta NREM relative power was negatively associated to the degree of underestimation of total sleep time (objective-subjective difference). Although this relationship was calculated across all of the three groups (leading to the same limitation encountered by Perlis et al. [8]), this finding suggests that a diminished EEG power in the low frequency range may lead to an underestimation of sleep. Moreover, these findings point to a difference in the pathophysiologic mechanisms of paradoxical and primary insomnia that are not revealed by conventional sleep time and the other continuity measures. Cause or Effect? Independent from the robustness of the association between hyperarousal and primary insomnia, as indexed by increased high frequency EEG activity during sleep onset and NREM sleep in PI, empirical evidence is intrinsically correlative. This status of knowledge does not allow to disentangle the question of whether patients with primary insomnia show trait-like or state-like differences compared to good sleepers. In other words, the empirical finding of a higher beta power (and/or lower delta power) is a cause or an effect of a “bad night”. In fact, the quality of sleep in primary insomnia is characterized by a considerable variability between “bad” and “good” nights [e.g., 46], and empirical evidence of an EEG marker of primary insomnia in both “bad” and “good” nights would be coherent with an interpretation in terms of trait-like changes. A similar interpretation would derive from the existence of stable EEG alterations in primary insomnia. Hence, longitudinal studies and/or studies with consecutive recordings in which these patients are recorded during both their good and bad nights are highly needed. Along this vein, a preliminary study compared in primary insomnia and normal sleepers the amplitude of P300 wave, elicited during a simple auditory discrimination task before and after the best and worst nights of sleep. Primary insomnia was associated to higher P300 amplitudes only after “bad” nights; this finding speaks against the hypothesis of a stable hyperarousal in these patients, contradicting the idea of trait-like changes in primary insomnia [47]. Sleep as a Local Process With the purpose of going beyond the intrinsically correlative nature of these studies, it should be useful to take into consideration the notion of sleep as a local process, trying to move the focus of the research from structural to functional characteristics of disorders in initiating and maintaining sleep. According to the 2-process model of sleep regulation, SWA depends on the duration of previous sleep and wakefulness and represents a marker of non-REM sleep intensity, with manipulations of sleep pressure leading to clear homeo-

Marzano et al.

static recovery processes [48, 49]. In recent years, it has been demonstrated that these recovery processes are mainly local and do not involve the whole cerebral cortex to the same extent, both in animals [e.g., 50, 51] and humans [e.g., 52]. Furthermore, experience-dependent plasticity in specific neural circuits during wakefulness induce localized changes of SWA during subsequent sleep [53, 54]. Finally, a critical window for primary insomnia, that is the sleep onset process, also shows specific regional differences in normal sleep. In fact it was demonstrated that more anterior scalp locations show sleep signs before the posterior cortical areas [55]. We here suggest that, if normal sleep is a local process exhibiting specific regional characteristics, also disorders in initiating and maintaining sleep should be characterized by local changes, which could provide new insights on the pathophysiological underlying mechanisms. Toward this direction, the above-quoted study on cortical evoked responses to auditory stimuli indeed found a higher P300 amplitude in primary insomnia than in controls only in correspondence with frontal areas [47]. This finding implicitly suggests that the hyperasousal associated to primary insomnia could selectively affect these cortical areas. THE POSSIBLE INTEGRATION OF BASIC AND CLINICAL RESEARCH As a first step toward the integration of knowledge from basic and clinical research focused on local sleep changes, we have considered our data from sleep onset recordings, collected within an in-progress study at the Laboratoire d'Explorations Fonctionnelles du Système Nerveux, Service de Neurologie CHU (Reims, France), designed to explore EEG topography of paradoxical insomnia. Ten patients with paradoxical insomnia (4 M and 6 F, mean age=33.0 ±2.81 yrs), matched with 10 healthy subjects (4 M and 6 F, mean age=30.2 ±2.80 yrs), participated in this preliminary study with a polygraphic recording (Deltamed, Paris, France) from four unipolar EEG channels (Fz-A1, Cz-A1, Pz-A1, Oz-A1) and standard electrooculogram (EOG) and electromyogram (EMG) channels. They were recorded for two consecutive undisturbed nights, and sleep onset (SO) has been defined as the first epoch (20 sec) of stage 2 [25]. The EEG was digitized on-line with a storage sampling rate of 128 Hz. After an off-line artifact rejection, the digitized EEG signals were analyzed by a Fast Fourier Transform (FFT) algorithm using a 4 sec resolution. Spectra of epochs with EOG/EMG artifacts were eliminated on the basis of a visual inspection of the 4-sec epochs. For the current analysis on the wake-sleep transition, power values were calculated across a 1-30 Hz frequency range in a 1-Hz resolution, considering two 5-min intervals, before and after SO. To account for interindividual differences in EEG power, relative values have been calculated dividing the Hz by Hz EEG power during the 5-min interval after SO by that before SO. Fig. (1) (panel A) depicts the typical EEG spectrum of sleep onset for the two groups in the two consecutive nights. At sleep onset, both groups show the increased EEG synchronization within the 1-7 Hz and the 12-15 Hz frequencyranges. The slow-frequency activity has higher values at the more anterior sites, while sigma power peaks at the centroparietal sites. These macroscopic EEG changes at the sleep



3451

Fig. (1). EEG changes at sleep onset, as a function of different recording scalp sites [frontal (Fz), central (Cz), parietal (Pz), and occipital (Oz)] in 10 patients with paradoxical insomnia (PPI) and 10 good sleeper controls. A. Mean EEG activity at 1-30 Hz recorded in the two groups during their wake-sleep transition. The first two panels (from left) show for each group the relative EEG changes, expressed as the ratio between the 5-min intervals subsequent to sleep onset (sleep onset) and those preceding sleep onset (presleep). The right panel shows between group differences, expressed as the ratio between EEG power in the PPI and in the control group. B. Results of the three-way mixed-design ANOVAs, Group (patients vs. controls) x Night (1st vs. 2nd night) x Derivation (Fz, Cz, Pz, Oz) on EEG power values for each 1-Hz bin, expressed in terms of F-values of the main effects and interactions. The red-shaded areas point to significant differences. To correct for multiple comparisons, the Bonferroni correction was applied. Considering the mean correlation between the variables (r=0.78), the alpha level was then adjusted to 0.02.

onset seem to characterize both groups. However, a different picture appeared when these values were compared between groups (right side of Fig. (1A)). The ratio between EEG power spectra in the patient and control groups showed a clear differentiation of the whole EEG spectrum between groups, as patients with paradoxical insomnia have higher beta power and lower synchronized activity. Moreover, patients with paradoxical insomnia in the second night show a higher increase of beta power in the 16-22 Hz range at the more fronto-central sites. The statistical analyses reported in Fig. (1B) confirm this topographical pattern of EEG changes: patients with paradoxical insomnia have higher beta power (the main effect for Group is significant at 24 and 27 Hz) and lower delta power at the central site (the Group x Derivation interaction is significant at 3 Hz) than control subjects. Patients with paradoxical insomnia also show a lower sigma power at the more anterior sites (the Group x Night and Group x Night x Derivation interaction is significant at 12 Hz) and, during the second night, higher beta at the Fz and Cz sites (the Group x Night x Derivation interaction is significant in the 17-20 Hz range) than control subjects. Although these results have to be considered preliminary due to the small sample size, they point to a functional im-

pairment in the process of synchronization in patients with paradoxical insomnia. The anterior scalp locations, which first synchronize EEG activity in normal subjects [55], exhibit more beta activity and less delta and sigma power in patients with paradoxical insomnia as compared to the good sleeper controls. In support of this interpretation, the pattern of topographical EEG changes patients with paradoxical insomnia seems to predict the magnitude of misperception. In fact, the subjective perception of sleep onset in the only patients’ group, expressed as the discrepancy between subjective and objective measures of sleep latency in the second night, negatively correlates with delta and sigma power. In other terms, larger discrepancy scores (longer subjective sleep latency than poligraphic ones) are associated with less synchonized EEG activity, and this relationship mainly affects more anterior scalp locations. Fig. (2) shows these correlations, which reach statistical significance at 4 and 15 Hz. Other findings coming from our experience of basic sleep research provide a further insightful scenario for interpreting the paradox of sleep misperception. In fact, the earlier EEG synchronization of anterior cortical areas in normal sleepers is associated with an inversion of the posterior-to-anterior direction in functional cortical coupling, which characterizes


Marzano et al.

Fig. (2). Correlations (Pearson’s r coefficients) between the magnitude of misperception, expressed as the difference between subjective and objective (polysomnographic) measures of sleep latency, and relative EEG power at sleep onset, recorded from different scalp sites [frontal (Fz), central (Cz), parietal (Pz), and occipital (Oz)]. Data refer to the second night of patients with paradoxical insomnia. The dotted line points to significant coefficients (alpha level adjusted to 0.02).

the presleep period [56]. This inversion of the functional cortical coupling, from a posterior-to-anterior direction toward an anterior-to-posterior direction, is affected by a heightened sleep pressure, being advanced during the sleep onset period in sleep deprived subjects [57]. Based on these findings, normal sleep onset could be described as a spreading of EEG synchronization from associative prefrontal to posterior areas, and this phenomenon should putatively correspond to the perception of sleep onset. In this framework, the less synchronized (and more desynchronized) EEG activity found in our patients with paradoxical insomnia could be interpreted in terms of a weaker (presumably, delayed) spreading of EEG synchronization signals from anterior to posterior areas. We suggest that this phenomenon could be one of the neurophysiological mechanisms of misperception, at least with respect to the degree to which individuals with paradoxical insomnia overestimate their sleep latency compared with the traditional polysomnography. The current results are obviously preliminary and some interpretations have to be considered unquestionably speculative. Nevertheless, we think that they have the merit of delineating the potential strength of our proposal. FINAL REMARKS In the present review of the literature on quantitative EEG analyses in insomnia, we have first underlined the sev-

eral methodological limitations that make hard the unequivocal interpretation of the main findings in the field. Evidence has been provided of CNS hyperarousal and of possible homeostatic dysfunctions in insomnia. However, these achievements should be confirmed by further studies with i) larger samples of completely drug-free patients, ii) withinsubjects designs (i.e., longitudinal studies), iii) polygraphic recordings taken from laboratory nights following the first night of adaptation. All these methodological improvements will allow to partial out the contribution of individual differences, pharmacological influences and first-night effects on both slow (delta) and fast (beta) frequency EEG bands. Similar considerations hold true also for the study of the effects of cognitive-behavioral therapy on sleep in PI. These effects should be tested in experimental paradigms not involving sleep curtailments, in order to identify the actual impact of this non-pharmacological approach on quantitative EEG indexes or, alternatively, in paradigms with control subjects undergoing to the same sleep curtailments. The literature on sleep onset in PI is characterized by the same limitations. Hence, the study of sleep onset may offer a privileged access to the understanding of the insomnia problems. In a previous study on normal sleepers we compared the two mostly used sleep onset definitions (first epoch of stage 1 vs. first epoch of stage 2) in their ability to discriminate sleep and wakefulness [25]. We reported that the crite-


rion of stage 2 onset shows a greater ability in differentiating sleep from wakefulness at a greater number of single-Hz frequency bins, with a uniformly increasing EEG power across the 1-16 Hz range after stage 2 onset. In a subsequent study, we demonstrated that sleep onset in normal sleepers is asyncronous, since the more anterior areas first synchronize their EEG activity [55]. Moreover, we showed that the alpha rhythm spreads anteriorly as the transition progresses, suggesting that the functional meaning of the alpha band during the sleep onset period should be partially revised. In other words, we suggested that the enhancement of EEG power in the 8-11 Hz after the SO point and its anterior spreading, rather than a sign of persisting arousal in the preliminary phases of sleep, is an expression of the generalized process of synchronization. The application of the same approach to the study of sleep onset in PI is in order. It would be interesting to assess whether the behavior of alpha rhythm is similar also in people with difficulties in falling asleep. Also the evaluation of the asynchronous development of the SO process across different antero-posterior areas is very promising, and could shed light on some pathophysiological mechanisms of insomnia. It could be hypothesized that SO in PI is characterized by a slower progression of EEG synchronization from the anterior to the posterior areas, leading to a longer persistence of higher levels of activation is some part of the brain.


3453

a. Control for Individual Variability • Between-subjects designs on relatively small sample size should be avoided. Sleep is characterized by large individual differences, which could constitute a confounding factor in the evaluation of the physiological bases of insomnia sleep. Therefore, future studies with larger samples of patients and withinsubjects designs (i.e., longitudinal studies) are needed, to establish generalizable characteristics in patients with insomnia. b. Careful Sample Selection and Characterization • Special attention should be paid to carefully discriminating between Idiopathic Insomnia, Paradoxical Insomnia, and so called Psychophysiological Insomnia. • Special attention should be paid to carefully discriminating between the insomnia phenotypes (Initial, Middle, Late or Mixed Insomnia). • Samples should be characterized for their clinical history, similarly to what is done with depression (e.g., age of onset, duration of illness, duration of index episode, etc.). c. Reduction of Data Variability

As a first example of the fruitful application of this approach to clinical research, we have here reported a preliminary comparison between normal sleepers and insomnia patients during sleep onset. Results showed that anterior scalp locations exhibit more beta activity and less delta and sigma power in patients with paradoxical insomnia as compared to controls. The evidence that sleep onset shows different quantitative and topographical EEG changes in normal sleepers and patients with paradoxical insomnia strongly supports the potentiality of our proposal.

• First night effect should be avoided. One single night of polygraphic recording seems inadequate to assess the existence of stable EEG alterations because the ‘first-night effect’ might interfere with results. Thus, also studies on PI need EEG data derived from laboratory study nights following the first one.

Generally speaking, we believe that the extension of the theoretical framework of “local sleep” to the study of sleep disorders has a great heuristic potential. In the near future, the routine use of several EEG derivations in clinical studies will prove to be fundamental to clarify the strength of the two main theories of insomnia: the hyperarousal and the homeostatic dysregulation. As regards the former, the possible differences between PI and normal sleepers in topographic scalp distribution of fast EEG rhythms should be addressed. As far as the latter is concerned, it would be easily elucidated by total sleep deprivation studies with full scalp recordings.

• The polysomnographic nights should be profiled for the sleep quality during the nights preceding the study. Recording nights should be also categorized according to whether the night was characterized (subjectively and/or objectively) as a “Bad” or “Good” night.

Finally, the application of the same approach and theoretical framework could be fruitfully extended to the study of other sleep disorders, such as hypersomnia. FUTURE DIRECTIONS We argued that literature on quantitative EEG analysis of sleep in PI from some intrinsic limitations that, although often ascribed to the constraints of standard assessment and clinical research methods, should be overcome in the near future. We suggest that future research would be largely more informative if (some of) the following indications/controls will be followed:

• Studies should be performed on completely drug-free PI, otherwise it is hard to discriminate the specific pharmacological contribution on rapid frequency EEG bands.

• Common methods should be used for data acquisition (e.g., filter settings, sampling rates, etc.). • Common definitions should be adopted for sleep onset, NREM and REM cycles, etc. • Common methods should be used for data processing (e.g., PSA or FFT, window sizes, tapers, pre-processing routines, frequency resolution, etc.). • Common methods should be used for controlling for individual variability (e.g., log transforms vs. idiographic relative power vs. relative power based on normal control’s values, etc.). d. Introduction of a New Research Paradigm • Last, but not least, we suggest that the study local (topographic) changes during pathological sleep, based on the analysis of the EEG recorded from multiple scalp derivations, could provide new insights on its


underlying mechanisms. This should ultimately contribute to a commendable integration of knowledge and methods between basic and clinical research.

Marzano et al. [15]

[16]

ABBREVIATIONS

[17]

SWA

=

Slow-Wave Activity

[18]

SWS

=

Slow-Wave Sleep

[19]

NREM

=

Non-rapid eye movement

PSG

=

Polysomnography

CNS

=

Central Nervous System

FDG

=

Fluorodeoxyglucose

PET

=

Positron emission tomography

CBT

=

Cognitive behavior therapy

CBT-I

=

Cognitive behavioral therapy for insomnia

BZ

=

Benzodiazepines

SO

=

Sleep onset

FFT

=

Fast Fourier Transform

PSA

=

Power spectral analysis

EOG

=

Electrooculogram

EMG

=

Electromyogram

PI

=

Patients with insomnia

[20]

[21] [22] [23] [24]

[25]

[26]

[27]

REFERENCES [1] [2] [3]

[4] [5] [6]

[7] [8] [9]

[10] [11]

[12] [13] [14]

Spielman A, Caruso L, Glovinsky P. A behavioral perspective on insomnia treatment. Psychiatr Clin North Am 1987; 10: 541-53. Perlis ML, Giles DE, Mendelson WB, Bootzin RR, Wyatt JK. Psychophysiological insomnia: the behavioural model and a neurocognitive perspective. J Sleep Res 1997; 6: 179-88. Borbély AA, Baumann F, Brandeis D, Strauch I, Lehmann D. Sleep deprivation: effect of sleep stages and EEG power density in man. Electroencephalogr Clin Neurophysiol 1981; 51: 483-93. Borbély AA. A two-process model of sleep regulation. Hum Neurobiol 1982; 1: 195-204. Dijk DJ, Brunner DP, Borbely AA. Time course of EEG power density during long sleep in humans. Am J Physiol 1990; 258: 65061. Dijk DJ, Beersma DGM. Effects of SWS deprivation on subsequent EEG power density and spontaneous sleep duration. Electroencephalogr Clin Neurophysiol 1989; 72: 312-20. Merica H, Gaillard JM. The EEG of the sleep onset period in insomnia: a discriminant analysis. Physiol Behav 1992; 52: 199-204. Perlis ML, Smith MT, Andrews PJ, Orff H, Giles DE. Beta/Gamma EEG activity in patients with primary and secondary insomnia and good sleeper controls. Sleep 2001; 24: 110-7. Besset A, Villemin E, Tafti M, Billiard M. Homeostatic process and sleep spindles in patients with sleep – maintenance insomnia: effect of partial (21 h) sleep deprivation. Electroencephalogr Clin Neurophysiol 1998; 107: 122-32. Pigeon WR, Perlis ML. Sleep homeostasis in primary insomnia. Sleep Med Rev 2006; 10: 247-54. De Gennaro L, Ferrara M, Vecchio F, Curcio G, Bertini M. An electroencephalographic fingerprint of human sleep. Neuroimage 2005; 26: 114-22. De Gennaro L, Marzano C, Fratello F, Moroni F, Pellicciari MC, Ferlazzo F, et al. The EEG fingerprint of sleep is genetically determined: A twin study. Ann Neurol 2008; 64: 455-60. Dauvilliers Y, Maret S, Tafti M. Genetics of normal and pathological sleep in humans. Sleep Med Rev 2005; 9: 91-100. Agnew HW, Webb WB, Williams RL. The first effect night effect: an EEG study of sleep. Psychophysiology 1966; 2: 263-6.

[28] [29]

[30] [31] [32] [33]

[34]

[35]

[36] [37] [38]

[39] [40]

Mendels J, Hawkins DR. Sleep laboratory adaptation in normal subjects and depressed patients (‘first night effect’). Electroencephalogr Clin Neurophysiol 1967; 22: 556-8. Schmidt HS, Huque MF, Dubey SD. The differential laboratory adaptation of sleep parameters. Biol Psychiatry 1997; 3: 33-45. Monroe L. Psychological and physiological differences between good and poor sleepers. J Abnorm Psychol 1967; 72: 255-64. Freedman RR, Sattler HL. Physiological and psychological factors in sleep-onset insomnia. J Abnorm Psychol 1982; 91: 380-9. Lichstein KI, Rosenthal TI. Insomniacs’ perceptions of cognitive versus somatic determinants of sleep disturbance. J Abnorm Psychol 1980; 89: 105-7. Mitchell KR. Behavioral treatment of presleep tension and intrusive cognitions in patients with severe predormital insomnia. J Behav Med 1997; 2: 57-69. Lutzenberger W, Pulvermuller F, Birbaumer N. Words and pseudowords elicit distinct patterns of 30-Hz EEG responses in humans. Neurosci Lett 1994; 176: 115-8. Makeig S, Jung TP. Tonic, phasic, and transient correlates of auditory awareness in drowsiness. Cogn Brain Res 1996; 4: 15-26. Freedman R. EEG power in sleep onset insomnia. Electroencephalogr Clin Neurophysiol 1986; 63: 408-13. Espie CA, Broomfield NM, MacMahon KM, Macphee LM, Taylor LM. The attention-intention-effort pathway in the development of psychophysiologic insomnia: a theoretical review. Sleep Med Rev 2006; 10: 215-45. De Gennaro L, Ferrara M, Bertini M. The boundary between wakefulness and sleep: Quantitative electroencephalographic changes during the sleep onset period. Neuroscience 2001; 107: 1-11. Johnson LC, Spinweber CL, Seidel WF, Dement WC. Sleep spindle and delta changes during chronic use of a short-acting and longacting benzodiazepine hypnotic. Electroencephalogr Clin Neurophysiol 1983; 55: 662-67. Aeschbach D, Dijk DJ, Trachsel L, Brunner DP, Borbély AA. Dynamics of slow-wave activity and spindle frequency activity in the human sleep EEG: effect of midazolam and zopiclone. Neuropsychopharmacology 1994; 11: 237-44. Lamarche CH, Ogilvie RD. Electrophysiological changes during the sleep onset period of psychophysiological insomniacs, psychiatric insomniacs, and normal sleepers. Sleep 1997; 20: 724-33. Curcio G, Ferrara M, Piergianni A, Fratello F, De Gennaro L. Paradoxes of the first-night effect: a quantitative analysis of anteroposterior EEG topography. Clin Neurophysiol 2004; 115: 1178-88. Borbély AA. From slow waves to sleep homeostasis: new perspectives. Arch Ital Biol 2001; 139: 53-61. Nofzinger EA, Buysse DJ, Germain A, Price JC, Miewald JM, Kupfer DJ. Functional neuroimaging evidence for hyperarousal in insomnia. Am J Psychiatry 2004; 161: 2126-9. Merica H, Blois R, Gaillard JM. Spectral characteristics of sleep EEG in chronic insomnia. Eur J Neurosci 1998; 10: 1826-34. Perlis ML, Kehr EL, Smith MT, Andrews PJ, Orff H, Giles DE. Temporal and stagewise distribution of high frequency EEG activity in patients with primary and secondary insomnia and in good sleeper controls. J Sleep Res 2001; 10: 93-104. Jacobs GD, Benson H, Friedman R. Home-based central nervous system assessment of a multifactor behavioral intervention for chronic sleep-onset insomnia. Behav Ther 1993; 24: 159-74. Cervena K, Dauvilliers Y, Espa F, Touchon J, Matousek M, Billiard M, et al. Effect of cognitive behavioural therapy for insomnia on sleep architecture and sleep EEG power spectra in psychophysiological insomnia. J Sleep Res 2004; 13: 385-93. De Gennaro L, Ferrara M. Sleep spindles: An overview. Sleep Med Rev 2003; 7: 423-40. Aeschbach D, Cajochen C, Landolt HP, Borbély AA. Homeostatic sleep regulation in habitual short sleepers and long sleepers. Am J Physiol 1996; 270: 41-53. Werth E, Dijk DJ, Achermann P, Borbely AA. Dinamics of the sleep EEG after an early evening nap: experimental data and stimulations. Am J Neurophysiol 1996; 271: 501-10. Brunner DP, Dijk DJ, Borbély AA. Repeated partial sleep deprivation progressively changes in EEG during sleep and wakefulness. Sleep 1993; 16: 100-13. Morin CM, Daley M, Ouellet MC. Insomnia in adults. Neurology 2001; 3: 9-18.

Quantitative Electroencephalogram (EEG) in Insomnia [41]

[42] [43] [44] [45]

[46] [47]

[48] [49] [50]

Kramer M. Hypnotic medication in the treatment of chronic insomnia: non nocere! Doesn’t anyone care? Sleep Med Rev 2000; 4: 529-41. Bastien CH, LeBlanc M, Carrier J, Morin CM. Sleep EEG power spectra, insomnia, and chronic use of benzodiazepines. Sleep 2003; 26: 313-7. Bonnet MH, Arand DL. Physiological activation in patients with Sleep State Misperception. Psychosom Med 1997; 59: 533-40. Bonnet MH, Arand DL. 24-Hour metabolic rate in insomniacs and matched normal sleepers. Sleep 1995; 18: 581-8. Krystal AD, Edinger JD, Wohlgemuth WK, Marsh GR. NREM sleep EEG frequency spectral correlates of sleep complaints in primary insomnia subtypes. Sleep 2002; 25: 630-40. Reite M, Buysse D, Reynolds C, Mendelson W. The use of polysomnography in the evaluation of insomnia. Sleep 1995; 18: 58-70. Devoto A, Manganelli S, Lucidi F, Lombardo C, Russo PM, Violani C. Quality of sleep and P300 amplitude in primary insomnia: a preliminary study. Sleep 2005; 28: 859-63. Borbély AA, Achermann P. Concepts and models of sleep regulation: an overview. J Sleep Res 1992; 1: 63-79. Tobler I, Franken P, Trachsel L, Borbély AA. Models of sleep regulation in mammals. J Sleep Res 1992; 1: 125-7. Oleksenko AI, Mukhameto LM, Polyakova IG, Supin AY, Kovalzon VM. Unihemispheric sleep deprivation in bottlenose dolphins. J Sleep Res 1992; 1: 40-4.

Current Pharmaceutical Design, 2008, Vol. 14, No. 32 [51] [52]

[53]

[54]

[55] [56]

[57]

3455

Huber R, Deboer T, Tobler I. Topography of EEG dynamics after sleep deprivation in mice. J Neurophysiol 2000; 84: 1888-93. Ferrara M, De Gennaro L, Curcio G, Cristiani R, Corvasce C, Bertini M. Regional differences of the human sleep electroencephalogram in response to selective slow-wave sleep deprivation. Cereb Cortex 2002; 12: 737-48. Huber R, Esser SK, Ferrarelli F, Massimini M, Peterson MJ, Tononi G. TMS-induced cortical potentiation during wakefulness locally increases slow wave activity during sleep. PLoS ONE 2007; 2 (3): e276 doi: 10.1371/journal.pone.0000276. De Gennaro L, Fratello F, Marzano C, Moroni F, Curcio G, Tempesta D, et al. Cortical plasticity induced by transcranial magnetic stimulation (TMS) during wakefulness affects EEG activity during sleep. PLoS ONE 2008; 3(6): e2483 doi: 10.1371/journal.pone. 0002483. De Gennaro L, Ferrara M, Curcio G, Cristiani R. Antero-posterior EEG changes during the wakefulness-sleep transiton. Clin Neurophysiol 2001; 112: 1901-11. De Gennaro L, Vecchio F, Ferrara M, Curcio G, Rossini PM, Babiloni C. Changes in fronto-posterior functional coupling during sleep onset in humans. J Sleep Res 2004; 13: 209-18. De Gennaro L, Vecchio F, Ferrara M, Curcio G, Rossini PM, Babiloni C. Antero-posterior functional coupling at sleep onset: changes as a function of increased sleep pressure. Brain Res Bull 2005; 65: 133-40.