Description

Cyclic alternating pattern and the gating mechanisms of sleep. Despite their EEG differences, slow EEG events (K-complexes and delta bursts) and AASM arousals (fast rhythms) share functional properties and, therefore, may be included in the comprehensive term “activating complexes.”

Such a variety of EEG manifestations relies on specific gates controlling the flow of internal and external inputs. The thalamic-basal forebrain gate is an ultimate step of resistance against arousing impulses. Initially the cortex tries to preserve sleep continuity with reinforcement of its gates that are indicated by the occurrence of K-complexes and delta bursts in sleep EEG. However, when the thalamic gate cannot control the afferent inputs, a cortical change is seen, translated by an alpha mixed or an alpha/beta frequency burst (32). The initial reaction of the cerebral cortex is a sleep-protective response as the majority of transient rapid activities are preceded by a slow high-amplitude EEG burst (29). Slow EEG events during non-REM sleep are an emergent property of cortico-thalamo-cortical networks. In particular, they originate from the dynamic interplay of three cardinal oscillators: the synaptically based cortical oscillator and two thalamic oscillators, ie, thalamocortical neurons and nucleus reticularis thalami. The functional implications of this dialogue provide permissive windows for cellular excitability and network plasticity during slow-wave sleep (12).

Cyclic alternating pattern sequences reflect the balance between sleep- and wake-promoting systems. Subtype A1 is a natural “delta injection” fueling the build-up and consolidation of deep NREM stages and defending sleep against perturbations. In contrast, cyclic alternating pattern subtypes A2 and A3 drive the sleeper toward more superficial vigilance states. Therefore, cyclic alternating pattern sequences represent a protective, short-term homeostatic mechanism of non-REM sleep in which the amount of slow-wave activity is buffered and sleep continuity preserved (30). When delta power increases in sleep after sleep deprivation, the cyclic alternating pattern system reacts with a robust decrease of cyclic alternating pattern rate (14). Supplementation of sleep with slow waves after deprivation counteracts the production of cyclic alternating pattern sequences, probably because the sleep promoting system is under saturation.

Cyclic alternating pattern and oscillation below 1 Hz. Besides cyclic alternating pattern, the other major EEG activity in the frequency range below 1 Hz is the so-called slow oscillation (02). This 0.5 to 0.9 Hz EEG rhythm, which characterizes states of reduced tonic arousal, was outlined during anesthesia and NREM sleep in both animals (cats and rats) and human subjects. The slow oscillation is generated in cortical neurons and consists of phases of depolarization characterized by intensive neural firing, followed by long-lasting hyperpolarization. Hence, the two phases of the slow oscillation are characterized by opposite neural phenomena: cortical excitation made up of synaptic potentials and cortical inhibition mainly due to disfacilitation in the network. The excitatory component of the slow oscillation is effective in grouping the K-complexes and delta waves, which do not occur in isolation, but are grouped into complex wave sequences. The coalescence of slow rhythms is especially visible during NREM sleep. In particular, high-voltage slow waves rarely appear as isolated features during slow-wave sleep, but in most cases they converge into collectives resulting in the phase A1 subtypes of cyclic alternating pattern. As NREM sleep progresses from stage 1 to stage 4, the differences in morphology and voltage between phase A (clusters of K-complexes and delta bursts) and the successive phase B (sleep stage background) become gradually less evident until the EEG is dominated by the uniform pattern of non-cyclic alternating pattern with the high amplitude slow waves recurring in the frequency range of oscillation below 1 Hz. Neurophysiological investigation has ascertained that the slow cortical oscillation (< 1 Hz) is absent at sleep onset but begins to organize in small territories, thereafter recruiting larger ones through coupling mechanisms as sleep deepens. Consolidated slow-wave sleep is characterized by a sustained oscillation below 1 Hz, which reflects a stable non-cyclic alternating pattern condition (85; 59).

Cyclic alternating pattern and the structure of sleep. Sleep architecture is based on the cyclic alternation of two major neurophysiological states: NREM and REM sleep. NREM sleep is composed of 3 or 4 stages in which EEG synchrony grows with the increasing depth of sleep. In contrast, EEG desynchrony is the dominant feature of REM sleep. The alternation of NREM and REM sleep constitutes the sleep cycle, and its recurrence during the night determines the classical stepwise profile of sleep macrostructure. The NREM portion of the sleep cycle starts with a slow descending branch sloping from the more superficial to the deeper NREM stages; continues with a central trough that represents the deepest stages of the sleep cycle; and ends with a rapid reverse ascending branch, expressed by the more superficial NREM stages that precede REM sleep. Accordingly, the NREM sleep architecture delineates a continuous pattern of build-up (descending branch), maintenance (trough), and resolution (ascending branch) of EEG synchrony. A detailed investigation has ascertained that the spontaneous EEG fluctuations centered on the 20 to 40 seconds periodicity of cyclic alternating pattern are involved in the subtle mechanisms that regulate the production and attenuation of slow-wave activities during sleep. There is evidence that the different components of cyclic alternating pattern have a sculpturing effect on the profile of the sleep cycle. Comparing spectral assessment and EEG visual scoring of NREM sleep in normal healthy subjects, the amount of slow rhythmic oscillations (spectral analysis) parallels the number of cyclic alternating pattern cycles (visual detection), with a striking agreement between spectral power gatherings and visually scored A phases (92). The regular EEG oscillations that accompany the transition from light sleep to deep stable sleep are basically expressed by the A1 subtypes (87). Within the sleep cycle, 90% of the A phases detected in the descending branches and 92% of the A phases detected in the troughs are subtype A1, whereas 64% of the A phases identified in the ascending branches are subtypes A2 (45%) or A3 (19%). These findings indicate that both slow and rapid EEG activating complexes are involved in the structural organization of sleep. The abundance of A1 subtypes in the descending branches and troughs can be the EEG expression of the cerebral mechanisms involved in REM-off activity, whereas the predominance of subtypes A2 and A3 (and arousals) in the ascending branches reflects the REM-on drive (92). As a confirmation, the A1 index and percentage positively correlate with slow wave sleep percentage and, overall, this cyclic alternating pattern subtype occurs more frequently during stage N3 of non-REM sleep (50). Therefore, besides their manifold EEG features, activating complexes are also characterized by a non-random distribution across the night, which assumes a clear-cut periodicity during NREM sleep within the framework of cyclic alternating pattern. For this reason, cyclic alternating pattern is considered the main expression of sleep microstructure (19).

The measures of cyclic alternating pattern: ontogenetic aspects. EEG features are highly sensitive markers of brain development. Accordingly, sleep reflects the physiological changes that accompany the different ages of the lifespan. In particular, slow-wave sleep dominates in the younger decades in contrast to superficial NREM sleep, which increases with the aging process. Cyclic alternating pattern parameters, as all the other measures of sleep microstructure such as K-complexes and sleep spindles, undergo dynamic changes across the maturational phases of life (52). The age-related values of cyclic alternating pattern have been measured and defined in order to establish the physiological ranges of normal sleep (57). Cyclic alternating pattern rate follows a U-shape dynamic course with aging; this sleep parameter is higher during the first years after birth (mainly during the first 2 years of life), subsequentially drops in middle age, and finally reaches a steep increase during advanced adulthood and senescence (57; 07; 08; 39; 49). Cyclic alternating pattern subtypes A1 gradually decrease with aging, whereas subtypes A2 and A3 progressively increase, contributing to the paraphysiological sleep lightening that occurs with aging. These data have been synthesized in a systematic review with meta-analysis (50).

Clinical applications

Clinical applications of cyclic alternating pattern

Cyclic alternating pattern and sleep disorders in adults. The cyclic oscillations of cyclic alternating pattern are physiological components of sleep structure in which they act as dynamic segments, pacing the state transition between different NREM sleep stages according to homeostatic and ultradian processes. However, cyclic alternating pattern sequences can occur also in response to external stimuli of different sensorial modality (tactile, thermal, acoustic, painful, etc.). Accordingly, the amount of cyclic alternating pattern increases when sleep is achieved under conditions of noise stimulation or in situations of sleep disruption, such as insomnia, depression, eating disorders, sleep apnea syndrome, periodic limb movements, nocturnal frontal lobe epilepsy, primary generalized and focal lesional epilepsy, and Prader-Willi syndrome in adults (86; 93; 88; 82; 83; 70; 55; 65; 58; 103; 18; 15; 72; 27). It is lowered by sleep-promoting conditions such as narcolepsy, multiple system atrophy, drug administration, CPAP treatment in obstructive sleep apnea, and nighttime recovery sleep after prolonged sleep deprivation (83; 69; 70; 56; 65; 71; 14; 23; 95; 99; 79; 16).

Cyclic alternating pattern is not only influenced by sleep disorders, but, in turn, modulates the occurrence and distribution of sleep-related events. In particular, phase A of cyclic alternating pattern triggers and paces the allocation of bruxism, sleepwalking, epileptic events, periodic limb jerks, and rhythmic movements during NREM sleep (102; 55; 60; 40; 68; 42; 26; 53; 36). In contrast, phase B of cyclic alternating pattern is closely related to the repetitive respiratory events of sleep-disordered breathing, and only the powerful autonomic activation during the following cyclic alternating pattern A phase can restore post-apnea breathing (88). Like an alternatively opening (phase A) and closing (phase B) gate, cyclic alternating pattern phases act as periodic permissive windows in NREM sleep, offering a favorable background for phasic and/or repetitive sleep-related manifestations, ie, periodic limb movements, epileptic seizures, sleep apneas.

Accordingly, a number of sleep disorders can be classified pathophysiologically on the basis of their relationship with cyclic alternating pattern and non-cyclic alternating pattern. In particular, periodic limb movement disorder, sleep bruxism, and epileptic manifestations can be considered as phase A-related disorders, whereas sleep apneas are a typical expression of a phase B-related disturbance (84).

The role of cyclic alternating pattern in primary insomnia. Primary insomnia appears to be the exaggeration of a physiological rhythm ordinarily involved in the sleep process. Previous studies have ascertained that acoustic stimuli enhance the physiological amount of cyclic alternating pattern rate and determine poor sleep and daytime dysfunction even without an increase of sleep fragmentation. In this perspective, cyclic alternating pattern operates as a double-edged sword. Although limited quantities of cyclic alternating pattern mediate physiological effects, larger amounts reflect the brain’s difficulties to consolidate and preserve sleep and, therefore, may be associated with detrimental consequences. Because primary insomnia is not supported by any other sleep, medical, psychiatric, or substance-induced problem, there is no evidence of an organ disorder. In any case, whatever the nature of disturbance, the outcome is the amplification of an otherwise physiological process.

Polysomnography investigation based on an extensive sample of Caucasian patients affected by primary insomnia have demonstrated that cyclic alternating pattern parameters consistently reflect the reduced quality of sleep in insomnia complainers and can substantiate the efficacy of hypnotic medication (62). Discriminant analysis indicates that cyclic alternating pattern rate is the most sensitive sleep measure of effective drug treatment, whereas correlation analysis shows that cyclic alternating pattern rate is the polysomnography parameter that better reflects subjective sleep quality (94). Similar findings have been confirmed in non-Caucasian subjects. In Japanese patients with psychophysiological insomnia, a randomized crossover comparative study with placebo showed that hypnotic treatment (zolpidem) increased sleep stability by significantly improving the overnight cyclic alternating pattern rate as well as subjective sleep quality (54).

Cyclic alternating pattern parameters are also useful tools to monitor the effects of intermittent hypnotic treatment. In a double-blind study carried out on adults with primary sleep maintenance insomnia lasting longer than 1 month, polysomnography measures and perception of sleep quality were assessed at baseline and during the following six consecutive nights of alternating treatment with zolpidem (10 mg) or placebo. Compared to baseline values, cyclic alternating pattern rate, cyclic alternating pattern time, subtype A2, and subtype A1 were significantly reduced with zolpidem treatment and correlated with sleep quality, whereas they did not rebound above baseline with placebo (66).

An intriguing aspect of insomnia is the established finding that people with this disorder often overestimate the time they take to get to sleep and underestimate the total amount of time they actually sleep. To investigate this mismatching phenomenon, a polysomnography study was carried out in 20 patients with a diagnosis of sleep state misperception or paradoxical insomnia (64). Recruitment of misperceptors without coexisting neurologic, medical, or psychiatric disorders was based on objective total sleep time (TST) of at least 6.5 h, objective sleep latency shorter than 30 minutes, underestimated difference between objective and subjective TST of at least 120 minutes, and subjective estimation of sleep latency more than 20% of objective sleep latency. Polysomnography data of misperceptors were compared with those of 20 normal gender- and age-matched subjects (controls). Patients and controls presented nonsignificant differences in the amounts of objective sleep time (464 vs. 447 minutes) and objective sleep latency (9 vs. 8 minutes). However, compared to controls, misperceptors reported a significantly shorter time of perceived sleep (285 vs. 461 minutes; p < 0.0001), and a significantly longer duration of perceived sleep latency (51 vs. 22 minutes; p < 0.0001). Arousal index (32/hour vs. 19/hour; p < 0.0001) and total cyclic alternating pattern rate (58% vs. 35%; p < 0.0001) were significantly higher in insomniacs. In the sleep period between objective and subjective sleep onset, cyclic alternating pattern rate was 64.4% in misperceptors and 45.1% in controls (p < 0.002). Insomniacs showed significantly higher amounts of cyclic alternating pattern rate in stage 1 (62.7% vs. 37.5%; p < 0.0001) and in stage 2 (53.3% vs. 33.2%; p < 0.0001), but not in slow-wave sleep. The percentage of subtypes A2, which include both sleep promoting and wake-promoting EEG features, was significantly higher (p < 0.001) in misperceptors (31%) compared to controls (24%). Interestingly, misperceptors reported a limited amount of subjective awakenings (mean: 4) in contrast to objective findings (mean: 11). The mismatch could be explained in part by the high amounts of cyclic alternating pattern between successive awakenings, which were merged together in a single experience. In other words, if sleep between two successive awakenings is superficial (expressed by sleep stages 1 and 2), unstable (as reflected by increased amounts of cyclic alternating pattern), and fragmented (increased arousal index), time separating the two events is perceived as continuous wake. These findings suggest that difficulty to maintain consolidated sleep is interpreted as wakefulness in misperceptors.

Sleep is a dynamic process with a self-regulating character. The nightly recurring sleep process is organized into consecutive cycles in which the sequence of NREM stages and the alternation between NREM and REM sleep show a quite stable tendency and a largely predictable pattern. These constraints produce the macrostructural development of sleep. However, transient EEG changes can interact with the expected development of sleep and ensure adaptation to both internal and external conditions. Cyclic alternating pattern and arousals represent rapid adaptive adjustments of vigilance during sleep. Failure of these compensatory processes conducts to non-restorative sleep. Therefore, assessment of sleep quality relies on a variety of polysomnography measures, including sleep duration (quantified by total sleep time and sleep efficiency), sleep intensity (reflected by stages 3 and 4), sleep continuity (altered by nocturnal awakenings and arousals), and sleep stability (impaired by excessive amounts of cyclic alternating pattern). These polysomnography measures are susceptible to deterioration in varying ways and proportions in accordance to the manifold clinical manifestations of insomnia. In the hierarchy of polysomnography measures, cyclic alternating pattern variables appear to be the most sensitive to any source of internal or external perturbation during sleep. Regardless of the specific characteristics of sleep alteration, insomnia ceases to be an indefinite mental disorder, but emerges as a subjective disturbance supported by measurable neurophysiological changes (62).

The importance of sleep instability. Objection could be raised on the usefulness of cyclic alternating pattern in routine clinical practice. Besides the consideration that the authors apply cyclic alternating pattern in their daily activities, attention should be focused on the philosophy that lies behind the cyclic alternating pattern scoring methodology. The main assumption of cyclic alternating pattern is that we must consider that during certain periods of the night the arousal level is unstable. The concept of instability is a basic issue of all complex systems and supports the dynamics of biological variability. Within certain ranges, instability warrants flexible and adaptive features to the complex system. In normal sleep, cyclic alternating pattern accompanies the stage transitions maintaining in-phase both the EEG and autonomic functions through regular fluctuations. This means that what we observe on the peripheral sensors (respiratory events, heart rate variations, blood pressure shifts, myoclonic jerks) has a consistent and synchronous expression on the EEG and vice versa. In other words, cyclic alternating pattern allows us to read sleep as a musical score where all the instruments play a coordinated harmony. This means that even without the EEG leads, the detection of an unstable cardiorespiratory pattern is certainly associated with an oscillatory behavior of the cerebral activities. To guarantee survival during prolonged unconsciousness (ie, sleep), a strong interaction among all the biological subsystems is mandatory.

Accordingly, life-threatening events, eg, repetitive apneas, enhance the brain’s necessity to increase the number of protective arousals determining a metronomic setting of oscillations expressed by prolonged cyclic alternating pattern sequences. Cyclic alternating pattern scoring incorporates all the conventional EEG arousals in NREM sleep and provides additional information supplied by the frontal EEG arousals (subtype A1), which are omitted by the official classification. All this remains unexplored using only the stepwise description of sleep stages. In particular, the new AASM rules have oversimplified the sleep process, merging stages 3 and 4 into stage N3 and overemphasizing the role of single arousals in stage scoring; the new AAMS rules are simply tools offered to accelerate scoring procedures in routine practice. However, it is a common experience that a number of patients remain inadequately diagnosed and treated in spite of an apparently normal sleep structure. A study carried out in patients with restless legs syndrome and periodic limb movements during sleep demonstrates that dopamine-agonist medication reduces the amount of leg jerks but maintains high levels of cyclic alternating pattern with persistence of non-restorative sleep (22). Similar controversies happen daily in sleep labs that ascertain maintenance of excessive sleepiness in patients with obstructive sleep apnea syndrome even when CPAP restores normal apnea-hypopnea index measures. An exploratory approach using the cyclic alternating pattern framework would probably shed light on these controversial cases. It is known that CPAP titration detached from the concomitant assessment of cyclic alternating pattern can jeopardize the effectiveness of ventilatory treatment (96). The persistence of cyclic alternating pattern and arousals even when respiratory events are controlled by autoadjusting equipment indicates an inadequate titration procedure.

Finally, the sensitivity of cyclic alternating pattern to medication in insomniac patients could be exploited in the evaluation of new sleep-promoting drugs. So far, hypnotic compounds have based their effectiveness on the paradigm of sleep-onset promotion, sleep-time enhancement, and wake after sleep onset (WASO) reduction. All the available hypnotic agents, especially those belonging to the gamma aminobutyric acid (GABA) family, share these features, which paradoxically offer a flat undifferentiated therapeutic scenario. The new sleep drugs in the pipeline, acting on alternative targets (melatonin, orexin, serotonin), need a new cultural framework that cannot be limited to sleep latency and sleep duration, but also needs to incorporate the issue of sleep and autonomic stability.

So far, these objectives have been neglected by most clinical trials for drug approval. Comparing gaboxadol and zolpidem in a model of situational insomnia, cyclic alternating pattern parameters showed a significant independence from other electrophysiological measures (including spectral analysis), disclosed a strong association with subjective sleep quality, and allowed discrimination between the administered drugs. Zolpidem has also been used in a non-insomnia setting. In one trial, most patients with idiopathic central sleep apnea experienced a decrease in central apnea/hypopneas with zolpidem. They also had improved sleep continuity, reduced total number of arousals, and decreased subjective daytime sleepiness, without worsening of oxygenation or obstructive events. In obstructive sleep apnea, trazodone increased respiratory effort–related arousal threshold in response to hypercapnia and allowed patients to tolerate a higher CO2 level. According to Younes, “the percentage of time spent in instability is related to how far removed from stability the system is, relative to how much changes in the stability factors can occur spontaneously during the night. Some of the factors that determine stability change from time to time during sleep. Chief among these is arousal threshold” (101). A study established that trazodone increases the respiratory arousal threshold in obstructive sleep apnea patients with low arousal threshold indicating the potential inclusion of non-GABAergic agents in the management of sleep-related breathing disorders (17).

Cyclic alternating pattern and obstructive sleep apnea. Obstructive sleep apnea can be considered as prototypical example of a cyclic alternating pattern phase-B disorder. In obstructive sleep apnea patients, the “deactivation” role of cyclic alternating pattern phase B creates the permissive framework for a transient (partial of complete) airflow interruption, whereas the more “interventionist” phase A guarantees respiratory restoration.

This condition exerts a profound effect on sleep stability, which can be measured at both macro- and microstructural sleep levels. Apneic patients frequently present decreased amounts of the deepest sleep stages, in a framework of a deeply fragmented sleep, as expressed by higher cyclic alternating pattern rate. However, the impact of the disorder on sleep profile strongly varies depending on obstructive sleep apnea severity; in fact, patients with milder obstructive sleep apnea present higher values of cyclic alternating pattern subtypes A1, whereas in moderate-to-severe obstructive sleep apnea patients, cyclic alternating pattern subtypes A2 and A3 prevail. In other words, if in milder obstructive sleep apnea cases sleep is still trying to reinforce its continuity with the cyclic alternating pattern subtypes A1 appearance, this balance is definitely lost in the more severe obstructive sleep apnea phenotypes, where the more disturbing subtypes A2 and A3 disrupt sleep continuity (25).

Sleep microstructure can also support a deeper understanding of mechanisms beneath excessive daytime sleepiness (EDS) in obstructive sleep apnea patients. It is well known that excessive daytime sleepiness does not necessarily correlate with apnea-hypopnea index (AHI) nor with disease severity. It has been demonstrated that obstructive sleep apnea patients with excessive daytime sleepiness presented higher level of cyclic alternating pattern oscillations, which is indicative of highly fragmented sleep, compared to those with a more refreshing sleep (34). This observation is not trivial, as it has been proved that excessively sleepy patients with obstructive sleep apnea are at higher risk for incident cardiovascular disorders. They frequently have higher autonomic imbalance and lower baroreflex sensitivity and abnormal nocturnal hearth rate variability (38). As cyclic alternating pattern oscillations have a strong bidirectional link with autonomic dynamics, the role of sleep microstructure towards patients’ health takes on a whole new meaning (24).

Finally, the cyclic alternating pattern “language” can confirm the correctness of obstructive sleep apnea therapy, as the utilization of continuous positive airway pressure (CPAP) treatment significantly modifies the sleep microstructure, with a decrease of cyclic alternating pattern subtypes A2 and A3 and a parallel increase of noncyclic alternating pattern sleep and subtype A1 during sleep stage N3 (10).

Cyclic alternating pattern, sleep-related epilepsy, and non-REM sleep parasomnias. Non-REM sleep parasomnias and sleep-related hypermotor (SHE) epilepsy share numerous semiological similarities, and their proper identification can be challenging for clinicians.

Although various aspects in the motor manifestation of these conditions can support the sleep-experts work-up, some intriguing information can be collected by the sleep texture as well (37). Both parasomnias and sleep-related hypermotor share an increase of cyclic alternating pattern oscillations, with a preponderance of cyclic alternating pattern subtypes A1, in a framework of increased slow-wave sleep instability (51). Notably, both paroxysmal and parasomniac events are frequently heralded by a cyclic alternating pattern sequence and commonly appear time-locked with cyclic alternating pattern phase A, confirming the permissive nature of these strongly synchronized events for the occurrence of (hypermotor) nocturnal events. The semiological commonalities between the two pathologies suggest the involvement of similar cortico-subcortical circuits, which find in the cyclic alternating pattern synthesis the common denominator of an arousal-related motor disinhibition, whether or not epileptic in origin.

These aspects collected from sleep microstructure highlight some pathophysiological clues on non-REM sleep parasomnias and sleep-related hypermotor, such as the derailment of the arousal system in NREM sleep.

Details on the cyclic alternating pattern variations in patients affected by epilepsy have been synthesized in a systematic review with meta-analysis (100).

Cyclic alternating pattern and neurodegeneration. The cyclic alternating pattern metric can be used to predict the risk of conversion to alpha-synucleinopathies in patients affected by REM-sleep behavior disorder (RBD), with lower values of cyclic alternating pattern rate and index being associated with stronger risk for early neurodegeneration (47). This observation reveals that REM-sleep behavior disorder cannot be considered as an exclusively “REM sleep-related” sleep disorder, as it has some deep consequences in the non-REM sleep features as well, confirming once again that sleep must be evaluated in its overall complexity, which is based on the reciprocal influence between REM and non-REM sleep stages.

The drop of cyclic alternating pattern rate below normal values had been observed in other neurodegenerative disorders such as Alzheimer disease and mild amyotrophic lateral sclerosis, revealing that the impoverishment of sleep oscillations can provide important information on brain health (41; 09; 11).

The association between non-REM sleep fragmentation, arousal-related disorders (including both REM and non-REM sleep parasomnias), and neurodegeneration have been suggested as markers of decline in Parkinson disease (80).

Brain criticality, avalanches, and cyclic alternating pattern. According to the neural criticality hypothesis the brain, as other complex systems, operates close to a phase transition to optimize information processing (98). Criticality is a property maximizing systems’ computational capabilities, enabling their flexibility and ameliorating their functional performances (77). In this framework, sleep plays a key role in tuning the brain towards criticality with scale-free cascades of spatiotemporal activity called “avalanches” following its inner physiological fluctuations (04). It has been demonstrated that the avalanche dynamic is largely modulated by the NREM/REM sleep cycles and that, during NREM sleep, there is a tight association between avalanches and CAP phase A (74). There is a strong correlation between avalanche occurrence and the CAP A1 subtype and weaker positive correlation with subtypes A2 and A3. Also, avalanche density partially overlaps with CAP subtype A1 distribution as it progressively decreases from the first to the last sleep cycle following the homeostatic decay of slow-wave sleep throughout the night (74). This strong association between brain criticality and CAP suggests some novel fascinating hypotheses to understand the neurobiological significance of sleep-related brain functions, including learning mechanisms and consciousness manifestation.

Automatic analysis of cyclic alternating pattern. There is consolidated evidence that cyclic alternating pattern parameters provide more detailed information and are significantly more sensitive than conventional sleep measures. However, the visual scoring of cyclic alternating pattern is time-consuming, and this can compromise an extensive utilization of the method. In other words, only the availability of an adequate system for the automatic detection of cyclic alternating pattern can really make it an easily exploitable tool. The problem of automatic cyclic alternating pattern recognition has been addressed in several studies, with interesting results (03; 20; 35). However, most proposed classification methods required the preliminary measurement of sleep macrostructure, making it necessary for the human scorer to visually classify the sleep profile. Moreover, they relied on the extraction of spectral parameters from the EEG signal to compute descriptors on time-windows and on the application of machine-learning algorithms for classifying each window as belonging to a phase A of cyclic alternating pattern or not. A preliminary study investigated a set of descriptors computed on 1-second-long windows with different techniques, and the information content of each descriptor was evaluated by means of ROC curves (44). Those descriptors were used in a subsequent work training four different classifiers for the classification of cyclic alternating pattern A phases, where the linear discriminant resulted to be the most accurate, with an average accuracy equal to 83% (45). However, due to the intrinsic characteristics of cyclic alternating pattern phases A, which consist of transient, non-stationary phenomena, it is likely that the signal will vary its characteristics over time, and it is not convenient to define a fixed window length. A dataset of 16 polysomnographic recordings from healthy subjects was employed, and the EEG traces first underwent an automatic isolation of NREM sleep portions by means of an Artificial Neural Network and then a segmentation process based on the Spectral Error Measure. The information content of the descriptors was evaluated by means of ROC curves and compared with that of descriptors obtained without the use of segmentation. Finally, the descriptors were used to train a discriminant function for the automatic classification of cyclic alternating pattern phases A. With respect to previous scoring methods, a significant improvement in terms of both information content carried by the descriptors and accuracy of the classification was obtained. Average cyclic alternating pattern rate obtained with automatic scoring was 44.83±12.41 vs. 43.83±10.13 measured with visual analysis. The correlation coefficient between the two series of cyclic alternating pattern rates was 0.73, and the P-value for the null hypothesis that the two populations had the same distribution was 0.62 (43). Although the sensitivity was still not very high (69.55±6.6), the specificity (90.49±2.8) and the accuracy (87.19±2.48) had a significant increase with respect to the previous study, indicating that EEG segmentation proves to be a useful step in the computation of descriptors for cyclic alternating pattern scoring (45). According to a systematic review, the median accuracy of the state-of-the-art methods for automatic CAP scoring is comparable to the upper limit of the specialist agreement range (ranging from 69% to 78%), implying that the automatic CAP analysis could be performed (48).

Conclusion. A certain amount of information remains uncovered if we limit our assessment to macrostructural events. As a marker of sleep instability, cyclic alternating pattern rate may be enhanced by a number of internal or external factors. Cyclic alternating pattern rate is a highly sensitive, poorly specific index as it cannot reveal the nature of perturbation. However, it certainly indicates that one or more factors are interfering with the processes of sleep consolidation and quantifies the magnitude of perturbation. Specificity of the disturbance factor is mainly supplied by the phase A subtype, which is differently modified by sleep disorders and medication. In contrast, the presence of non-cyclic alternating pattern is closely related to a global condition of stability when all the subsystems that control and influence sleep mechanisms have achieved a reciprocally balanced interaction.

From the cyclic alternating pattern/non-cyclic alternating pattern perspective, analysis of sleep microstructure is not limited to the finding of a single event (eg, an isolated arousal), but to the identification of a pattern (presence or absence of a cyclic alternating pattern sequence) that translates a physiological state involving cerebral activities, autonomic functions, and behavioral features. In other words, what happens upstairs (brain) is reflected at the lower levels (autonomic and muscle parameters) and vice versa. Once we attribute an activation relevance to all cyclic alternating pattern A phases (including K-complexes and delta bursts), we overcome the barren contrast between visible and non-visible arousals, which for years has frozen the possibility of understanding the flexible capacity of the brain to use different EEG features in different neurophysiological situations. Conventional EEG arousals are the peak of an iceberg composed of other more subtle, but equally powerful, activating events.