American Journal of Respiratory and Critical Care Medicine

Common polysomnographic measures of sleep-disordered breathing have shown a disappointing ability to predict important consequences such as excessive daytime sleepiness. Using novel analytic techniques, this study tested the hypothesis that numerous, brief disruptions in cortical activity could occur on a breath-to-breath basis during nonapneic sleep. Spectral analysis proved the existence of respiratory cycle-related electroencephalographic changes in each of 38 adult patients evaluated by polysomnography for sleep-disordered breathing. Furthermore, the tendency for sigma (13–15 Hz) electroencephalographic power to vary with the respiratory cycle predicted next-day sleepiness as measured by the multiple sleep latency test. The predictive value was enhanced when the analysis was limited to those 27 patients who had sleep-disordered breathing (more than 5 apneas or hypopneas per hour of sleep). In contrast, nocturnal rates of apneas and hypopneas, as well as minimal oxygen saturation, did not predict sleepiness as well. On average, sigma power increased notably during inspiration, whereas delta (1–4 Hz) power showed a simultaneous decrease. We conclude that electroencephalographic activity shows detectable changes during nonapneic respiratory cycles in adults evaluated for sleep-disordered breathing. Quantification of these changes, which may reflect numerous inspiratory microarousals, could prove useful in prediction of excessive daytime sleepiness.

The severity of sleep-disordered breathing (SDB) is often assessed by polysomnography. However, standard measures based on recorded rates of apneas and hypopneas or extent of oxygen desaturation frequently fail to explain important SDB outcomes, such as excessive daytime sleepiness (1). In a large clinical series, data from 1,146 patients who underwent nocturnal polysomnography followed by the multiple sleep latency test showed that the apnea–hypopnea index explained only 11% of the variance in daytime sleepiness (2). In the community-based Sleep Heart Health Study, snoring still showed a significant association with subjective sleepiness after the apnea–hypopnea index was taken into account (3). Results such as these raise the possibility that standard measures of apnea severity may not accurately reflect underlying processes that cause sleepiness in SDB (4).

Attempts to improve respiratory assessment during polysomnography have included innovative methods to monitor airflow (5), respiratory effort (6), and autonomic signs of intrapleural pressure swings (7). Efforts to improve assessment of brain activity during sleep have focused on arousals (8, 9), alternative EEG leads to detect them (10), autonomic signs of arousals (11), respiratory event-related arousals (6, 12), and EEG signal analysis during sleep (13). Other investigations have focused on changes in EEG spectral power before, during, and after apneic events (1416). However, most of these approaches remain inextricably linked to apneas or other discrete, visually identified respiratory events, the rates of which are known to be suboptimal correlates of health outcomes. None of the newer methods have been shown to improve prediction of sleepiness substantially, and none have replaced older, standard measures in clinical practice.

To address this problem, we developed a novel signal analysis algorithm that focuses specifically on nonapneic sleep (17). We hypothesized that increased respiratory effort known to occur in SDB during nonapneic respiratory cycles, which generally occupy the majority of the night, might arouse the cortex in a subtle but recurrent manner detectable only by computer. Signal analysis was used to demonstrate respiratory cycle-related EEG changes (RCREC) during sleep in eight of nine children scheduled to undergo adenotonsillectomy for clinically suspected SDB (18). The magnitude of RCREC predicted sleepiness before surgery, and the change in RCREC after surgery correlated with improvement in sleepiness. In contrast, the standard pediatric apnea–hypopnea index did neither. A subsequent preliminary report showed that among 34 additional adenotonsillectomy subjects, preoperative RCREC predicted postoperative improvement in both sleepiness and hyperactive behavior, whereas the apnea–hypopnea index and EEG arousal index predicted neither (19). Moreover, the predictive value of RCREC, but not that of the standard measures, was particularly strong among those subjects who had obstructive sleep apnea.

Despite these promising results in children, RCREC have not been investigated previously in adults. Here we describe studies, in a series of adult patients at risk for SDB, to determine whether RCREC exist, and if so whether they predict sleepiness.

Subjects and Measures

Subjects were consecutive patients studied at the Michael S. Aldrich Sleep Disorders Laboratory between June 1, 2000 and May 31, 2001, who met the following criteria: (1) they were at least 18 years old; (2) full-night polysomnography was performed, mainly to assess for clinically suspected SDB; (3) at least 6 hours of sleep was obtained; and (4) a multiple sleep latency test was administered on the following day. All requests for multiple sleep latency tests were made at the time of a clinic visit, before results of polysomnography were known, because the referring clinician thought that an assessment of sleepiness would be important to clinical care. Laboratory protocol requires medications likely to affect multiple sleep latency test results—such as stimulants or antidepressants—to be discontinued at least 10 days before the test is administered. This study was approved by the University of Michigan Institutional Review Board.

Digital polysomnography, scoring, and reliability safeguards in our laboratory were performed according to standard procedures and have been described previously (2, 20), and in more detail in the online supplement. The multiple sleep latency test, a well-validated objective measure of daytime sleepiness, also was administered according to a standard protocol (21). Sleep onset was defined as the first appearance of Stage 1 sleep. Apneas were defined by 10 or more seconds of cessation of thermocouple-defined airflow. Hypopneas were defined as 10-second or longer decrements (by at least 20%) in airflow, chest excursion (piezoelectric belts), or abdominal excursion followed by an oxygen desaturation of 4% or more, an arousal (22), or an awakening (20). The apnea–hypopnea index was calculated as the number of apneas or hypopneas per hour of sleep.

The RCREC were computed as described previously (17), except that spectral power was determined by digital filtering, rather than short-time Fourier transform, of the EEG signal recorded at the C3–A2 electrode. Additional details about methods used can be found in the online supplement. Digital filtering was implemented in MATLAB version 6.5.1 (The MathWorks, Natick, MA). Zero-phase forward and backward digital filtering was used to produce zero-phase distortion. To estimate the power in individual frequency bands, fifth order Butterworth passband filter functions were used. As in our previous work, the first 3 hours of scored sleep, approximately two sleep cycles, were analyzed. The RCREC data from the first 3 hours resemble that from the entire night (17), but the early portion of the night may reveal more about sleepiness in patients with SDB (23, 24). Respiratory cycles defined by the airflow signal maxima were divided into four segments: early expiration, late expiration, early inspiration, and late inspiration. The shortest 5%, longest 5%, smallest 5% (amplitude), and largest 5% of the respiratory cycles were rejected to avoid analysis of apneas, hypopneas, and other usual breaths. The EEG power was calculated for each respiratory cycle segment, normalized by division by the power during the entire cycle, and averaged across all corresponding segments for the roughly 1,500 to 3,000 breaths taken within the 3-hour period. For each subject, the RCREC were defined as the maximum difference between any two of the four mean segment-specific powers. The RCREC were calculated on the basis of power in several different EEG frequency ranges: delta (1–4 Hz), theta (5–7 Hz), alpha (8–12 Hz), sigma (13–15 Hz), and beta (16–30 Hz). In short, the RCREC represent the degree to which EEG activity in the specified frequency range varies in synchrony with the respiratory cycle.


The statistical significance of EEG variation with the respiratory cycle, for any given patient, was tested by a two-factor analysis of variance (ANOVA) in MATLAB (17, 18): the two factors were respiratory cycle number (e.g., 1–2,500) and respiratory segment number (1, 2, 3, or 4). As the underlying distribution of RCREC is not well known, the nonparametric Spearman correlation coefficient rho was used to test the strength of bivariate associations with measures of apnea severity (apnea–hypopnea index and minimal oxygen saturation) and sleep architecture (total sleep time, total recording time, and percent sleep time spent in each sleep stage). Linear regression models were used to test the strength of association between the mean sleep latency (MSL, on the multiple sleep latency test) and explanatory variables that had appeared useful in the bivariate analyses. For each model reported, a normal probability plot of residuals showed a normal distribution, and plots of residuals against each explanatory variable suggested adequate fit of the linear model. Analyses across subjects were performed with SAS version 8.01 (SAS Institute, Cary, NC), and the level of significance was set at p < 0.05.


Thirty-eight patients met inclusion criteria. Their demographic and polysomnographic data are summarized in Table 1

TABLE 1. Demographic and sleep data for 38 subjects

All Subjects (n = 38)

Apnea–Hypopnea Index > 5 (n = 27)

Apnea–Hypopnea Index ⩽ 5 (n = 11)

p Value*
Age, yr 41.7 ± 13.9 45.2 ± 13.7 33.2 ± 10.30.0130
Sex, no. (% male)20 (53%)18 (67%)2 (18%)0.0066
Total recording time, min493.0 ± 21.8489.2 ± 15.0502.3 ± 32.30.2216
Total sleep time, min415.1 ± 39.5405.3 ± 35.6439.2 ± 39.70.0141
Stage 1 sleep, % 23.2 ± 15.3 26.1 ± 16.516.2 ± 8.80.0227
Stage 2 sleep, % 50.9 ± 11.5 50.4 ± 12.552.3 ± 9.00.6478
Stage 3 or 4 sleep, % 9.6 ± 7.5 8.3 ± 7.813.0 ± 5.70.0760
Stage REM sleep, %16.2 ± 6.715.3 ± 7.218.5 ± 4.80.1804
Apnea–hypopnea index 18.5 ± 21.0 25.0 ± 21.8 2.6 ± 1.3< 0.0001
Minimum oxygen saturation, %86.3 ± 7.884.2 ± 8.391.5 ± 2.10.0002
Mean sleep latency, min
 6.9 ± 3.9
 6.5 ± 4.0
 7.7 ± 3.7

*For comparison between subjects with an AHI greater than 5 and subjects with an AHI of 5 or less; by t test or χ2 test, as appropriate.

On the Multiple Sleep Latency Test.

Values represent means ± SD, or percentages.

. Diagnoses established at the time of polysomnographic interpretation, with consideration of available historical data, included obstructive sleep apnea (n = 26), upper airway resistance syndrome (n = 5), primary snoring (n = 6), and possible narcolepsy in the absence of SDB and cataplexy (n = 1). Three subjects with obstructive sleep apnea and one with primary snoring had two or more sleep-onset REM periods on multiple sleep latency tests, but were thought unlikely to have narcolepsy because of their clinical histories.

For the purposes of this study, 27 subjects met a common polysomnographic criterion for significant SDB (apnea–hypopnea index > 5.0); their data are shown, separate from those for the remaining 11 subjects, in Table 1. Subjects with SDB, in comparison to those without SDB, were older and more likely to be male. Subjects with SDB had less total sleep time, a higher percentage of Stage 1 sleep, a higher apnea–hypopnea index, and lower minimum oxygen saturation. Mean sleep latency was not significantly lower among subjects with SDB than among subjects without SDB.

Respiratory Cycle–related EEG Changes

The magnitude of RCREC, within the five different frequency bands, is shown in Table 2

TABLE 2. Respiratory cycle-related eeg changes within each frequency band

All Subjects (n = 38)

Apnea–Hypopnea Index > 5 (n = 27)

Apnea–Hypopnea Index ⩽ 5 (n = 11)

Mean ± SD
No. (%) of Subjects with
 Significant RCREC*
Mean ± SD
No. (%) of Subjects with
 Significant RCREC*
Mean ± SD
No. (%) of Subjects with
 Significant RCREC*
p Value
Delta (1–4 Hz)0.094 ± 0.06725 (66)0.092 ± 0.06419 (70)0.097 ± 0.0766 (55)0.8415
Theta (5–7 Hz)0.061 ± 0.03223 (61)0.060 ± 0.03016 (59)0.065 ± 0.0397 (64)0.6391
Alpha (8–12 Hz)0.058 ± 0.03323 (61)0.060 ± 0.02620 (74)0.053 ± 0.0483 (27)0.6330
Sigma (13–15 Hz)0.097 ± 0.06129 (76)0.092 ± 0.05621 (78)0.108 ± 0.0738 (73)0.4854
Beta (16–30 Hz)
0.065 ± 0.042
30 (79)
0.067 ± 0.045
22 (81)
0.060 ± 0.031
8 (73)

*ANOVA, p < 0.05.

t Test for difference between subjects with an AHI greater than 5 and subjects with an AHI of 5 or less.

Definition of abbreviation: RCREC = respiratory cycle-related EEG changes.

, along with the numbers of subjects for whom RCREC were statistically significant. Every subject showed statistically significant RCREC in one or more frequency bands. The RCREC in one frequency band tended to correlate more closely with RCREC in adjacent rather than more distant frequencies, but no correlation exceeded rho = 0.56. The tendency for EEG power to vary with respiratory cycles was most prominent (largest RCREC) in sigma and delta frequencies (Table 2). The 27 subjects with an AHI greater than 5, in comparison with the 11 subjects with an AHI of 5 or less, showed no significant difference in average magnitudes of RCREC within specific EEG frequency ranges (Table 2).

To determine whether patterns of EEG variation with respiratory cycles showed consistency across subjects, respiratory cycle segment-specific EEG powers were averaged across subjects and plotted in Figure 1

(left column). These graphs and ANOVA for overall differences between segments showed that EEG activity varied in a consistent way, with respiratory cycle segments, only within delta, alpha, and sigma frequency bands.

Correlates of Objective Sleepiness

Plots of mean relative EEG power for each respiratory cycle segment, averaged separately across sleepy subjects and less sleepy subjects, revealed two important observations (Figure 1, middle and right columns, respectively). First, sigma and delta RCREC were more prominent and consistent among sleepy subjects than among less sleepy subjects. Among 15 subjects with a mean sleep latency of less than 5 minutes, sigma mean relative power was highest during late inspiration and lowest during late expiration. This pattern was much less pronounced in the 23 remaining subjects. Similarly, sleepy subjects showed prominent changes in delta mean relative power during the respiratory cycle, whereas more alert patients did not. In contrast to results for sigma and delta frequencies, no strong differences between sleepy and less sleepy subjects were noted in theta, alpha, and beta ranges. The second important observation is that between early and late inspiration (segments 3 and 4) among sleepy subjects, sigma mean relative power showed a prominent increase, whereas delta mean relative power showed a notable, simultaneous decrease.

In bivariate analyses, no demographic or standard sleep variable listed in Table 1 provided a statistically significant prediction of next-day sleepiness, as reflected by the MSL (all p > 0.05), except for percent Stage 3 or 4 sleep, apnea–hypopnea index, and minimal oxygen saturation (Table 3)

TABLE 3. Spearman correlation coefficient rho (p value) for association between mean sleep latency and previous night polysomnographic measures

All Subjects
 (n = 38)

Apnea–Hypopnea Index > 5
 (n = 27)

Apnea–Hypopnea Index ⩽ 5
 (n = 11)
Stage 3 or 4 sleep, %0.39 (0.0153)0.24 (0.2293) 0.77 (0.0051)
Stage REM sleep, %0.31 (0.0557)0.43 (0.0246)−0.41 (0.2104)
Apnea–hypopnea index−0.34 (0.0396)−0.28 (0.1579) −0.42 (0.2026)
Minimal oxygen saturation0.40 (0.0135)0.44 (0.0221)−0.30 (0.3774)
Delta RCREC−0.16 (0.3322)−0.33 (0.0936) 0.15 (0.6590)
−0.49 (0.0017)
−0.69 (< 0.0001)
 −0.07 (0.8418)

MSL, on the multiple sleep latency test.

All Table 1 and Table 2 polysomnographic variables that are not listed here showed no association or trend with mean sleep latency (p > 0.10).

Definition of abbreviation: REM = rapid eye movement.

. Sigma RCREC showed a comparatively robust association with MSL (rho = –0.49, p = 0.0017), whereas other frequency-specific RCREC did not. Among the 27 subjects who had SDB, the correlation between MSL and sigma RCREC became notably stronger (rho = –0.69, p < 0.0001), as did that between MSL and delta RCREC (rho = –0.33, p = 0.0936). In contrast, the correlation between MSL and the apnea–hypopnea index did not become stronger (Table 3 and Figure 2).

In a multiple regression model of MSL, explanatory variables were limited at first to those standard polysomnographic measures for which bivariate analyses had shown an association or trend with MSL (as listed in Table 3). Sigma RCREC was then added to this model. Results before and after addition of sigma RCREC are shown in Table 4

TABLE 4. All subjects*: regression of sleepiness on standard sleep variables before and after addition of sigma respiratory cycle-related eeg changes to the model

Explanatory Variables

Total Model

p Value
100 × R2
p Value
Before addition of sigma RCREC
 Stage 3 sleep, %0.1470.0940.126322.60.0696
 Stage REM sleep, %0.1720.1210.1627
 Apnea–hypopnea index−0.0060.0440.8895
 Minimal oxygen saturation−0.0010.1130.9918
After addition of sigma RCREC
 Stage 3 sleep, %0.1520.0830.076441.10.0034
 Stage REM sleep, %0.2020.1070.0689
 Apnea–hypopnea index−0.0230.0390.5602
 Minimal oxygen saturation−0.0720.1020.4864
 Sigma RCREC

*n = 38.

Mean sleep latency (MSL).

Standard sleep variables were those that showed associations with MSL in initial bivariate correlations (Table 3).

. Before taking sigma RCREC into account, the best standard variables explained only a marginally nonsignificant 22.6% of the variance in sleepiness. After taking sigma RCREC into account, the model explained 41.1% of the variance in sleepiness (p = 0.0034). The same analyses were repeated after confining the sample to those 27 subjects who had an apnea–hypopnea index greater than 5 (Table 5)

TABLE 5. Subjects* with apnea–hypopnea index greater than 5: regression of sleepiness on standard sleep variables before and after addition of sigma respiratory cycle-related eeg changes to the model

Explanatory Variables

Total Model

p Value
100 × R2
p Value
Before addition of sigma RCREC
 Stage 3 sleep, %0.0510.1020.628632.30.0624
 Stage REM sleep, %0.3060.1370.0365
 Apnea–hypopnea index0.0100.0460.8363
 Minimal oxygen saturation0.0110.1160.9256
After addition of sigma RCREC
 Stage 3 sleep, %0.0590.0710.418569.2< 0.0001
 Stage REM sleep, %0.3550.0950.0012
 Apnea–hypopnea index−0.0120.0320.7238
 Minimal oxygen saturation−0.1300.0850.1399
 Sigma RCREC
< 0.0001

*n = 27.

Mean sleep latency (MSL).

Standard sleep variables were those that showed associations with MSL in initial bivariate correlations (Table 3).

. In this subset, standard variables together explained 32.3% of the variance in sleepiness (p = 0.0624). After the addition of sigma RCREC to the model, it explained 69.2% of the variance (p < 0.0001).

This study of patients evaluated for SDB demonstrates for the first time in adults that EEG activity varies in synchrony with nonapneic respiratory cycles. Furthermore, the extent of synchronization between respiration and EEG power, at least in the sigma frequencies (13–15 Hz), predicts next-day sleepiness as measured by the multiple sleep latency test. The predictive value was particularly strong among subjects who proved to have SDB: in this subgroup, sigma RCREC alone predicted nearly 50% of the variance in sleepiness in a nonparametric correlation analysis. The addition of sigma RCREC to a regression model that already included two common polysomnographic measures of SDB severity, the rate of apneas and hypopneas and extent of oxygen desaturations, substantially improved the ability to predict sleepiness. Therefore, sigma RCREC appears to provide unique predictive ability not already reflected by standard measures. These findings could have important implications for clinical practice, research, and understanding of SDB.

Although sleepiness is one of the most important and reversible consequences of SDB (25), the mechanisms by which SDB causes sleepiness are not completely understood. Arousals seen at the termination of apneas, hypopneas, or “respiratory event–related arousals” are thought to fragment sleep sufficiently to contribute to daytime sleepiness. Artificially induced arousals at rates similar to those seen in upper airway resistance syndrome, for example, can produce daytime sleepiness (26). Short, brief, but frequent arousals are suspected to produce sleepiness more readily than longer but less frequent awakenings (8). However, the extremes of arousal brevity, frequency, or nature most likely to produce sleepiness have not been defined. Some studies also suggest an effect of hypoxemia on sleepiness, independent of the rates of apneic events or associated arousals (2).

The current data replicate previous findings in that they show mild unadjusted associations between sleepiness and both the apnea–hypopnea index and minimum oxygen saturation (2). The substantial additional contribution of RCREC in predicting sleepiness among our subjects suggests that this outcome of SDB may depend not only on apneic events, associated arousals, or oxygen desaturations, but also to a major extent on RCREC or the processes that underlie them. The present results are correlative, and cannot prove causation, but they are consistent with the hypothesis that frequent, brief, computer-detected cortical responses to increased work of breathing may be an important mechanism by which SDB causes sleepiness.

Perhaps the most surprising finding is that RCREC in sigma frequencies, rather than other EEG frequencies, were most predictive of sleepiness. The two-process model of sleep and wakefulness might have predicted that disruption of slow wave (delta) activity would most significantly affect sleepiness (27). During non-REM sleep, responsiveness to sensory stimulation decreases as EEG delta power is increased (28). Spectral analysis in adult sleep apnea patients, although based on 4-second windows and other methods quite different from our own, implicated deficient delta power early in the night, and perhaps a lack of the normal decrement in delta power through the night, as responsible for sleepiness in SDB (24). Our own previous investigation of RCREC in children suggested that delta RCREC best predicts improvement in sleepiness after treatment for SDB (18). However, sigma EEG frequencies were not studied in either of these previous investigations that implicated abnormal slow wave activity as a cause of SDB-related sleepiness (18, 24). In cross-sectional analyses among children scheduled for adenotonsillectomy, high frequency (alpha) rather than delta RCREC showed associations with sleepiness (18).

The physiological significance of sigma frequency activity, also called spindle frequency activity (29), is not completely understood. Although spindles themselves are discrete, 0.5- to 2-second EEG patterns, their occurrence shows close correspondence to the spectral power, within relevant frequencies, analyzed over longer periods (30). The EEG activity in this range is produced by cortical projections from the reticular nucleus of the thalamus. During non-REM sleep, neurons within this nucleus become hyperpolarized, generate EEG spindles, and block afferent input to the cortex, thus helping to maintain sleep (31). Spindle frequency activity is notable for particularly high inter- and intrahemispheric coherence: when this activity occurs in one part of the cortex, other regions tend to show the same (32, 33). This uniquely high coherence has led to the speculation that spindle frequency activity could have important physiological significance (32).

Sigma frequency activity does vary (reciprocally) with delta power during sleep cycles, and potentially could reflect increased central arousal (34, 35). If so, then recurrent brief increases in sigma power with labored inspirations may represent a form of microarousals that disrupts the restorative process of sleep. Our sleepy subjects, but not our more alert subjects, experienced a consistent increase in sigma activity during the course of inspiration along with a simultaneous decrease in delta activity (Figure 1). Consistent with the hypothesis that these RCREC may represent significant inspiratory microarousals, the only published comparisons of RCREC from different sleep stages, in a single child with SDB, showed that sigma RCREC was most prominent specifically during non-REM Stages 3 and 4 sleep (17). Data from normal adult men show that spindle frequency activity is particularly prominent at transitions into and out of non-REM portions of sleep cycles (30). We speculate that spindle frequency RCREC could repeatedly interrupt the progressive hyperpolarization of thalamocortical neurons thought necessary for the development of slow wave sleep.

Among the limitations of this study is the fact that these findings cannot go so far as to prove that RCREC represent microarousals, or that they cause sleepiness. The results do encourage future studies that involve prospective patient intervention and animal models to test causal hypotheses. Alternative explanations for RCREC are feasible. Wakefulness is known to affect the drive to breathe (36), but the idea that cortical microarousals could gait the respiratory cycle—thought to be generated at medullary brainstem levels (37)—is difficult to conceive. Another possibility is that RCREC reflect some third variable that is responsible for sleepiness. Our analyses included multiple between-subject comparisons in a limited sample size, and therefore these findings merit confirmation in larger samples. The existence of RCREC raises many questions that have yet to be addressed, including the relative influence of RCREC during specific sleep stages. Different EEG recording sites also remain to be investigated. Current observations were made at a central EEG derivation, where sigma activity is probably strongest (30).

Findings in this study are not likely to be an artifact of the type of equipment used to obtain standard polysomnographic data. Nasal pressure monitoring can identify subtle hypopneas that are not detected by thermocouples, and calibrated inductance plethysmography may generate a more accurate measure of chest and abdominal movement than that provided by piezoelectric belts (12). No studies to our knowledge have shown that nasal pressure monitoring or inductance plethysmography can improve prediction of health-related outcomes such as sleepiness. However, as we did not use the most sensitive methods to identify subtle hypopneas, we cannot exclude the possibility that some may have occurred during the “nonapneic sleep” in which we identified RCREC.

In conclusion, this study demonstrates that RCREC exist in some adults with SDB. The finding raises the possibility that SDB exerts its influence throughout sleep, rather than solely through the minority of the night spent within readily apparent apneas and hypopneas. Suspicion of this possibility arose from esophageal pressure and nasal pressure recordings that often show continuous, increased upper airway resistance outside defined respiratory events (38, 39). The new measure may serve to illustrate and quantify an important physiological pathway that translates these numerous, labored breaths into neurological consequences. Any new measure, to be clinically useful, must fail to correlate well with older, standard measures that do not adequately predict known consequences of the disorder. The RCREC demonstrate this property: with virtually no correlation with the apnea–hypopnea index, sigma RCREC instead correlate with an important outcome, excessive daytime sleepiness. The RCREC could prove to have clinical utility as an objective predictor of SDB-engendered sleepiness, especially in the common clinical situation in which a Multiple Sleep Latency Test is unavailable or impractical.

This early study of a novel analytic approach sought to compare RCREC and standard polysomnographic measures, side by side, in a limited number of patients. This research design and sample size cannot provide compelling evidence that standard measures have limited utility in prediction of an important SDB outcome, but larger studies have already shown this (2, 3). Our findings do not diminish the possibility that other new or modified technology also may improve the value of polysomnography for assessment of SDB. However, in the absence of demonstrated outcome-based clinical value for most such methods, our initial findings with a novel approach to SDB do raise several attractive possibilities. The RCREC analysis is automated, potentially instantaneous, invariant between scorers, and achievable without any new hardware. Initial data on clinical utility are promising, and many aspects of the RCREC analysis that could improve its usefulness still remain to be investigated. These include identification of optimal EEG and respiratory leads, analysis of RCREC within specific sleep stages, alternative approaches to segmentation of the respiratory cycle, and perhaps examination of apneic sleep in addition to nonapneic sleep.

The authors thank Ralph Lydic, Ph.D., for his critical reading of this manuscript and insightful comments. The authors also thank Morton B. Brown, Ph.D., and Kenneth E. Guire, M.S., for assistance with statistical analyses.

1. Guilleminault C, Partinen M, Quera-Salva MA, Hayes B, Dement WC, Nino-Murcia G. Determinants of daytime sleepiness in obstructive sleep apnea. Chest 1988;94:32–37.
2. Chervin RD, Aldrich MS. Characteristics of apneas and hypopneas during sleep and relation to excessive daytime sleepiness. Sleep 1998;21:799–806.
3. Gottlieb DJ, Yao Q, Redline S, Ali T, Mahowald MW. Does snoring predict sleepiness independently of apnea and hypopnea frequency? Am J Respir Crit Care Med 2000;162:1512–1517.
4. Chervin RD, Guilleminault C. Assessment of sleepiness in clinical practice. Nat Med 1995;1:1252–1253.
5. Norman RG, Ahmed MM, Walsleben JA, Rapoport DM. Detection of respiratory events during NPSG: nasal cannula/pressure sensor versus thermistor. Sleep 1997;20:1175–1184.
6. Guilleminault C, Stoohs R, Clerk A, Cetel M, Maistros P. A cause of excessive daytime sleepiness: the upper airway resistance syndrome. Chest 1993;104:781–787.
7. Bennett LS, Barbour C, Langford B, Stradling JR, Davies RJ. Health status in obstructive sleep apnea: relationship with sleep fragmentation and daytine sleepiness, and effects of continuous positive airway pressure treatment. Am J Respir Crit Care Med 1999;159:1884–1890.
8. Stepanski E, Lamphere J, Badia P, Zorick F, Roth T. Sleep fragmentation and daytime sleepiness. Sleep 1983;7:18–26.
9. Roehrs T, Zorick F, Wittig R, Conway W, Roth T. Predictors of objective level of daytime sleepiness in patients with sleep-related breathing disorders. Chest 1989;95:1202–1206.
10. O'Malley EB, Norman RG, Farkas D, Rapoport DM, Walsleben JA. The addition of frontal EEG leads improves detection of cortical arousal following obstructive respiratory events. Sleep 2003;26:435–439.
11. Zou D, Grote L, Eder DN, Peker Y, Hedner J. Obstructive apneic events induce α-receptor mediated digital vasoconstriction. Sleep 2004;27:485–489.
12. American Academy of Sleep Medicine Task Force. Sleep-related breathing disorders in adults: recommendations for syndrome definition and measurement techniques in clinical research. Sleep 1999;22:667–689.
13. Guilleminault C, Do Kim Y, Chowdhuri S, Horita M, Ohayon M, Kushida C. Sleep and daytime sleepiness in upper airway resistance syndrome compared to obstructive sleep apnoea syndrome. Eur Respir J 2001;17:838–847.
14. Svanborg E, Guilleminault C. EEG frequency changes during sleep apneas. Sleep 1996;19:248–254.
15. Black JE, Guilleminault C, Colrain IM, Carrillo O. Upper airway resistance syndrome: central electroencephalographic power and changes in breathing effort. Am J Respir Crit Care Med 2000;162:406–411.
16. Bandla HPR, Gozal D. Dynamic changes in EEG spectra during obstructive apnea in children. Pediatr Pulmonol 2000;29:359–365.
17. Chervin RD, Burns JW, Subotic NS, Roussi C, Thelen B, Ruzicka DL. Method for detection of respiratory cycle-related EEG changes in sleep-disordered breathing. Sleep 2004;27:110–115.
18. Chervin RD, Burns JW, Subotic NS, Roussi C, Thelen B, Ruzicka DL. Correlates of respiratory cycle-related EEG changes in children with sleep-disordered breathing. Sleep 2004;27:116–121.
19. Chervin RD, Burns JW, Subotic NS, Roussi C, Ruzicka DL, Weatherly RA, Dillon JE, Hodges E, Giordani BJ. Respiratory cycle-related EEG changes (RCREC) predict neurobehavioral outcomes in childhood OSA [abstract]. Sleep 2004;27(Suppl):A99.
20. Rechtschaffen A, Kales A. A manual of standardized terminology, techniques and scoring system for sleep stages of human subjects. Los Angeles: Brain Information Service/Brain Research Institute, UCLA; 1968. p. 1–12.
21. Carskadon MA, Dement WC, Mitler MM, Roth T, Westbrook PR, Keenan S. Guidelines for the Multiple Sleep Latency Test (MSLT): a standard measure of sleepiness. Sleep 1986;9:519–524.
22. American Sleep Disorders Association. EEG arousals: scoring rules and examples. Sleep 1992;15:173–184.
23. Morisson F, Décary A, Petit D, Lavigne G, Malo J, Montplaisir J. Daytime sleepiness and EEG spectral analysis in apneic patients before and after treatment with continuous positive airway pressure. Chest 2001;119:45–52.
24. Heinzer R, Gaudreau H, Decary A, Sforza E, Petit D, Morisson F, Montplaisir J. Slow-wave activity in sleep apnea patients before and after continuous positive airway pressure treatment: contribution to daytime sleepiness. Chest 2001;119:1807–1813.
25. Engleman HM, Martin SE, Kingshott RN, Mackay TW, Deary IJ, Douglas NJ. Randomised placebo controlled trial of daytime function after continuous positive airway pressure (CPAP) therapy for the sleep apnoea/hypopnoea syndrome. Thorax 1998;53:341–345.
26. Philip P, Stoohs R, Guilleminault C. Sleep fragmentation in normals: a model for sleepiness associated with upper airway resistance syndrome. Sleep 1994;17:242–247.
27. Borbély AA. A two process model of sleep regulation. Hum Neurobiol 1982;1:195–204.
28. Borbély AA, Baumann F, Brandeis D, Strauch I, Lehmann D. Sleep deprivation: effect on sleep stages and EEG power density in man. Electroencephalogr Clin Neurophysiol 1981;51:483–493.
29. Borbély AA, Achermann P. Sleep homeostasis and models of sleep regulation. In: Kryger MH, Roth T, Dement WC, editors. Principles and practice of sleep medicine, 3rd ed. Philadelphia, PA: W. B. Saunders; 2000. pp. 377–390.
30. Dijk DJ, Hayes B, Czeisler CA. Dynamics of electroencephalographic sleep spindles and slow wave activity in men: effect of sleep deprivation. Brain Res 1993;626:190–199.
31. Steriade M, McCormick DA, Sejnowski TJ. Thalamocortical oscillations in the sleeping and aroused brain. Science 1993;262:679–685.
32. Achermann P, Borbély AA. Coherence analysis of the human sleep electroencephalogram. Neuroscience 1998;85:1195–1208.
33. Ferrara M, De Gennaro L, Curcio G, Cristiani R, Bertini M. Regional differences of the temporal EEG dynamics during the first 30 min of human sleep. Neurosci Res 2002;44:83–89.
34. Uchida S, Maloney T, March JD, Azari R, Feinberg I. Sigma (12–15 Hz) and delta (0.3–3 Hz) EEG oscillate reciprocally within NREM sleep. Brain Res Bull 1991;27:93–96.
35. Uchida S, Maloney T, Feinberg I. Beta (20–28 Hz) and delta (0.3–3 Hz) EEGs oscillate reciprocally across NREM and REM sleep. Sleep 1992;15:352–358.
36. Fink BR, Hanks EC, Ngai SH, Papper EM. Central regulation of respiration during anesthesia and wakefulness. Ann N Y Acad Sci 1963;109:892–900.
37. Richter DW, Spyer KM. Studying rhythmogenesis of breathing: comparison of in vivo and in vitro models. Trends Neurosci 2001;24:464–472.
38. Guilleminault C, Stoohs R, Clerk A, Simmons J, Labanowski M. Excessive daytime somnolence in women with abnormal respiratory efforts during sleep. Sleep 1993;16:S137–S138.
39. Guilleminault C, Stoohs R, Kim Y, Chervin R, Black J, Clerk A. Upper airway sleep-disordered breathing in women. Ann Intern Med 1995;122:493–501.
Correspondence and requests for reprints should be addressed to Ronald D. Chervin, M.D., M.S., Michael S. Aldrich Sleep Disorders Laboratory, University Hospital, 8D8702, P.O. Box 0117, 1500 E. Medical Center Drive, Ann Arbor, MI 48109-0117. E-mail:


No related items
American Journal of Respiratory and Critical Care Medicine

Click to see any corrections or updates and to confirm this is the authentic version of record