INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Keywords: polysomnography; nocturnal myoclonus syndrome; sleep apnea; obstructive; sleepiness; Multiple Sleep Latency Test

At sleep disorders centers, periodic leg movements during sleep (PLMS) are most often identified during polysomnography performed to assess for sleep-disordered breathing (SDB). When PLMS occur at rates of five or more per hour of sleep, they may be associated with excessive daytime sleepiness and periodic limb movement disorder may be diagnosed (1). However, the extent to which PLMS contribute to excessive daytime sleepiness is controversial (2).

Frequent arousals can cause daytime sleepiness (3), and some PLMS are associated with arousals that can be scored on polysomnograms according to standard electroencephalogram (EEG) criteria (4). Data from at least two clinical series, though, failed to reveal any positive correlation between the number of PLMS or PLM-arousal complexes per hour of sleep and severity of sleepiness as measured objectively by the Multiple Sleep Latency Test (MSLT) (5, 6). Other studies have shown that increases in heart rate after PLMS are similar whether or not EEG-defined arousals occur (7, 8), that spectral analysis may detect EEG changes after PLMS not accompanied by visible EEG-arousals (9), and that nonvisible arousals may still produce excessive daytime sleepiness (10). Therefore, in short, the significance of PLMS and arousals sometimes scored with them remains poorly understood.

We previously assembled a clinical database of polysomnographic results from more than 1,100 patients to explore associations between SDB, sleep architecture, and excessive daytime sleepiness (11). This series of patients with suspected or confirmed SDB is the largest published, to our knowledge, that contains both nocturnal polysomnographic and daytime MSLT data, and this report extends our analysis to include data on associations between PLMS and sleepiness.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects and Measures

Data were extracted from a database of over 9,000 studies performed for clinical indications in our laboratory, between January 1, 1988 and February 1, 1997, in patients with suspected or confirmed SDB. Current inclusion and exclusion criteria have been described previously (12), except that age now was required to be 18 yr or more. Data from 1,124 subjects form the basis for this report. Nocturnal polysomnography and scoring of studies followed standard clinical procedures, as previously published (12).

Periodic limb movements were scored when they met criteria for duration (0.5 to 5 s), periodicity (5 to 120 s between each movement), and number (3 or more in a row). For this study, periodic limb movement disorder (PLMD) was considered present when the rate of leg movements per hour of sleep (periodic limb movement index [PLMI]) was at least 5 (1). For a subsample of 321 studies, the occurrence of an arousal at the same time or else just after each PLM was also scored (4). Arousals were scored for those patients studied within the last 4 yr of data collection and in specific sleep recording rooms; assignment to these rooms was essentially but not formally random. The PLM-arousal index and the PLM-nonarousal index were calculated as the number of events per hour of sleep (1).

The MSLTs followed standard methods for collection of electroencephalographic, electro-oculographic, and chin electromyographic data (15). Each patient's mean sleep latency was calculated as the time, averaged across all nap attempts, from "lights out" to the first epoch of stage 1 sleep. Measures of subjective problem sleepiness and sleep propensity also were available for some subjects. A total of 873 subjects had answered a simple, previously reported (13) laboratory questionnaire item about problem sleepiness"How often do you have a major problem with sleepiness during the daytime?"with a Likert scale response between 1 (never) and 5 (almost always). Subjects who answered 4 (often) or 5 were considered to have problem sleepiness. A subset of 201 subjects, studied in the last 2 yr of data collection, also had completed the Epworth Sleepiness Scale (ESS), a validated subjective measure of sleep propensity (16).

Analysis

Linear regression models were used to test for associations between MSLT-defined sleepiness as an outcome variable and PLMI, apnea/ hypopnea index (AHI; number per hour of sleep), minimal oxygen saturation, age, and sex as explanatory variables. Sleepiness was represented as the square root of the mean sleep latency (MSLRT) because the mean sleep latency did not follow a normal distribution. For some analyses, subjects were divided into one group with more severe SDB (AHI > 45) and another with less severe SDB (AHI  45). Data for subjective sleep propensity were analyzed with similar models, as the ESS score followed a normal distribution. Problem sleepiness was modeled with logistic regression. In tests of statistical significance, the level was set at 0.05.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

Ages of the 1,124 subjects ranged from 18 to 85 yr; the mean was 46 ± 12 (s.d.) yr and 790 (70%) were male (Table 1). Nine hundred forty-four subjects (84%) had an AHI > 5; 792 subjects (70%) had an AHI > 10; 266 subjects (24%) had a PLMI > 5; and 399 subjects (35%) had a mean sleep latency < 5 min. In comparison to the 854 subjects with milder SDB (AHI  45), the 270 (24%) with more severe SDB were 4 yr older on average and were more predominantly male. They had higher rates of apneas and hypopneas, lower minimal oxygen saturation, more daytime sleepiness, and slightly lower rates of periodic leg movements (Table 1).

Correlates of Objective Sleepiness

Simple regression models showed that objective sleepiness (lower MSLRT) was associated with higher AHI, lower minimal oxygen saturation, severe SDB, and male sex (p < 0.0001 for each model, Table 2). The AHI accounted for 11% of the variance in MSLRT and minimal oxygen saturation accounted for 12%. Sleepiness showed no association with age (p > 0.05). Sleepiness showed statistically significant associations with lower PLMI and absence of PLMD (p < 0.05), but each of these explanatory variables accounted for less than 1% of the variance in MSLRT. Patients with the highest PLMIs (PLMI > 20, n = 46) had a mean MSL of 7.7 ± 5.0, whereas those with the lowest (PLMI = 0, n = 665) had a mean MSL of 7.1 ± 4.6. 

Terms that represented the interaction of PLMI with AHI, severe SDB, and minimal oxygen saturation also were tested. The interaction of PLMI with AHI was not significant (p = 0.81) but that of PLMI with severe SDB was (p = 0.0035), as was that of PLMI with minimal oxygen saturation (p = 0.0053). To better assess the influence of SDB on the association between PLMI and MSLRT, the subjects were divided into groups with and without severe SDB for further analysis.

Subjects with and without Severe SDB

Among the 270 subjects with AHI > 45, minimal oxygen saturation explained 13% of the variance in MSLRT, AHI explained 4%, and PLMI also explained 4% (p < 0.002 for each, Table 3). A multiple regression of MSLRT on PLMI, AHI, minimal oxygen saturation, and age showed that after accounting for these potential confounds, PLMI still explained a small but significant amount of the variance in MSLRT (100*part R² = 1%, p = 0.0475), as did minimal oxygen saturation (9%, p < 0.0001) but not AHI (0%, p = 0.3552). Lower PLMI was independently associated with more sleepiness, as shown graphically in Figure 1A.

View larger version (26K): [in this window] [in a new window]   Figure 1.   Median and 10th, 25th, 75th, and 90th percentiles for mean sleep latency (MSL) among patients with and without five periodic leg movements per hour of sleep (PLMD). In patients with severe SDB (A), absence of PLMD was associated with lower MSL, but this pattern was not seen in patients without severe SDB (B).

Among the 854 subjects with AHI  45, simple regression models showed that minimal oxygen saturation explained only 2% of the variance in MLSRT, AHI explained 3% (p < 0.0001 for each), but PLMI explained none (p = 0.9408, Figure 1B).

Subjects with Moderate or High Rates of PLMS

Many patients had few or no PLMS, and theoretically could have obscured an association between sleepiness and PLMI among other patients. However, among the 266 patients with PLMI > 5, PLMI still failed to show any association with MSLRT in a simple regression model (p = 0.91), and adjustment for AHI, minimal oxygen saturation, age, and sex did not change this result (p = 0.48).

PLMS with and without Arousals

The 321 subjects for whom PLM-arousal and PLM-nonarousal indices were available did not differ from the remaining 803 subjects in age, sex, or PLMI (p > 0.30 for each). However, the 321 subjects did have somewhat lower AHI (25.5 ± 24.7 versus 33.4 ± 35.8, p < 0.0001), higher minimal oxygen saturation (80.1 ± 10.0 versus 77.7 ± 15.8, p = 0.003), and higher mean sleep latency (7.9 ± 4.5 versus 7.1 ± 4.6, p = 0.02). The mean PLM-arousal index was 1.6 ± 3.2 and the mean PLM-nonarousal index was 2.2 ± 4.2.

A higher PLM-arousal index was associated with less sleepiness and explained 2% of the variance of MSLRT (Figure 2 and Table 4). The PLM-nonarousal index explained none of this variance. The difference between the PLM-arousal and nonarousal indices, adjusted for the overall PLMI, explained 2% of the variance in MSLRT (Table 4). That difference, after additional adjustment for AHI, minimal oxygen saturation, age, and sex, still accounted for 2% of the variance in MSLRT (p = 0.0198, not shown in table).

View larger version (29K): [in this window] [in a new window]   Figure 2.   Median and 10th, 25th, 75th, and 90th percentiles for mean sleep latency (MSL) among patients with and without (A) more than five PLM-arousals per hour of sleep, and (B) more than five PLM-nonarousals per hour of sleep. Presence of PLM-arousals, but not PLM-nonarousals, was associated with a relative increase in the MSL.

Subjective Sleepiness

Among the 873 subjects for whom data on problem sleepiness were available, the mean age was 46 ± 12 yr, 624 (71%) were male, the mean AHI was 32.6 ± 34.6, the mean minimal oxygen saturation was 77.9 ± 15.0, the mean PLMI was 3.4 ± 6.3, and the mean sleep latency on the MSLT was 7.3 ± 4.5 min. Problem sleepiness, which was reported in 457 (52%) of the subjects, showed small associations with younger age (odds ratio [OR] = 0.99), higher AHI (OR = 1.01), severe SDB (OR = 1.36), and lower minimal oxygen saturation (OR = 0.99) (p < 0.05 for each), but not sex, PLMI, or PLMD. Among patients with PLMI > 20 who answered the questionnaire item about problem sleepiness (n = 35), 51% reported this symptom, whereas among those with PLMI = 0 (n = 533), 53% did the same. The interactions of PLMI with AHI, severe SDB, and minimal oxygen saturation were all nonsignificant. No association was found between problem sleepiness and either PLM-arousal or PLM-nonarousal indices among the 220 subjects for whom these measures were also available. Among 198 subjects with PLMI > 5 and available problem sleepiness ratings, the latter still showed no association with PLMI, either before (p = 0.57) or after (p = 0.79) adjustment for AHI, minimal oxygen saturation, age, and sex.

Among the 201 subjects for whom ESS data were available, the mean age was 43 ± 12 yr, 132 (66%) were male, the mean AHI was 23.3 ± 34.0, the mean minimal oxygen saturation was 82.9 ± 9.9, the mean PLMI was 3.7 ± 6.2, the mean sleep latency on the MSLT was 7.8 ± 4.7 min, and the mean ESS score was 11.3 ± 5.1. Higher ESS scores showed no association with age, sex, AHI, severe SDB, minimal oxygen saturation, PLMI, or PLMD (all p > 0.05). Among patients with PLMI > 20 and available ESS scores (n = 6), the mean score was 12.8 ± 3.8, whereas among those with PLMI = 0 (n = 106), the mean was 11.6 ± 5.4. The interactions of PLMI with AHI, severe SDB, and minimal oxygen saturation were all nonsignificant. No association was found between ESS scores and either PLM-arousal or PLM-nonarousal indices among the 88 subjects for whom these measures were also available. Among 52 subjects with PLMI > 5 and available ESS scores, the latter still showed no association with PLMI, either before (p = 0.80) or after (p = 0.41) adjustment for AHI, minimal oxygen saturation, age, and sex.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The results of this study cast doubt on two prevalent beliefs: that incidentally discovered PLMS may cause excessive daytime sleepiness, and that such PLMS are more likely to cause sleepiness when they are accompanied by nocturnal arousals. Among patients suspected or confirmed to have SDB, increased numbers of PLMS were not associated with increased sleepiness as objectively measured by a test often considered to be a gold standard, the MSLT (17). In subsets of patients for whom additional data on subjective problem sleepiness or sleep propensity had been gathered, no association with PLMS could be demonstrated. For the first time in a large sample, PLMS with and without arousals were compared with MSLT results, and increased rates of either form of PLMS failed to predict excessive daytime sleepiness. Surprisingly, our data suggested that in some cases the opposite may occur. Among patients with more severe SDB, the rate of PLMS showed a small but highly significant association with less severe MSLT-defined sleepiness. In the subset of patients for whom arousals had been scored, PLMS with arousals predicted less sleepiness, and PLMS without arousals showed no predictive value.

The belief that PLMS may cause sleepiness is expressed in early reports on PLMS (18), in widely used sleep texts (21, 22), and in the International Classification of Sleep Disorders (1). Experimentally induced arousals are known to cause daytime sleepiness (3, 23), and the assumption is often made that PLMS may cause sleepiness because they are often followed by arousals. In contrast, our findings fail to provide any evidence that PLMS, associated or not with arousals, produce daytime sleepiness. A previous study of PLM-arousals in patients with or without SDB produced similar findings (6). Our results now show that the absence of an association between PLMS and sleepiness is not a by-product of potential confounding by SDB severity, age, and sex, or of an unwarranted focus on arousals. A recent study found that arousals recorded with PLMS more often precede rather than follow the movements (24). The same patients showed few differences in sleep architecture between the portions of the nights affected by PLMS and those unaffected by the movements (25). Some researchers have concluded that periodic limb movement disorder with consequent sleepiness does not exist independent of other sleep or neurodegenerative disorders (2, 26). Our results appear to provide some support for this contention.

Our counterintuitive findings that higher rates of PLMS, and especially PLM-arousals, are associated with less severe MSLT-defined sleepiness are consistent with previously published data from 518 patients with obstructive sleep apnea (6). In that study, the PLM-arousal index explained a statistically significant 1% of the variance in MSLT results; the finding was not explained, but the investigators noted the small magnitude of association and speculated that the relationship may have resulted from an untested confounding variable known to co-vary with PLMI, namely age (27). Results of our models that adjusted for age suggest otherwise. An association between increased sleepiness and less frequent PLM-arousals raises the possibility that excessive sleepiness may reduce the frequency of arousals more than the underlying frequency of PLMS. However, this hypothesis does not explain why the overall PLMI shares a small amount of the variance with sleepiness, and a larger amountequivalent to that shared by AHI and sleepinessamong subjects with more severe SDB.

One possibility is that PLMS, which commonly occur near sleep onset, prolong sleep latency during MSLT naps. In patients with more severe SDB, apneas and hypopneas may further delay onset of a daytime nap when they are triggered by PLMS and associated arousals. Another possible explanation is that PLMS reflect some physiologic mechanism by which the brain reduces sleepiness. Associated arousals do not necessarily result from PLMS, and both could be by-products of a third neurophysiological event that promotes wakefulness in sleepy or sleep-disordered individuals (24, 28). Such an explanation might account for observations that PLMS are common among patients with several different types of sleep disorders (21). The cyclic alternating pattern (CAP), prominent in the sleep EEG of patients with disparate sleep disorders, sometimes may represent brain mechanisms that maintain sleep in the face of arousing stimuli (29, 30). Some evidence suggests that CAP may serve as a gating mechanism for PLMS (31).

Several limitations of the current study deserve mention. A periodic leg movement, especially if associated with an arousal, may be accompanied by mild hyperventilation that can make normal ventilation between successive PLMS difficult to distinguish from hypopneas. Our polysomnograms were scored by technicians who had undergone rigorous training and who had demonstrated good proficiency on reliability records, but even the most expert technicians following standard scoring rules can have difficulty distinguishing subtle hypopneas and PLMS in some cases. Clinical research recommendations published after the current data were collected suggest monitoring of nasal pressure to increase sensitivity for subtle hypopneas (32), arousals from which can be difficult to distinguish from PLMS. With this technique, we might have identified PLMS more accurately and improved our ability to detect an association with sleepiness. However, PLMIs derived in the presence or absence of nasal pressure monitoring are likely to be strongly correlated, and either would be expected to show an association with measures of sleepiness, in a sample size this large, if PLMS do in fact contribute to sleepiness.

Only 29% of our records were scored for PLM-arousals and nonarousals, and these records were not a strictly random sample of all those performed. The subsample had somewhat less severe sleep apnea and slightly less severe sleepiness, perhaps because they were studied in more recent years, when increased physician awareness of sleep disorders may have led to referral of subjects less severely affected than those referred in previous years. We are not aware of a specific way in which this or any other bias might explain the observations we report in the subsample. Finally, our study did not include patients whose primary problem was restless legs syndrome, periodic limb movement disorder, or narcolepsy. In some patients with sleep disorders other than SDB, PLMS may still be a cause of excessive daytime sleepiness.

In conclusion, our results have several important implications. Considerable time is taken in many sleep laboratories to record and score PLMS among patients evaluated for SDB, and an impressive 24% of our patients had PLMS frequent enough to satisfy criteria for PLMD. However, the utility of such knowledge remains to be demonstrated: in this study, we could not show a link between incidental PLMS and increased objective or subjective sleepiness. Rates of PLMS associated with arousals seem to provide no better prediction of sleepiness, and in fact weakly predict alertness, perhaps because patients who are more sleepy are not able to arouse when a PLM occurs. These data raise the possibility that unless the clinical history suggests clear symptoms of PLMS, neurodegenerative disorders, or restless legs syndrome, routine scoring of PLMS and PLM-arousals may be less helpful than commonly thought, especially if it leads to medication for an asymptomatic polysomnographic finding.