Moderate to severe but not mild RLS is associated with greater sleep-related sympathetic autonomic activation than healthy adults without RLS.

Abstract

Introduction:

Restless legs syndrome (RLS) patients have been found to have high rates of transitory increases in the activity of the sympathetic autonomic nervous system with increases in heart rate and blood pressure. These were identified by evaluating heart rate or blood pressure changes independent of any leg movement analyses. There has been an implicit assumption this high rate of sympathetic activations is abnormal, but there has been no direct comparison for similar measures with a healthy population free of RLS. Thus, it is not known if during sleep the rates of sympathetic nervous system activation are greater for RLS than for a healthy population. The objectives of this study were to determine if: (1) RLS patients compared to healthy controls have a greater frequency of sympathetic nervous system activation (significant heart rate increases) with a higher percentage of leg movements associated with these activations; (2) the sympathetic activation frequency and its relation to leg movements correlate significantly with RLS severity in RLS patients; and (3) there is some minimum threshold for RLS severity defining an RLS population where most (e.g. 85%) have abnormally high rates of sympathetic activation.

Methods:

Sleep data on 32 RLS patients and 21 matched healthy controls were obtained from a prior study. All leg movements during sleep (LMS) and periodic leg movements in sleep (PLMS) were identified following the new WASM criteria; LMS that were not PLMS were considered isolated leg movements in sleep (ILMS). All episodes with significant heart rate increases were identified following procedures established by Cassel et al, 2016, i.e. a slope of linear regression ≥ 2.5 beats per minute over 5 consecutive heartbeats. Severity of RLS was evaluated using the International Restless Legs Study Group Scale (IRLS).

Results:

RLS patients had significantly more heart rate increases than controls (67.88/hr vs. 9.87/hr). RLS patients had a significantly greater percentage of both LMS and PLMS occurring with heart rate increases than controls (44% vs. 30%; 48% vs. 18%, respectively). These measures correlated significantly with IRLS and also PLMS/hr. 85% of the RLS patients with IRLS scores > 22 or PLMS > 50/hr had rates of sympathetic activation that were > 90^th percentile for the healthy controls.

Conclusion:

This is the first paper documenting that RLS patients showed clearly increased sympathetic activation when identified independent of PLMS. This, however, occurs for more severe RLS and not milder RLS. It has been proposed that the abnormally high rate of sympathetic activation for RLS patients relates to development of adverse cardiovascular health consequences observed in some studies. Thus, these data may provide a basic standard for possible use in epidemiological studies to identify the level of RLS severity more likely to have adverse health consequences (e.g. cardiovascular disease). Since 2/3^rd of RLS patients have mild to even intermittent disease, including all RLS is likely to miss the potential health consequences of RLS.

Introduction:

Arousal events during sleep likely represent “preparation in anticipation of rapid assumption of the upright posture and a fight or flight reaction” [1]. Arousal events disrupt sleep processes and exhibit 3 different features, each supporting the fight or flight response: (1) cortical arousal with EEG changes toward wake state; (2) metabolic and sympathetic autonomic arousal preparing for activity, e.g. transitory increases in heart rate and blood pressure; (3) movement arousal, e.g. leg movements in sleep (LMS) preparing for movement by activating extensive motor systems supporting human’s unique biped movement, such as flexion at the knee, dorsiflexion of the foot, extension of the big toe, and slight inversion of the foot, which are characteristics of a step and of leg movements in sleep (LMS) [2]. Note that the sleep arousal processes as defined here are not limited to those events with cortical EEG changes but include neurophysiological arousal events producing any of these three markers. These arousal events fragment sleep and, when unusually frequent, have been associated with increased risk of adverse health events [3–8]. The long-term health consequences of fragmented sleep have been considered to result from multiple metabolic effects of disrupting sleep processes [9], including abnormally frequent arousals producing sympathetic activation with transitory but relatively large increases in heart rate and blood pressure [10, 11].

While the three features of sleep arousal events tend to co-occur, this is certainly not always the case. For example, leg movements in sleep (LMS) have, on average, significant sympathetic activation with heart rate and blood pressure increases [12–16], but this may not be true for all LMS. Similarly, not every significant heart rate increase in sleep occurs with LMS [10]. Each of the three features of the arousals during sleep also occurs with differing degrees of intensity. The more intense arousal events, however, appear to occur with all three features present [17].

In disease conditions associated with increased arousal during sleep, such as restless legs syndrome (RLS), [18] arousal events would be expected to occur not only more commonly but also more intensely and therefore more often together, presumably in relation to disease severity. Motor activity such as sleep leg movements (LMS) and sympathetic activations such as transitory heart rate increases (HRup) should therefore be increased in RLS in relation to disease severity. Moreover, while not all LMS will occur with HRup, the percentage of LMS with HRup should also be increased in RLS patients in relation to disease severity. A presumably high rate of HRup of 50–60/hr in sleep has been documented for the RLS population [10], but it is actually not known if this is a high rate since no similar data have been reported for a comparable healthy adult population. It is important to note that these HRup episodes during sleep are determined independent ly of any movement or LMS evaluation. Thus, it is not known if RLS patients compared to matched healthy adults actually have either (1) a higher rate of arousal producing more episodic cardiac activation (HRup) or (2) a higher percentage of HRup occurring with LMS, indicating greater intensity of the underlying arousal processes. It is also not known if the HRup events not related to LMS are greater for RLS than healthy controls. Does the increased sympathetic activation occur for RLS even when there is no associated leg movement or is it related only to the increased leg-movement associated arousals? Moreover, it is similarly not known if occurrence of these arousal-related events relate to RLS severity. This may occur for all RLS patients or there may be some mild RLS that has limited increase in rates of transitory cardiac activation compared to healthy normals. Thus, defining the level of RLS severity that identifies those with significantly increased sympathetic arousal becomes important for evaluating potential health risks associated with RLS. Answering these important questions requires a study identifying the cardiac activation events during sleep and comparing these for RLS patients vs. matched healthy controls who have no indication of RLS. This basic comparison has not been previously reported.

This study tested two major hypotheses indicating increased arousal events for RLS, i.e. RLS compared to healthy controls free of RLS would have (1) higher rate of HRup per hour of sleep (HRup/hr Sleep) and (2) more intense arousals indicated by greater percentage of LMS/hr with HRup/hr (%LMS w/ HRup) (Hypotheses 1 and 2). A corollary to hypothesis 2 was also tested to determine if the rate of HRup without LMS was greater for RLS. If the two basic hypotheses were confirmed, then it was further hypothesized that these findings would relate to RLS severity, i.e. HRup/hr sleep and the %LMS w/ HRup would correlate significantly with the severity of RLS. (Hypotheses 3 and 4). If these correlations are significant, then the regression relations can be reviewed to determine a potential critical RLS severity level above which most (e.g. 85% or more) of the patients would have HRup rates ≥ 90^th percentile of these rates observed for the healthy normals.

Exploratory analyses included asking similar questions for periodic leg movements per hour of sleep (PLMS/hr) and the non-periodic, isolated leg movements per hour of sleep (ILMS/hr).

Methods:

Subjects:

32 RLS and 21 controls included in this study (Table 1) were all participants in a study on transcranial magnetic stimulation [19]. The study was approved by the Johns Hopkins Institutional Review Board. The RLS patients were diagnosed by a validated structured diagnostic interview (HTDI) [20], and the diagnoses were further confirmed by an RLS expert’s clinical interview. The RLS subjects were gradually tapered off any RLS medication and had to be off all RLS medications for at least 12 days prior to two consecutive nights of sleep studies. Healthy control subjects were similarly evaluated with the HTDI and clinical interview for no indication of any symptoms like RLS and for no family history of RLS. All subjects wore leg activity meters placed on the ankle (PAM-RL Phillips Respironics) [21] measuring PLM in bed for each of five consecutive nights before starting the study. Subjects were excluded from this study if either the average PLM over the 5 nights of home recording or the PLM/hr of sleep on the 2^nd night polysomnogram (PSG) was ≥ 15 for control subjects or ≤ 15 for RLS patients. PLMS rates ≥ 15/hr have been associated with a family history of RLS and therefore to increase confidence of no unrecognized familial RLS, controls were required to have PLM ≤ 15/hr [22]. This served to more clearly separate the controls from the RLS disorder including reducing risks of significant genetic overlaps with RLS as seen for subjects with PLMS [23, 24]. Patients with RLS clearly related to or significantly exacerbated by other medical conditions were excluded from the study. Both RLS and controls had to be healthy and free of any significant disease or medication use that would disturb sleep or either exacerbate or engender RLS symptoms. The subjects with apnea/hypopnea rates >10/hr on the 1^st night’s PSG recording were excluded from the study.

Table 1:

Subject Characteristics

	Control	RLS	t test prob
Sample size	21	32	----
age	56.3 ± 8.2	59.0 ± 9.3	0.29
gender (Males/Total %)	23.8	43.8	0.06*
Total sleep time (hr)	6.76±0.53	5.74±1.66	<0.005
LMS/hr sleep	18.65±32.70	110.41±60.05	<0.0001
PLMS/hr sleep	0.78±1.02	87.81±55.48	<0.0001
ILMS/hr sleep	17.20±32.67	22.60±31.91	0.55
RLS severity (IRLS)**	----	25.9±6.3	----
Correl. PLMS/hr to IRLS**	----	0.38	<0.025

LMS= leg movement in sleep, PLMS= periodic leg movement in sleep ILMS= LMS not a PLMS

Data are averages ± st, dev. except where specified otherwise.

^*p value for difference in male distribution in RLS and control was calculated using χ² value= 3.44 and p value = 0.06

^**RLS severity off RLS medication was not obtained for one RLS patient.

RLS clinical severity:

RLS severity was determined using the International Restless Legs Syndrome Study Group severity scale (IRLS). The IRLS severity rating was obtained for RLS patients following standard procedures [25] the day after the 2^nd overnight sleep study, which was at least 12 days off any RLS medication when most of the withdrawal symptoms had abated [26].

Sleep Studies:

All subjects accepted in the study had polysomnogram studies for two consecutive nights in the sleep lab with 8 hours in bed. The studies followed AASM standards [27] with channels recording EOG (one for each eye), EMG (one for each anterior tibialis), EMG chin muscle, EEG (6 channels for frontal, central and occipital recordings from each side of the head referenced to the ipsilateral mastoid), finger oximetry, chest and abdominal respiratory bands, and nasal pressure. The respiratory and oximetry measures were not obtained on the 2^nd night. The data in this study came from the 2^nd night.

Leg movements in sleep (LMS):

LMS from the sleep studies were scored using a validated [28] MATLAB program adjusted to meet the current WASM criteria [29] with a final visual review of all events by an RLS expert. All candidate leg movements that were not periodic leg movements in sleep (PLMS) were considered isolated leg movements in sleep (ILMS).

Heart rate measurements:

The episodes of significant transitory increased heart rate (HRup) were determined as in prior studies [13] [10]. A significant HRup event was defined by a slope of linear regression ≥ 2.5 beats per minute over 5 consecutive heartbeats. When this occurred an HRup episode was defined as a 20-second window divided into 7 seconds before the HRup event onset and 13 seconds after the onset. Testing for a HRup event then started again evaluating the first 5 consecutive heartbeats after the end of the 20-second window. When there was no HRup event the analyses would shift forward by one heartbeat, i.e. the 2^nd heartbeat of the previous analysis would be the 1^st heartbeat of the new analysis. This would be repeated until a HRup episode was found or the record ended.

A LMS was considered associated with a HRup episode if the LMS occurred within the first 7 seconds of the 20-second window as was done in the prior studies [13] [10].

Statistics:

Data were checked visually for significant differences from normal distribution or for indications of significant outliers. The primary outcomes were the HRup/hr of sleep and the %LMS with HRup, %PLMS with HRup, and %ILMS with HRup. These were calculated for each subject, and then differences between RLS and control subjects were tested using t tests adjusted for differences in the variances (F test). Correlations were evaluated using the Pearson-Product moment correlation. The HRup rates were evaluated for mean, standard deviations and also medians, ranges and the 90^th percentile.

The primary hypotheses were evaluated sequentially with hierarchical analyses in order of the hypotheses. Testing was discontinued after a preceding test failed significance with p > 0.05. (Table 2)

Table 2:

Primary Hypotheses Tested in Order (average ± st. dev.)

	Control n=21	RLS n=32	T test prob
hypothesis 1
HRup/hr_sleep (avr±sd)	9.87±10.53	67.88±48.65	<0.0001
hypothesis 2
%LMS w/ HRup (avr±sd)	30 ± 27%	44 ± 24%	<0.025
Residual to hypotheses 2*
(HRup no LMS)/hr sleep (avr ±sd)	7.43 ±9.65	14.55 ± 17.7	<0.04
	correlation	r value
hypothesis 3
correlation between HR-up/hr and RLS severity**		0.42	<0.025
hypothesis 4
correlation between %LMS w/ HRup and RLS severity**		0.37	<0.025

^*Residual – the HRup events remaining after removing those related to an LMS. (HRup not related to LMS as % of all HRup= about 75% for controls and 21% for RLS)

^**RLS severity off RLS medication was not obtained for one RLS patient, The sample size here is 31.

Note that the t-tests were for the directional hypotheses.

Results:

Population characteristics:

There were no significant differences between the control and RLS populations except for those expected for RLS, i.e. shorter sleep time, more LMS/hr and PLMS/hr of sleep. There was somewhat unexpectedly no difference in density of isolated LM in sleep (ILMS/hr). There were fewer males in the control population than for RLS but the difference was not significant (24% vs. 44%, p =0.06). (Table 1) The IRLS severity score was not obtained the day after the 2^nd sleep study for one patient. The remaining 31 RLS patients showed a wide variation on the IRLS from mild to very severe (range: 12 – 36; avr ± sd: 25.9 ±6.3). (Table 1)

Primary hypotheses:

There were significantly more HRup/hr for RLS than for controls (Avr ± sd: 67.88 ± 48.65 vs 9.87±10.53 respectively, p < 0.0001) and a significantly higher %LMS w/ HRup for RLS than for controls (avr ± sd: 44±24% vs 30±27% respectively, p < 0.025), thus confirming major hypotheses 1 and 2 (Table 2). The residual to hypothesis 2 that the rate/hr sleep of HRup was not associated with LMS was also confirmed for RLS > controls (avr ±sd : 14.6 ± 17.7 vs. 7.4±9.7 respectively). Both HRup/hr of sleep and %LMS w/ HRup displayed significant correlations with RLS severity (r = 0.42, 0.37 and p < 0.025, < 0.025, respectively), confirming hypotheses 3 and 4. (Table 2; Figure 1). Separate analyses for females were done to check for possible gender bias given the somewhat smaller number of male controls. This produced the same results as the total sample except the correlations were not significant. Thus, the analyses for females with reduced sample sizes confirmed the 1^st two hypotheses but not the correlations in hypotheses 3 & 4. (see supplemental tables).

Figure 1A(upper panel):

Episodes of significant heart rate increase (HRup) per hour of sleep vs. RLS severity on the IRLS scale at 12 days or more off RLS treatment. The red line is the 90^th percentile for HRup/hr of the healthy controls without PLMS or family history of RLS (see Figure 3-Left). RLS patients with IRLS > 22 tend to persistently exceed the red line. Figure1B (lower panel): Percentage of leg movements in sleep (LMS) with HRup vs. RLS severity on the IRLS scale at 12 days or more off RLS treatment.

The 90^th percentile of the HRup/hr of sleep for normals was 27/hr (Figure 4). The HRup rates were < 27/hr for 50% (5/10) of the patients with mild RLS (IRLS ≤22) but only 14% (3/21) of those with moderate to severe RLS (IRLS >22), (Figure 1A). There was no significant gender effect with HRup <27 for 3 of 18 females and 5 of 14 males. HRup/hr ≥ 27/hr (90^th percentile for healthy controls) occurred for 86% of the patients with IRLS > 22. (See supplemental fig. 1. Note the IRLS score for one patient was not obtained).

Figure 4:

Episodes of heart rate up per hour of sleep vs. periodic leg movement in sleep per hour in RLS patients. The red line indicates the 90^th percentile of number of HR up/hr of sleep for controls. The RLS patients with higher rates of PLMS (PLMS/hr ≥ 40–55) tend to persistently exceed the red line.

Exploratory Hypotheses: The %PLMS w/ HRup was significantly higher in RLS than in controls (p < 0.001), but %ILMS w/ HRup showed no major difference between RLS and controls (p = 0.82). (Table 3) Similarly, for RLS patients, both the %PLMS w/ HRup and PLMS/hr had significant (p<0.05) correlations with RLS severity (r= 0.31, 0.38, respectively), whereas %ILMS w/ HRup and ILMS/hr had no significant correlation with RLS severity (r=0.11, −0.28). (Table 3, Figure 2). The PLMS/hr correlation with HRup/hr showed that 50% (5/10) of the patients with PLMS/hr < 50 had HRup/hr ≤ 27, the 90^th percentile for the healthy controls. The patients with PLMS/hr ≥ 50 in contrast, had only 14% (3/ 22) with HRup/hr less than the 90^th percentile for the healthy controls. HRup/hr ≥ 90^th percentile of healthy controls occurred for 86% of the patients with PLMS/hr ≥ 50 (See supplemental fig 2)

Table 3:

Exploratory Hypotheses Tested (means ± sd)

	Control n=21	RLS n=32	t test prob
%PLMS w/ HR-up	18 ± 28%	48 ± 26%	<0.001
correlation with RLS severity (R)*		0.31	<0.05
PLMS/hr	0.78±1.02	87.81±55.48	<0.0001
correlation with RLS severity (R)*		0.38	<0.025
%iLMS w.HR-up	30 ± 23%	31 ± 18%	0.82
correlation with RLS severity (R)*		0.11	>0.05
ILMS/hr	17.20±32.67	22.60±31.91	0.15
correlation with RLS severity (R)*		−0.28	>0.05

^*RLS severity off RLS medications was not obtained for one RLS patient.

The samples size for the correlation = 31

Discussion:

The results confirmed our primary hypotheses. RLS patients compared to healthy matched controls had six times (68 vs 10) more episodes of heart rate increase per hour of sleep, indicating profoundly increased sympathetic autonomic activation. The rate of sympathetic activation events as indicated by heart rate increases is relatively low for these healthy controls without PLMS, occurring about once every 6.0 minutes, whereas the rate for RLS occurred more than once a minute. (Table 2) Not only are there more sympathetic activations for RLS, but the leg movement events are almost 1.5 times more likely to occur with an sympathetic activation for RLS than for controls (44% vs 30% of the LMS events). (Table 2) This is also true for only the LMS that are PLMS, for which the sympathetic activation with PLMS is over 2 times more likely to occur for RLS than controls (48% vs 18% of the PLMS events). (Table 3) Thus, the sleep arousal process associated with a PLMS is both more common and more intense, i.e. more likely to also occur with a significant sympathetic activation for RLS patients than for healthy controls.

One weakness of this study was the slight imbalance between genders for the control sample. The separate analyses for females only showed essentially the same primary results as for the total sample except the hypotheses based on correlations were not confirmed. This was partly due to a much reduced sample size below that required for a good representation of the entire population. The small sample size for males, particularly control males (n=5), limited meaningful statistical analyses. Thus, at this point, there is no clear indication of a significant gender bias for these results. This was further strengthened by the lack of a significant gender bias in the number of HRup/hr < 90^th percentile of controls.

The strong relations of both RLS severity and PLM density to the degree of sympathetic activation deserve a special consideration. It has been suggested that very frequent sympathetic activations contribute to the development of cardiovascular disease and high blood pressure that has been associated with RLS [11]. These data would suggest that, if this is the case, there could be an association between cardiovascular disease development and RLS patients with IRLS > 22 or with PLMS/hr ≥ 50. Studies on the health consequences of RLS should therefore probably focus on more severe RLS or on those with higher PLMS/hr. One epidemiological study has shown this effect: RLS relation to poor health outcomes was stronger in that study for subjects who had more severe RLS, as indicated by symptoms daily vs.1–6 days a week [30].

Duration of disease has also been noted to be a factor for cardiovascular health consequences of RLS in women [31]. Another study reported cerebral small vessel disease was worse for RLS than controls for RLS patients with > 10 year history of RLS but not those with RLS for < 10 years [32]. Presumably, the patients with longer disease had a longer exposure to the RLS-related risk factors for disease.

Overall, at this point there are now at least two major factors to consider when evaluating the relation of RLS to adverse health, particularly cardiovascular health consequences: the disease severity and the duration of active disease. It is important to note that only about 1/3^rd of all RLS patients report at least moderate symptoms while almost 2/3^rd report mild to minimal, even intermittent symptoms [33]. so the data in this study would indicate that epidemiological studies that fail to consider RLS severity will likely include mostly RLS without significantly high rates of sympathetic arousals and thereby fail to detect adverse health consequences of the disorder, as seen in two recent large epidemiological studies [34, 35]. Given the broad spectrum of RLS disease severity and the relatively low percentage of all RLS with severe disease, it is important that future work takes into consideration disease severity and duration.

It also deserves note that the PLMS/hr rates were high in this particular patient population even for the mild RLS. (Figure 2A) Nonetheless, most mild RLS patients had about the same percentage of the PLMS with cardiac activation as did controls. (Table 3; Figure 2A) This raises some questions about whether or not PLMS alone suffice for adverse health consequences. They may have to be part of a larger arousal process that also includes leg movements occurring with other markers of arousal, particularly cortical arousal. Identifying the LMS associated with cortical arousal may require determining the LMS features related to such arousals. Future studies on health consequences of PLMS and of RLS need to consider RLS severity in addition to just the diagnosis and possibly also consider LMS features rather than just counting PLMS.

Figure 2A (upper panel):

Percentage of periodic leg movements in sleep with heart rate increase (%PLMS w/ HRup) vs RLS severity on the IRLS scale at 12 days off RLS treatment. Figure 2B (lower panel): Percentage of isolated leg movements in sleep with heart rate increase (%ILMS w/HRup) vs. RLS severity on the IRLS scale at 12 days off RLS treatment.

Curiously, the sleep arousal associated with isolated leg movements in sleep (ILMS) was not associated with the increased sympathetic arousal for RLS compared to controls. (Figure 2B) There were neither significantly more ILMS/hr nor a higher percentage of ILMS with sympathetic arousal for RLS compared to controls. This suggests that the sleep arousal process producing ILM may be similar for RLS and controls while, in contrast, that producing PLMS appears to differ. The healthy controls in this study were selected to exclude any with significant PLMS, and this may have effectively removed any significant PLMS with associated arousal process. Healthy controls who have PLMS may have the same relation of PLMS to heart rate increase as that seen here for RLS patients. The healthy controls with high PLMS tended to be older subjects, which may reflect an age-related increase in PLMS. However, one study showed that this age-related increase in PLMS occurred mostly for patients with a family history of RLS and not for older adults without RLS in their family [22]. Moreover, the genetics of RLS overlap those associated with more severe PLMS [23]. Thus, healthy controls with higher rates of PLMS/hr (e.g. ≥ 50) may have both genetic and sleep disruption increasing risks of adverse health outcomes, particularly cardiovascular disease.

Conclusions:

These data demonstrated that RLS patients compared to healthy controls without PLMS or a family history of RLS showed markedly increased sympathetic autonomic activation. The increased rate of sympathetic activation occurred both without any association to LMS and with association to PLMS, but not with association to ILMS. Moreover, this arousal process tended to exceed the 90^th percentile of that observed for healthy controls mostly for moderate and severe RLS (IRLS ≥ 22) and those with PLMS/hr ≥ 50. Future epidemiological studies should focus on health consequences of these more severe RLS patients and possibly adults without RLS who have these high rates PLMS/hr. The sample considered at risk for adverse health consequences could be further refined by identifying the LMS with significant cardiac or cortical arousal based on the features of the leg movements.

This study also raises the need for better technology to measure leg movements, which would allow for more accurate characterization of significant features of LMS and PLMS in healthy normal controls, RLS patients, and even those with family history of RLS. This could help better identify the LMS associated with significant arousal and the relation of both PLMS and RLS to health consequences.

Figure 2C:

PLMS/hr vs. RLS severity on the IRLS scale at 12 days off RLS treatment. Figure 2D: ILMS/hr vs. RLS severity on the IRLS scale at 12 days off RLS treatment.

Figure 3-Left:

Box plot of distribution of heart rate increase per hour of sleep in healthy controls. The 90^th percentile of the box plot (27) was used to draw the red line in Figure 1A and 4. Figure 3-Right: Box plot of distribution of heart rate increase per hour of sleep in RLS patients. Note differences in units on the y axis. 90^th percentile for controls =27 which is about 25^th percentile for RLS patients.

Highlights.

• RLS has 6-fold more episodic heart rate increases than controls in sleep.

• The % of leg movements with increased heart rate is greater for RLS than controls

• No difference between RLS and controls was seen for isolated leg movements in sleep

• Increased autonomic arousal occurred mainly for moderate to severe but not mild RLS