Novel assessment of microvascular changes in idiopathic restless legs syndrome (Willis–Ekbom disease)


CorrespondenceKirstie N. Anderson, MBBS, DPhil, Department of Neurology, Royal Victoria Infirmary, Queen Victoria Road, Newcastle upon Tyne NE1 4LP, UK. Tel.: +44(0)191 2823833; fax: +44(0)191 2825027; e-mail:


Many patients with restless legs syndrome (Willis–Ekbom disease) complain of burning sensations in their feet associated with the desire to move, such that they seek cooler environments. This pilot study aimed to characterise the microvascular skin changes in 12 patients with restless legs syndrome compared with 12 age- and sex-matched controls. Patients with moderate or severe restless legs syndrome and controls underwent detailed thermovascular assessment in a controlled temperature room at three different stages (normothermic phase 23 °C, hot phase 30 °C, cold phase 18 °C). Microvascular activity was recorded during all phases by bilateral great toe laser-Doppler flowmetry and also by whole-body thermography. Patient and control measurements were compared. The study protocol was well tolerated. Parameters extracted from the laser-Doppler flowmetry measurements were used to model a logistic function using binary logistic regression. This demonstrated a statistically significant difference between patients with restless legs syndrome and healthy controls (< 0.001). Visual inspection of the body thermography image sequences showed increased lower limb movement in patients with restless legs syndrome patients compared with controls. Thermography analysis also showed significant differences between foot temperatures in patients with restless legs syndrome compared with controls during the hot phase (= 0.011). Notably, patients with restless legs syndrome had more uniform foot temperatures, whereas controls had a wider variability in surface temperature across the feet. This novel study demonstrates impaired microvascular circulation in patients with restless legs syndrome in comparison to matched controls and a potential mechanism for the sensation of burning feet. The protocol also provides an experimental paradigm to test therapeutic interventions for the future.


Restless legs syndrome (RLS; Willis–Ekbom disease) is defined as dysaesthesia within the lower limbs associated with rest and to some extent relieved by movement. Although up to 25% of patients have symptoms during the day, there is characteristically a circadian rhythm with patients more severely affected in the evening and at sleep onset. Moderate or severe RLS has been shown to disrupt sleep. Standardised RLS diagnostic criteria were introduced and then revised by the International Restless Leg Study Group (Allen et al., 2003), and prevalence estimates range from 3 to 7% in large population studies (Allen et al., 2005). A unifying pathophysiology remains debated, as does the explanation for the clear circadian pattern that is typical. RLS can occur as an idiopathic disease or in association with other conditions, such as uraemia and small fibre peripheral neuropathy. Chronic venous insufficiency was proposed as a potential mechanism for disease by Karl Ekbom in his original disease description (Ekbom, 1945), but vascular abnormalities have been explored very little in RLS.

Idiopathic RLS sufferers describe a range of unpleasant sensations, but tend to complain of heat and burning more than other sensations when asked about symptoms in detail (Karroum et al., 2012). Patients have been shown to have enhanced perception of heat stimuli compared with controls (Edwards et al., 2011), and anecdotally patients often describe kicking off bedcovers or choosing to rest their feet on cooler surfaces. Thermal thresholds remain intact in idiopathic RLS patients (Bachmann et al., 2010), but detailed microvascular studies looking for abnormalities in thermoregulation have not been performed to date. This may be of particular relevance in RLS as there is known to be a clear circadian pattern to thermoregulation. In sleep our core body temperature reaches its nadir, and distal extremity vasodilatation associated with central cooling is strongly associated with sleep onset (Kräuchi, 2007). Therefore, we undertook a case–control study assessing the microvascular circulation in symptomatic, idiopathic RLS patients in comparison to controls at different ambient temperatures.

Materials and Methods

Patients and controls

All subjects had English as a first language and were able to give signed consent. Ethical favourable opinion was obtained from the Durham and Darlington Local Ethics Committee.

Twelve patients (four males; age 53.3 ± 14.5 years) with primary idiopathic RLS were recruited from the Regional Sleep Service, Newcastle. The diagnosis of RLS was made in accordance with the four revised essential criteria as published by the International Restless Legs Syndrome Study Group (Allen et al., 2003; Walters, 1995). All patients had a normal neurological examination with no evidence of peripheral neuropathy. All but one patient had previously had normal nerve conduction studies and electromyography to exclude peripheral neuropathy. All patients were otherwise well without co-morbidities that might affect peripheral circulation, and none used excess alcohol. One was a smoker of 10 cigarettes a day. All patients had a positive response to dopaminergic therapy and all patients had associated periodic limb movements of sleep demonstrated on overnight polysomnography, and in all cases patients rated symptoms as moderate or severe at the time of the study. In addition, 12 healthy age- and sex-matched controls (four males; age 51.7 ± 14.6 years) were recruited from healthy volunteers working within NuTH and Newcastle University. None of the controls had RLS, and all were free of medication that might affect quantitative sensory testing, or any relevant neurological or rheumatological diseases during the 6 months prior to any testing. Any prescription medication remained stable for a 4-week period before the study commenced. Patients had no alteration of any prescription medication for the 4 weeks prior to the study, all were taking dopamine agonists, and none was using levodopa or any other RLS therapies at the time of the study. All participants abstained from caffeine and alcohol on the day of testing.

Quantitative microvascular testing protocol

The study was undertaken within the thermal physiology suite within Medical Physics at the Freeman Hospital, Newcastle upon Tyne. The suite allows rapid and precise temperature control throughout the room to ensure precisely defined ambient temperatures. Bilateral great toe laser-Doppler flowmetry (LDF) waveforms, thermal imaging temperature data for hands, feet and whole-body thermography measurements were made during a range of room temperatures. Each study lasted approximately 2 h, and was carried out at the same time of day for all subjects (10:00–00:00 hours). LDF was performed using the Moor Instruments floLAB system, and thermal imaging measurements collected using a FLIR A40 camera (Flir Systems, Kent, UK) for whole-body imaging and a FLIR SC300 (Flir Systems) camera for close up views of the feet. The measurement stages are defined below.

Stage 1. An acclimatisation period of 20 min within the thermal physiology suite. Patients were within a comfortable normothermic environment with non-discernible draughts. The mean ± SD local temperature and humidity across all the 24 subjects for this acclimatization and following normothermic phase was 23.3 ± 0.4 °C and relative humidity (RH) 38 ± 10%, respectively.

Stage 2. Baseline testing (normothermic room temperature phase). Resting thermovascular measurements were performed for 10 min, comprising simultaneous LDF (bilaterally at great toe pads) and thermography (lower legs, hands and face). Patients underwent a brief assessment of RLS severity using the validated IRLSS rating scale to confirm current restless leg severity (on a visual analogue rating scale). During measurements the subjects were asked to keep their legs relaxed but still.

Stage 3. Dynamic testing (hot room temperature phase). The measurement suite ambient temperature was quickly warmed to 30 °C, and a repeat set of 10-min thermovascular measurements performed and the severity of RLS reassessed. The measured local temperature and humidity across all the subjects during this phase was 30.2 ± 0.5 °C and RH 28 ± 8%, respectively.

Stage 4. Dynamic testing (cold room temperature phase). The measurement suite ambient temperature was quickly cooled to 18 °C, and a repeat set of 10-min thermovascular measurements performed and the severity of RLS reassessed. The measured local temperature and humidity across all the subjects during this phase was 18.3 ± 0.7 °C and RH 39 ± 8%, respectively.

Quality control in thermography temperature measurements was made by referencing the thermal images to a portable calibrated Blackbody thermal reference source (LAND LandCAL P80P, set to 42 °C) held in the viewing region of interest.

Data analysis


Mean and standard deviation of the LDF waveform was calculated for each great toe of each subject in three different 10-min windows at the end of each phase (normothermic, hot, cold phases). Movement noise artefact was reduced with a median filter using a window of N = 100 data points. All analysis was then performed on the filtered signal.

Bilateral toe LDF waveforms were characterised in terms of amplitude and variability over time, considering values from each body side along with right–left average and difference. Absolute values during each phase and across-phase changes were investigated for differences between controls and patients. Correlations between right and left toe LDF waveforms were calculated for the normothermic, hot and cold room temperature phases, and these were tested for statistical significance utilising the non-parametric Mann–Whitney U-test.

A validated binary logistic regression approach (Hosmer and Lemeshow, 2000) was then used with parameters extracted from the LDF measurement to model the logistic function:

display math

where n represents the number of variables included. The following four variables (parameters defined on the difference of the two toe LDF waveforms) were used: (i) mean value during the hot phase; (ii) mean value during the cold phase; (iii) percentage relative variation of the mean value between the hot and cold phases; (iv) variation of the mean value divided by the difference in the local temperature between the hot and cold phases.


Thermal imaging data were calculated by computing the mean temperature and standard deviation over the imaged skin area for regions of hands, feet and face at each of the three temperature phases. Right and left sides were averaged for hands and feet as there was no significant difference between left and right in any subject. Group differences at each individual temperature phase and across-phase changes were investigated in terms of absolute values from each body site and segmental differences. The thermal gradients were also calculated with estimated core temperature measured using the established region of the inner canthus of the eye (SO/TR, 2009). Mean temperature and standard deviation were also calculated for each foot, for an area (manually drawn) including the dorsum, the lateral sides and the toes. An area including the fingers and the palm was used for the hands.


The RLS score was calculated at normal, hot and cold phases for healthy controls and patients with RLS. As expected, no controls had RLS symptoms at any stage of the thermovascular and microvascular testing median (LQ–UQ) 1.0. The RLS subjects all had RLS symptoms, but there was no significant difference between the cold and hot phases in patient estimates of their RLS severity. The IRLS scores for the patients were 1.0 (1.0–2.0) with normothermic room, 2.0 (1.0–2.5) with hot temperature, 2.0 (1.5–3.5) with cold temperature.

LDF measurements

Differences between controls and patients with RLS according to values computed by the logistic function were evaluated using the Mann–Whitney U-test. Comparing the values from the controls with the RLS group, a significant difference was obtained (P < 0.001; Fig. 1). Receiver operating characteristic analysis then gave an area under the curve of 0.96. The best clinical performance was obtained using a cut-off of 0.4. The clinical performance was thus: negative predictive value (NPV) = 100%; positive predictive value (PPV) = 92%; sensitivity = 100%; specificity = 92%; and accuracy = 96%.

Figure 1.

A comparison of the LDF blood flow signals between patients with restless legs syndrome (RLS) and healthy controls using the binary logistic regression model approach. The model with four parameters showed that the two subject groups could be clearly separated. The P-value is from the Mann–Whitney U-test. *Denotes outliers.


Foot thermography

There was no significant difference between values at the right and left foot for both the mean temperature or the standard deviation for either controls or patients with RLS, therefore they were averaged for all further comparisons. Studying the whole-body thermography, the mean whole-body temperature was not significantly different between healthy controls and patients with RLS either during the normal, the hot or the cold phase.

However, there was a clear difference between the RLS feet compared with controls in the hot phase. The standard deviation of the foot temperature became significantly lower in RLS feet compared with controls in the hot phase. There was no difference during the standard or cold phases. Therefore, during the hot phase, the feet of the patients with RLS had a more uniform temperature compared with controls where the feet were cooler than the core body temperature (Fig. 2). This difference in the standard deviation remained consistent also after normalisation to mean temperature (P = 0.05). Hence, control subjects showed a greater temperature gradient between hotter and cooler areas (Fig. 3a), while RLS patients had a more uniform feet temperature (Fig. 3b). The difference, however, was not present in the hands of the patients with RLS when compared with controls (P = 0.86).

Figure 2.

(a) The thermography area of interest (delimited by a carefully hand-drawn black contour) used in this analysis. (b) Temperature standard deviation of patients with restless legs syndrome (RLS) and healthy controls during baseline, hot and cold phases. The P-values (from Mann–Whitney U-test) show controls versus patients' statistical difference at each phase (NS, not significant). *Denotes outliers.

Figure 3.

(a) Thermography example for a healthy control during the hot phase of the measurement. (b) Thermography example for a patient with RLS during the hot phase of the measurement.

The power to discriminate between RLS and non-RLS subjects was further investigated by linear discriminant analysis, and the best separation of the two groups was found when plotting the mean temperature during the normal phase versus the standard deviation of the hot phase (Fig. 5a). This combination gave a sensitivity of 92% (11/12) and specificity of 83% (10/12) in spite of this being a pilot study and the numbers in each group relatively small. Also, it was visually evident from the videos that patients moved their limbs mainly during the hot phase, and all patients with RLS moved more than the controls.

Segmental differences in the standard deviation of the temperature were significantly different between RLS and controls during the hot phase and when considering the head–feet and hands–feet thermal gradients (Fig. 4). This again suggested a more uniform temperature distribution within the feet of the patients with RLS in comparison to control subjects. There was no significant difference between the RLS versus the control group during the cold phase. Using linear discriminant analysis, the best separation of RLS versus control data for the segmental hands–feet gave a sensitivity of 92% and a specificity of 83%, with overall accuracy of 88% (Fig. 5b). Also, NPV = 91% and PPV = 85%.

Figure 4.

Segmental differences in thermography. These show significant differences between groups for standard deviation measurements of temperature for the hot phase, suggesting a more uniform temperature distribution within the feet of patients with restless legs syndrome (RLS) patients compared with healthy controls. There was no significant difference found between subject groups for the cold phase of the measurement. The P-values (from Mann–Whitney U-test) show any statistical differences between phases (NS, not significant). *Denotes outliers.

Figure 5.

Linear discriminant classification showing the separation between subject groups. The classification areas use a grey scale to show the confidence in the result, with confidence estimated also using a binary logistic regression-derived method. (a) Measures only from foot thermography. (b) Segmental differences obtained from whole-body thermography are also considered. RLS, restless legs syndrome.


This study assessed the microvascular circulation using both LDF measurements and whole-body thermography to look for differences between patients with RLS in comparison to controls at a wide range of different ambient temperatures in the physiological range. These two techniques are well-established methods of quantifying blood flow within the dermal microvascular bed. LDF measures the blood flow flux [Red blood cell (RBC) velocity × concentration], and infrared thermography measures spatially the surface temperature to give information on thermoregulatory blood flow (Ring and Ammer, 2012). Given that the skin microvascular circulation is controlled by the sympathetic nervous system, it has been used in a number of conditions to look for localised sympathetic dysfunction. However, this is the first study to use thermography and LDF in patients with RLS.

In our case–control study, the healthy control subjects had cooler feet with a more variable pattern of cooling when the room ambient temperature was increased. In contrast, the feet of patients with RLS had more uniform foot temperatures, suggesting impaired spatial cooling at the higher temperatures. This difference was only present in the feet and not the hands, and was only present during the warm phase and not the cool phase.

Abnormal perception of temperature at the skin surface in patients with RLS has been demonstrated by a number of authors. Happe and colleagues demonstrated impaired cutaneous responses to stimuli of different temperatures using thermal stimulation (Happe and Zeitlhofer, 2003). Stiasny-Kolster et al. (2004) also demonstrated that patients with idiopathic RLS have a static mechanical hyperalgesia to pinprick stimuli, pointing to a central sensitisation or disinhibition of A-fibre-mediated pinprick perception. More recently, a combined approach using both skin biopsies and qualitative sensory testing demonstrated that secondary but not primary RLS patients had thermal hypoaesthesia to cold and hot stimuli alongside hyperalgesia to pinprick (Bachmann et al., 2010). Therefore, the discomfort in restless legs varies depending upon temperature.

Increased temperatures are recognised to exacerbate RLS symptoms for some patients, and self-help measures recommended to ease symptoms include sleeping in cool temperatures or using a fan to blow cool air onto the limbs (Chaudhuri and Muzerengi, 2009). Patients with RLS often report burning sensations or a feeling of heat, and seek to place their feet on cooler surfaces. In a case–control study with 31 patients with RLS and 18 controls, there was an enhanced temporal summation of heat but not to cooler temperatures (Edwards et al., 2011). In our study, controls and patients with RLS showed no significant difference in microvascular circulation at ambient temperatures or cooler temperatures, but patients with RLS were visibly more restless in association with clear microvascular changes compared with controls in the hot phase.

Restless legs syndrome is a disorder with both sensory and motor symptoms, and this may explain some of the difficulties in defining the precise pathophysiology, but a dopaminergic dysfunction in subcortical systems seems to play a fundamental role. Clemens et al. (2006) hypothesised that RLS was associated with a dysfunction of the dopaminergic A11 neurons within the subparafasicular thalamus. These neurons exert potent modulatory actions upon spinal networks, mainly via D3-like receptors, and this is postulated to shift descending control to an excitatory state with sympathetic overdrive and increased norepinephrine. This provides a model for both the increased locomotor activity during RLS and the circadian variability. Dopamine levels show circadian variation with a peak at night, and mouse knockout models with absent D3 receptors had abnormal reflex excitability and increased locomotor activity during rest when descending tonic inhibitory activity is at its greatest (Clemens and Hochman, 2004). The skin microvascular circulation is controlled by the sympathetic nervous system, so finding impaired microvascular circulation in patients with RLS in comparison to controls is consistent with sympathetic dysfunction in these patients.

The circadian variability of patients with RLS is striking and remains one of the key diagnostic criteria. The normal circadian variation in both core and distal skin temperature has been well described, and core body temperature steadily drops as the evening progresses, but in contrast there needs to be distal vasodilatation in comparison to the core temperature close to sleep onset (Kräuchi, 2007). Impairments in the skin microcirculation may therefore be linked to difficulty initiating sleep, which is a characteristic feature of more severe RLS. They may also provide a mechanistic explanation of the symptoms of burning feet.

There are several potential criticisms of the data as presented, the sample sizes are small, and any assessment of the sensitivity and specificity of the techniques described to distinguish between patients with RLS and controls must be interpreted with caution. The binary logistic model used for regression analysis of the LDF is validated for use with multiple biomedical variables but has not been used in LDF measurements before, so again a larger sample size would be helpful to help to further validate this assessment. It may be that thermography alone is sufficient to detect microvascular changes between patients and controls. Another potential criticism is that the length of the assessments may have contributed to increasing patient discomfort and RLS symptoms; however, if the responses seen were simply a measure of the duration of the study then one might have expected progressive increase in severity of symptoms over time irrespective of the temperature of the room, and this was not the case. Patients were more symptomatic in the earlier phase when the room temperature was hot and this phase showed the greatest difference in thermography, so we do not think study duration skewed results.

One further issue is the possible effect of medication on thermovascular responses. There are no microvascular effects of dopamine agonists described in the human literature, but it is possible that medication itself contributes to some of the differences seen.

Previous studies to assess skin responses or sympathetic function require painful stimuli, but the technique described has the advantage of being non-invasive and well tolerated by patients and controls. Further larger studies beyond this pilot study are required to validate these findings, but these data provide potential support for sympathetic dysfunction in RLS, and the microvascular techniques of thermography and LDF form a well-tolerated experimental paradigm to visually assess patients with RLS and possibly enable future therapies to be trialled.

In summary, patients with RLS have been shown to have abnormal microvascular circulation with impaired cooling at higher temperatures in comparison to controls. This provides a potential mechanism for the sensation of burning feet and increased discomfort at higher ambient temperatures.


We are grateful to all the patients who took part in the study, and to Phillips R&D for financial support. Jennifer Caffarel within Phillips provided helpful discussion.

Authors' Roles

K.A.: study concept and design, recruitment to study, obtaining funding and drafting and revision of manuscript; C.Di.M.: analysis of data, statistical analysis, critical revision of manuscript; J.A.: study conception and design, acquisition of data, analysis of data, critical revision of manuscript.

Conflict of Interest

This work was supported by an unrestricted grant from Philips R&D, there are no conflicts of interest. All authors were full-time employees of the Newcastle upon Tyne Hospitals NHS Foundation Trust for the preceding 12 months with no other financial disclosures.