Four female RLS patients, with a PLM index >5 on a polysomnogram (PSG), were included. The diagnosis of RLS was made based on the diagnostic criteria of the National Institutes of Health workshop on RLS.15 A validated Korean-language version of the John Hopkins telephone diagnostic questionnaire16 was administered in a face-to-face interview to exclude conditions that may mimic RLS. RLS severity was determined using the International Restless Legs Scale.17 Four age-matched female subjects who had no sleep disturbance or neurological or psychiatric disorders were included as healthy controls. All of the subjects provided written informed consent, and the experimental protocol was approved by the Institutional Review Board of our local institute.

PSG recordings were made in all of the participants (controls and RLS patients with PLMS) over a single night in our sleep laboratory. The overnight PSG was recorded using 19 electrodes placed on the scalp (in accordance with the international 10-20 system) for a two-channel electrooculogram, an electromyogram of the submental and anterior tibialis muscles, and an electrocardiogram that used surface electrodes. The electroencephalogram signals were sampled at 200 Hz. A thermistor (to monitor nasal and mouth breathing), cannula (to monitor nasal air pressure), oximeter (to measure oxygen saturation), piezoelectric bands (to determine thoracic and abdominal wall motions), body position sensor, and snoring sensor were also attached. Patients were recorded on videotape using an infrared video camera, and were observed continuously by a PSG technician. Sleep stage was scored in 30-s epochs according to the standard criteria listed in the American Academy of Sleep Medicine sleep scoring manual. Light sleep (LS) was defined as either stage-N1 or -N2 sleep, and slow-wave sleep (SWS) was defined as stage-N3 sleep. PLM were scored according to the American Academy of Sleep Medicine criteria.18

NIRS data were acquired simultaneously with the PSG data. A custom-manufactured wireless NIRS system was used to measure the cerebral hemodynamics responses.19 This system comprises six LED light sources (three 780-nm sources for detecting [Hb] and three 850-nm sources for detecting [HbO]) and eight silicon photodiode detectors. The light-intensity signals of only six user-defined channels were measured; two channels were selected on the forehead of the subjects to measure the hemodynamics on the left and right sides. The source and detector of the optical probe were separated by 3 cm. The six optical probe sources were multiplexed with an onset time of 10 ms per LED. The sampling frequency was set at 14 Hz.

We calculated the changes in [HbO] and [Hb] from the light-intensity data based on the following modified version of Beer-Lambert Law:20 where ΔOD is the change in optical density, λ is the wavelength, L is the distance between the source and detector, ɛ^λ _HbO2 and ɛ^λ _Hbr are the extinction coefficients of HbO and Hb, respectively, and B is the differential path length factor (DPF), which compensates for increases in the light path length caused by light scattering in biological tissues. We used DPF values of 6.625 and 6.4 for light at 780 and 850 nm, respectively, which were interpolated from previously reported wavelength-dependent DPF values.21 From Equation 1, ΔHbO and ΔHb can be derived using the following equations: The optical density signals were de-trended by third-order polynomial fitting to remove signal drifts not caused by actual changes in the hemodynamics.22 A finite-impulse-response bandpass filter based on the Parks-McClellan algorithm was applied to extract the LFO bands of the hemodynamics responses (i.e., from 0.04 to 0.12 Hz). The measured amplitudes were normalized by dividing by the maximum value for each subject.

We calculated the instantaneous phase difference between oscillations of [Hb] and [HbO] by subtracting the instantaneous phase of [HbO] from the instantaneous phase of [Hb] LFOs. The instantaneous phases of both [Hb] and [HbO] were computed by a complex extension of the original signal s(t) generated by the Hilbert transform:13 where s is the original signal, and š is the Hilbert-transformed original signal. The analytic signal, which represents a complex extension of the original signal, is defined as follows: Analytic signal a(t) is a complex signal whose real part is the original signal and imaginary part is its Hilbert transform. E(t) is the instantaneous amplitude of the analytic signal and θ(t) is its instantaneous phase. We analyzed the phase differences of LFOs only when the instantaneous amplitudes of the analytic signals-as calculated using the oscillations of ΔOD at 780 and 850 nm-were both above a threshold of 2×10^-3 cm^-1. This threshold was chosen based on the noise values obtained using a tissue-like phantom. We did not analyze data segments that included motion artifacts or signal processing noise, since they can cause a large amount of distortion. All 30-s NIRS data epochs without artifacts were selected, with amplitude and phase differences averaged across each sleep stage from each subject. In RLS patients with PLMS, LS stages containing PLM sequences were included as PLM epochs. We included all PLM sequences regardless of the arousal state, because the arousal events associated with PLM were irregular and had short durations. The number of epochs considered for data processing for each of the eight subjects is reported in Supplementary Table 1 in the online-only Data Supplement.

The Mann-Whitney U test was used to compare clinical and PSG characteristics between RLS patients with PLMS and controls. To evaluate changes in the cerebral hemodynamics during the sleep stages, NIRS LFO signals during wakefulness, LS (i.e., LS without PLM in RLS patients with PLMS), SWS, and rapid eye movement (REM) sleep were compared using the Kruskal-Wallis test. In RLS patients with PLMS, we compared the amplitude and phase of [Hb] and [HbO] between LS with and without PLM using the Wilcoxon signed-rank test in order to evaluate hemodynamics changes during LS with PLM. A probability value of 0.05 was considered to be indicative of statistical significance. All of the analyses were performed using the SPSS for Windows software package (version 12.0, SPSS Inc., Chicago, IL, USA).

The clinical and PSG characteristics in each group are summarized in Table 1. The median age of the RLS patients with PLMS was 53 years (range: 52-56 years), compared to 52 years (range: 43-58 years) for the normal controls. All of the subjects were female. The age, body mass index, total sleep time, sleep latency, wakefulness after sleep onset, sleep efficiency, proportion of time spent in each sleep stage, arousal index, and apnea-hypopnea index did not differ significantly between the groups. RLS patients with PLMS exhibited a median PLM index score of 14.2/h and a median PLM arousal index score of 4.6/h, which were both higher than in the normal controls (p=0.018 and p=0.019, respectively).

Fig. 1 displays the amplitude and phase differences of the measured [HbO] and [Hb] LFOs during the different sleep stages [wakefulness, LS (LS without PLM in RLS patients with PLMS), SWS, and REM sleep]. Overall, a progressive decreasing trend in the median amplitude of [HbO] and [Hb] LFOs was observed between wakefulness and SWS in both the patient and control groups. In contrast, there was a trend toward a progressive increase in the median phase difference between [HbO] and [Hb] LFOs between wakefulness and deeper sleep stages in both groups. The [HbO] and [Hb] LFO phase differences differed significantly with the sleep stage in the normal control group (p=0.033); however, there was no significant difference in the group of RLS patients with PLMS.

The [HbO] and [Hb] LFO amplitude and phase differences were compared between the patient and control groups according to sleep stage. Significant differences in the amplitude of [HbO] and [Hb] LFOs were observed between the patients and controls during REM sleep. The amplitude of [HbO] LFOs measured in the right frontal area was higher in RLS patients with PLMS than in controls during REM sleep (median amplitude=0.59 vs. 0.37, p=0.043). In contrast, the amplitude of [Hb] LFOs was higher in the left frontal area in RLS patients with PLMS than in controls (median amplitude=0.55 vs. 0.13, p=0.021). No differences between the two groups in [HbO] and [Hb] LFO amplitudes were found during LS (LS without PLM in RLS patients with PLMS) and SWS. There were no significant differences in [HbO] and [Hb] LFO phase differences between the RLS patients with PLMS and controls during any sleep stage (Supplementary Table 2 in the online-only Data Supplement). We compared [HbO] and [Hb] LFO amplitude and phase differences between RLS patients with PLMS during LS with PLM and normal controls during LS. RLS patients with PLMS exhibited higher [HbO] LFO amplitudes in the right frontal area (median amplitude=0.77 vs. 0.33, p=0.043) and higher [Hb] LFOs in the left frontal area (median amplitude=0.67 vs. 0.15, p=0.043) compared to controls with no significant PLM. However, there were no significant phase differences between the two groups (Supplementary Table 3 in the online-only Data Supplement).

To assess the effect of PLM during LS in patients with RLS and PLMS, we compared [HbO] and [Hb] LFO phase difference and amplitude changes during LS without PLM with the values obtained during LS with PLM (Fig. 2). Phase differences between [HbO] and [Hb] LFOs tended to decrease when PLM was present in both frontal areas (p=0.068). Furthermore, the amplitude of [HbO] LFOs tended to increase when PLM was present in the right frontal area (p=0.068).