Introduction

Heart rate variability (HRV), a surrogate marker of autonomic system modulation, was found to be blunted in patients with Parkinson disease (PD) [1,2,3,4,5,6]. Some doubts persist regarding the type of autonomic dysfunction, and its relation with sleep stage and association with rapid eye movements (REM) sleep behaviour disorder (RBD). While some studies have shown a predominant affection of the sympathetic nervous system (SNS) [3, 4, 6], others have found signs of both SNS and parasympathetic nervous system (PNS) dysfunction [1, 2], depending on sleep stage [2], or motor dysfunction severity [1]. Some studies evaluated HRV only in the wake state [3], while others assessed polysomnography sleep data, finding differences between REM and non-REM stages [1, 4, 5]. Reduction in heart rate response to arousals or periodic limb movements was documented in idiopathic RBD (iRBD) [5]. Some studies have shown reduction in the variation of SNS modulated variables, more significantly in wake stages and in patients with PD compared to patients with iRBD [6]. Others found a reduction in very low and low frequencies components in iRBD compared to controls [7]. Lanfranchi et al. found an attenuation in normal HRV increase between NREM and REM stages in patients with iRBD compared to controls, affecting high and low frequency components as well as respiratory frequency [8]. Valappil and collaborators found that HRV during wakefulness was significantly decreased in patients with idiopathic REM sleep behavior disorder compared with control subjects, suggesting abnormalities of both sympathetic and parasympathetic function [9]. Some authors have suggested that HRV reduction could be specific for RBD, and not PD [3]. Others, however, have found alterations both in patients with PD with RBD (PD–RBD) and patients with PD without RBD (PD–NRBD) [2, 6]. Another study has found that circadian variation in low frequencies may accurately discriminate between patients with PD-RBD and PD–NRBD [10].

Our objective was to assess HRV changes across sleep stages (wake, NREM and REM), in patients with PD and iRBD, in order to understand if HRV blunting described in previous studies is related with the presence of PD or specifically with the presence of RBD, independently of PD status. This study tries to clarify previous findings, by including variables that were evaluated separately in past investigations. Our study includes three groups that differ from each other either in relation with the presence of RBD or the presence of PD (patients with idiopathic RBD, PD with RBD and PD without RBD) and three sleep stages that differ regarding the preponderance of SNS or PSN modulation (wake, Non-REM and REM sleep). We expect to find a blunting of the normal changes of HRV between sleep stages, related with the presence of RBD.

Methods

Participants

Patients with PD and iRBD were consecutively selected from the Movement Disorders and Sleep Disorders outpatients’ clinic of Egas Moniz Hospital and diagnosed according to the UK Brain Bank criteria [11] and the International Classification of Sleep Disorders (ICSD) III [12], respectively. We excluded patients with cardiac disease, dementia, signs or symptoms of Parkinsonian-plus disorders or any additional neurodegenerative diseases, and use of alcohol and drugs with influence on the autonomic nervous system. Dysautonomia, per se, was not an exclusion criterion (patients were not excluded if they did not present with the other relevant criteria for Parkinson-plus disorders).

Video-polysomnography (PSG)

All participants underwent one-night PSG, performed with a digital polygraph (XLTEK-TREX, Natus Medical Inc., Middleton, USA) that included electrooculography, six channel electroencephalography, electrocardiography, electromyography of the mentalis, right and left tibial muscles, recording of nasal air flow, thoracic and abdominal respiratory effort, oxygen saturation, microphone, and digital EEG-synchronized videography with infrared camera. All PSG were done in a sleep lab setting. Patients were forbidden to consume caffeine or alcohol in the day previous to PSG, but were allowed to take their usual medication.

Sleep staging and REM sleep muscular tone were scored according to the American Academy of Sleep Medicine (AASM) recommendations [13].

Heart rate variability analysis

One 5-min epoch with stable ECG, without arousals, motor or respiratory events, was selected from pre-sleep wakefulness, the N2 stage and REM stage (first epoch in each stage that fulfilled the criteria). Pre-sleep wakefulness was assessed with the patients already in bed, with lights off. ECG was sampled at 200 Hz. For each epoch, beat-to-beat variability was analyzed in the time and frequency domains, according to the guidelines proposed by the Task Force of the European Society of Cardiology and the North American Society and Electrophysiology [14], using Kubios HRV 2.2 software (University of Eastern Finland, Kuopio, Finland). ECG recordings were manually checked for technical and physiological artifacts. When artifact-free data was non-sufficient to allow a proper analysis, Kubios artifact correction options were applied using a medium (0.25 s) threshold for detecting RR intervals differing “abnormally” from the local mean RR interval and interpolating a RR value. Time domain variables included mean RR interval—Mean RR; standard deviation of RR intervals—SDNN and root mean score square of successive RR interval differences—RMSSD. As for frequency domain variables, spectral components of RR variability were quantified by a fast Fourier transform decomposition algorithm and classified as very-low-frequency (VLF < 0.04 Hz), low-frequency (LF 0.04–0.15 Hz) and high-frequency (HF 0.15–0.4 Hz) bands. VLF component was excluded from analysis, given that it could be considered dubious on short term readings. We calculated the normalized unit (n.u.) values for both LF and HF frequencies. Mean RR and SDNN are considered as general measures of HRV. RMSSD, HF and HF n.u. are mainly influenced by the PNS. LF reflect the compound influence of the PNS and the SNS. The physiological meaning of LF n.u. is debatable, some authors suggesting that it could correspond to SNS influence, others that it also reflects the compound influence of PNS and SNS systems [14]. Some authors have considered normalized values to be mathematically redundant measures [15]. Although the use of normalized values is debatable, they have been frequently reported in previous studies. We thus decided to include them, for the sake of comparison, albeit acknowledging that this part of the results should be subject to cautious interpretation.

Respiratory frequency analysis

Because respiratory frequency variation (RF) is also a marker of PNS modulation, we calculated the mean and standard deviation values of RF for each sleep stage segment.

Data analysis

The normality of variable distribution was tested with the Kolmogorov–Smirnov test. Demographic, sleep structure, dopaminergic medication, motor stage and respiratory frequency variables were compared between the PD–RBD, PD–NRBD and iRBD groups, using one-way ANOVA (for normally distributed variables) or the Kruskal–Wallis test (for non-normally distributed variables).

Sleep stage related differences in HRV and respiratory variables were compared between groups using repeated measures ANOVA. Because frequency domain variables were not normally distributed, their values were log transformed and used in calculations instead of the original values. We have assessed state differences, group differences and the interaction between each group factor and state changes. Mauchly’s test was used to test sphericity and the Greenhouse–Geisser correction was used whenever sphericity could not be assumed. Significance was set at p < 0.05.

In a first analysis, the changes in HRV across sleep stages were compared between the three groups by using repeated measures ANOVA with three levels of the within subject factor (sleep stage—wakefulness, N2 and REM) and three levels of the between subject factor (groups PD–RBD, PD–NRBB and iRBD.

In a second analysis, also using repeated measures ANOVA, we have performed two different comparisons: patients with PD (PD–RBD + PD–NRBD) vs. patients with idiopathic RBD; patients with RBD (PD–RBD + iRBD) vs. patients without RBD (PD–NRBD).

Finally, to test if there were significant differences in the time of the night from which REM and NREM sleep segments were selected, we assessed if NREM and REM segments were taken from the first half vs. second half of the night, and compared the frequencies between the three groups, by using Chi Square tests (wake segments were taken from pre-sleep wakefulness, which corresponded to the first half of the night in all patients).

Ethics

Patients and controls signed informed consent forms. The ethics committee of the institution approved the investigation protocol.

Results

We included 36 patients, out of 40 (two presented artifacts that prevented HRV analysis and two had atrial fibrillation): ten had iRBD and 26 had PD (eight PD–NRBD, 18 PD–RBD). Table 1 shows the comparison of demographic, motor function, dopaminergic treatment and sleep structure related values for the three groups. There were no significant differences in age, gender, sleep structure, Apnea–Hypopnea index or periodic limb movements of sleep between the three groups. DED and Hoehn and Yahr stage were higher in PD–RBD and PD–NRBD than in iRBD patients, as would be expected, with no significant differences between the two PD groups.

Table 1 Demographic, medication, motor stage, sleep structure and respiratory data in PD-NRBD, PD-RBD and iRBD
Full size table

Table 2 contains HRV and respiratory frequency values for the three groups according to sleep stage. There were significant sleep stage variations in mean RR, RMSSD and HF. There were significant group differences regarding mean RR and no significant group*stage interactions.

Table 2 Heart rate and respiratory frequency variability according to PD–NRBD, PD–RBD and iRBD status and sleep stage
Full size table

Comparisons between patients with RBD (PD–RBD + iRDBD) and patients without RBD (PD–NRBD) are depicted in Fig. 1 which also contains the results of the repeated measures ANOVA. There were significant stage variations in mean RR. Values in mean RR, RMSSD, HF, HF n.u. increased from wake to N2 and decreased from N2 to REM, while LF and LF n.u. changed in an inverse pattern. Stage vs. group interactions were significant for SDNN, HF and HF n.u.. Variations between stages were steeper in patients without RBD than in patients with RBD.

Fig. 1
figure 1

Heart rate variability according to sleep stage and group: patients with RBD (PD–RBD + iRBD) vs. patients without RBD (PD–NRBD). Values are estimated marginal means. Significance values obtained by repeated measures two away ANOVA with three repeated measures (wake, NREM and REM) and one independent factor (group). SDNN standard deviation of RR intervals, RMSSD root mean score square of successive RR interval differences, Log LF low frequency (log transformed), Log HF high frequency (log transformed), n.u. normalized units

Full size image

Comparisons between patients with PD (PD–RBD + PD–NRB) and patients without PD (iRBD) are shown in Fig. 2. Mean RR changed significantly between sleep stages, but there were no significant differences between groups or group vs. stage interactions.

Fig. 2
figure 2

Heart rate variability according to sleep stage and group: patients with PD (PD–RBD + iRBD) vs. patients without PD (iRBD). Values are estimated marginal means. Significance values obtained by repeated measures two away ANOVA with three repeated measures (wake, NREM and REM) and one independent factor (group). SDNN standard deviation of RR intervals, RMSSD root mean score square of successive RR interval differences, Log LF low frequency (log transformed), Log HF high frequency (log transformed), ms milliseconds, n.u. normalized units

Full size image

The percentage of patients from which NREM sleep segments were taken from the first half of the night was 75.0 in iRBD, 100.0 in PD–NRBD and 85.7 in PD–RBD (p = 0.520). The percentage of patients from which REM sleep segments were taken from the first half of the night was 75.0 in iRBD, 45.0 in PD–NRBD and 22.9 in PD–RBD (p = 0.143).

Discussion

The present investigation has compared several measures of HRV between patients with PD (with and without RBD) and iRBD, collected from three different sleep stages. In a first comparison between these three groups, we did not find significant group vs. stage interactions. In a second analysis, groups were defined according to: (1) the presence vs. the absence of RBD; and (2) the presence vs. the absence of PD. When comparing patients with vs. without RBD, we found a significant stage vs. group interaction regarding HF and HF n.u., patients with RBD showing a flattening in HRV changes between stages. The comparison between patients with and without PD yielded no significant differences between groups or group vs. stage interactions. Taken together, our results suggest that HRV is lower in patients with RBD, and that the variation of HRV between stages is significantly blunted in this group compared with patients without RBD. The present data widens the significance of previous findings by showing that HRV blunting in RBD patients encompasses several sleep stages and not only the wake state, as in Postumas’ et al. [3]. It also suggests that this HRV attenuation particularly affects HF measures. Given that HF power reflects parasympathetic modulation [14], it could signify increased derangement regarding this part of the autonomic nervous system in RBD patients. In non-diseased subjects, total HRV decreases from wake to NREM stages, increasing again from NREM to REM. PNS tone predominates in NREM stage, while SNS tone is higher in wake and REM stages [16]. Our data shows an attenuation of this normal variation pattern, related to the presence of RBD. These findings are in accordance with the notion that HRV blunting could be connected with the neurodegenerative process that causes RBD symptoms and not with the presence of Parkinsonian signs per se. RBD is believed to be caused by spreading of alpha-synuclein to pontomedullary structures and particularly the sublaterodorsal nucleus and sub-coeruleus complex [17]. Nearby structures, like the dorsal motor nucleus of the vagus, the nucleus of the solitary tract and the rostral ventrolateral medulla are part of the central control of PNS autonomic system, and also become affected in the early stages of disease [17].

Divergences regarding studies that did not find RBD to influence HRV [2, 6], could in part be explained by methodological differences. Palma et al. [2] study was not directed at evaluating differences according to RBD (only six out of 33 patients with PD had RBD). Sorensen’s study [5] was directed at evaluating HR response to PLMS and arousals. Slower heart rate response to stimuli like arousals, limb movements or complex motor events (as also recently described [18]) in PD patients compared with iRBD patients could be attributed to the former being in a more advanced stage of alpha-synuclein related neurodegeneration. The findings of these studies conflict with our present investigation, and some of the previous ones, the difference being that the latter have investigated HRV in steady state sleep segments and not the response to HR increasing stimuli. We hypothesize that this difference in methodology could be the origin of the discrepancy, but our data does not permit to further test this hypothesis. The finding of PNS modulating deficits partially conflicts with the study by Sorensen and collaborators [6], who found differences in LF, that they interpret as a sign of SNS dysfunction. However, LF values are believed to reflect both SNS and PNS influences, which problematizes the interpretation of this variable [14]).

This investigation presents some limitations. By analyzing only a 5-min segment in each stage of sleep, we could have narrowed its generalizability, as we did not account for circadian changes in HRV. On the other hand, our analysis shows that there were no significant differences between the three groups regarding the time of the night NREM and REM sleep segments were taken from, which may suggest that circadian differences should not have affected our results. Mean age of both PD and iRBD groups, although reflecting the demographic features of patients attended at our center (and not differing significantly between groups), is higher than in some of the previous studies, which should be taken in account when comparing results.

In conclusion, our study suggests that HRV reduction, a marker or autonomic system dysfunction, is predominantly related with RBD, irrespectively of the presence of PD diagnosis. RBD patients showed lower HRV values in all stages and significant attenuation of HRV changes between sleep stages (particularly in PNS variables).