Dysfunction of the cardiac parasympathetic system in fatal familial insomnia: a heart rate variability study

Abstract

Study Objectives

Although sympathetic hyperactivity with preserved parasympathetic activity has been extensively recognized in fatal familial insomnia (FFI), the symptoms of parasympathetic nervous system failure observed in some patients are difficult to explain. Using heart rate variability (HRV), this study aimed to discover evidence of parasympathetic dysfunction in patients with FFI and the difference of parasympathetic activity between patients with FFI and Creutzfeldt–Jakob disease (CJD).

Methods

This study enrolled nine patients with FFI, eight patients with CJD and 18 healthy controls (HCs) from May 2013 to August 2020. All participants underwent a nocturnal video-polysomnography with lead II electrocardiography, and the data were analyzed using linear and nonlinear indices of HRV during both wake and sleep states.

Results

Compared to the HC and CJD groups, the FFI group had a continuously higher heart rate with a lower amplitude of oscillations. The low frequency (LF)/high frequency (HF) ratio and ratio of SD1 to SD2 and correlation dimension D2 (CD2) were significantly different in the FFI group compared to the HC group. The root mean square of successive differences (RMSSD), HF and SD1 in the FFI group were significantly lower than in the HC group. RMSSD, SD1, and CD2 in the FFI group were all significantly lower than in the CJD group.

Conclusions

Cardiovascular dysautonomia in FFI may be partly attributable to parasympathetic abnormalities, not just sympathetic activation. HRV may be helpful as a noninvasive, quantitative, and effective autonomic function test for FFI diagnosis.

Statement of Significance.

It has long been believed that dysautonomia in those with fatal familial insomnia is caused by sympathetic overactivity with preserved parasympathetic activity. However, our study demonstrated that cardiovascular dysautonomia in FFI may be partially due to reduced parasympathetic function using heart rate variability, a noninvasive and quantitative autonomic function test. It may be a potential approach to help with the diagnosis of fatal family insomnia in research and practice to extract data from polysomnography to compute heart rate variability. Subsequent studies with bigger sample sizes and additional autonomic function tests should be conducted.

Introduction

Fatal familial insomnia (FFI) is a rare genetically determined prion disease linked to the D178N mutation coupled with the M129 genotype in the prion protein gene [1]. Unlike Creutzfeldt–Jakob disease (CJD), the most typical prion disease, FFI is characterized by prominent organic sleep disturbances accompanied with neuropsychiatric symptoms and dysautonomia [1–3]. In previous studies, dysautonomia, such as high heart rate (HR) and blood pressure, was reported in more than 80% of patients with FFI and it has long been attributed to sympathetic overactivation with preserved parasympathetic activity [4–7]. However, our previous review of 131 cases of genetically verified FFI showed that symptoms mainly regulated by the parasympathetic nervous system (PNS), such as constipation, urinary retention, and erectile failure, occurred in >1 per 10 patients [8]. This suggests that parasympathetic function may also be impaired in patients with FFI. Since previous studies on patients with FFI did not find evidence of decreased parasympathetic function, we decided to use heart rate variability (HRV) for this study, which has never been used in FFI before [4].

HRV reflects the regulatory function of the central autonomic network through analyzing the variance in beat-to-beat intervals of the heart [9]. Importantly, it is a noninvasive, reproducible method that has been widely used in diabetes, coronary heart disease, and sleep-related diseases, but it has never been used in FFI research [10–12]. The most important advantage of HRV is that it can quantitatively estimate complete PNS function and autonomic balance using only indicators obtained from a simple noninvasive electrocardiogram (ECG) signal extracted from a nocturnal video-polysomnography (PSG), allowing for sensitive and convenient detection of possible PNS dysfunction in patients with FFI [13, 14]. Furthermore, HRV is also expected to estimate the differences between FFI and CJD regarding PNS function and autonomic balance.

In the present study, we propose the hypothesis that parasympathetic function may be impaired in patients with FFI. Towards this goal, we explored cardiac parasympathetic function and sympatho-vagal balance in FFI, CJD, and healthy controls (HC) using HRV.

Methods

Participants

Nine patients with FFI, eight patients with CJD, and 18 HCs who were matched for sex and age were recruited from the Department of Neurology at Xuanwu Hospital, Capital Medical University, Beijing from May 2013 to August 2020. None of them was using medication for the treatment of sleeping disorders or those that affect HR. The patients with FFI and with CJD were identified according to the criteria [15] for FFI and CJD established in 2009 and 2012, respectively [16]. All participants underwent genetic testing to confirm and exclude diagnoses. All patients with FFI carried a PRNP 129MM mutation. Patients with heart disease, arrhythmias, sleep disorders of any kind, or other neurological illnesses were excluded from the study. The HCs were recruited from the community according to the following inclusion criteria: (1) no evidence of neurological or other physical diseases, such as respiratory, cardiac, renal, hepatic, and endocrinal diseases, based on clinical history, physical examination, and routine laboratory tests and (2) no medication within 14 days. Written informed consent was provided by participants or their families for participation in this study according to the Declaration of Helsinki. This study was approved by the Ethics Committee and local Institutional Review Board (IRB) of Xuanwu Hospital, Capital Medical University.

Polysomnography recording and sleep analysis

All patients underwent a nocturnal video-PSG in a temperature-controlled (21°C–24°C), humidity-controlled (40%–60%), partially soundproof room. They had already slept in the video-PSG test room the day before PSG testing to avoid first-night effects. They were required to abstain from all autonomic nervous system stimulants, such as coffee, tea, and alcohol, in the 24 h before the sleep study. Following the American Academy of Sleep Medicine (AASM) criteria [17], polysomnography recordings (E-Series; Compumedics Limited, Abbotsford, Australia) included measurements of six channels (F3-M2, F4-M1, C3-M2, C4-M1, O1-M2, and O2-M1) of the scalp electroencephalogram (EEG), two electrooculogram (EOG) electrodes applied to the cantus of each eye, submental and bilateral anterior tibial electromyography (EMG), nasal/oral airflow, percutaneous oxygen saturation, chest/abdominal respiratory effort, a lead II electrocardiography (ECG), and body position [18]. The EEG, ECG, and EMG signals were recorded at a 500-Hz sampling frequency. Digital bandpass filtering was applied to the EEG and ECG channels in the 0.3–70 Hz range and to the EMG channels in the 10–100 Hz range, and notch filters were applied at 50 Hz. The impedance of the recording electrodes was maintained below 5 kΩ. Sleep monitoring started at 10 pm and lasted at least 8 hours. A technician was present, and a video monitor was used throughout the registration.

Sleep analysis was independently performed by three experienced scorers, two of whom were blinded to the patient characteristics. The first scorer explored the data, performed a preliminary scoring according to the criteria of AASM, and isolated the specific PSG periods for which the criteria were clearly not applicable [19–22]. These unscored periods were scored blindly by the other two scorers according to “the alternative criteria” developed based on previous research [17]. The overall agreement for scoring of sleep analysis was 88.5%. Disagreements were resolved by a final discussion among the scorers.

HR and HRV Analysis

HR was defined as the number of heart beats per minute, and HRV was defined as the variations of the interval between two consecutive heart beats, both of which were obtained from the ECG channel of PSG [23]. The Pan–Tompkins’s method was used to detect R-waves of the ECG, and the time series of RR intervals were calculated by the time intervals between each pair of successive R-wave peaks. The entire algorithm was implemented in MATLAB. HRV analysis with successive 5 min was performed during quiet wakefulness before sleep and in the following non-rapid eye movement (non-REM) sleep stages N2 and N3 and rapid eye movement (REM) sleep of the first sleep cycle. The N1 stage was excluded from the analysis considering it is transient with an unstable condition in PSG. For each sleep–wake stage, 5-min intervals were chosen as to avoid stage shifts, micro-awakenings, fast-frequency EEG arousals, body movements, extrasystoles, breathing issues, and artifacts (identified visually).

HRV was evaluated with linear (time domain methods and frequency domain methods) and nonlinear measurements. In the time domain, the root mean square of successive differences between normal heartbeats (RMSSD) was performed. We calculated each successive time difference between heartbeats in ms, then squared each of the values, and the result was averaged before the square root of the total was obtained. The RMSSD reflects the beat-to-beat changes in HR and is used to estimate PNS changes reflected in HRV. The frequency domain indices, calculated with the absolute amount of signal energy within component bands, included the power spectral density (PSD) of low-frequency (LF) band (0.04–0.15 Hz) and high-frequency (HF) band (0.15–0.4 Hz) and the LF/HF ratio. While the LF band is sometimes equivocal and mainly reflects baroreceptor activity during rest, the HF band reflects parasympathetic activity under the influence of breathing and is highly correlated with RMSSD [9, 14, 24]. Although the LF/HF ratio is commonly used to estimate the balance between SNS and PNS activity in clinical research, it has been controversial because the proportion of SNS that contributes to LF power is not clear [25–27].

The nonlinear analysis used to quantify the unpredictability and complexity of a series of interbeat intervals were also calculated in this study, including the Poincaré plot and correlation dimension D2 (CD2). The Poincaré plot is graphed by plotting every R–R interval against the prior interval and analyzed by fitting an ellipse to the plotted points. The standard deviation of the distance of each point from the y = x-axis (SD1) and the y = x + average R–R interval (SD2) specified the ellipse’s width and length, respectively. The correlation dimension, which estimates the minimum number of variables required to construct a model of system dynamics, was also included as an index to reflect the balance between SNS and PNS [28]. The SD1/SD2 ratio and CD2 were used to measure autonomic balance.

Statistical Analysis

For demographic data and PSG findings, categorical variables were compared using Fisher’s exact test while continuous variables were compared using t-tests, nonparametric Kruskal–Wallis H-Test or one-way analysis of variance according to normality of data distribution determined using the Shapiro–Wilk test. The mixed-model repeated-measures analysis of variance test and post hoc tests (least significant differences, LSD) were applied to compare the differences of HRV parameters during wakefulness, stage N2, stage N3, and REM sleep periods. All statistical analyses were performed using SPSS v.18.0 (IBM Corp., Armonk, NY), and all figures were drawn on GraphPad Prism 7 (Graphpad Software, La Jolla, CA). A p-value less than .05 was considered significant.

Standard protocol approvals, registrations, and patient consent

This study was approved by the Ethics Committee and local Institutional Review Board (IRB) of Xuanwu Hospital, Capital Medical University and was carried out in compliance with the Declaration of Helsinki. All patients or their families provided written informed consent for study participation.

Data availability

Anonymized data will be shared on request from any qualified investigator.

Results

Demographics, clinical feature, and polysomnography findings

There were no differences in age (FFI group = 57.44 ± 7.31, CJD group = 51.50 ± 12.59, HC group = 53.65 ± 7.51; F = 2.157, p = .340) and sex (male/female: FFI group = 6/3, CJD group = 5/3, HC group = 12/8; value = 0.118, p = .943) among the three groups (Table 1). All three groups had body mass index <24 kg/m². In addition, the disease duration of the FFI group matched that of the CJD group (FFI group = 8.33 ± 4.06 months, CJD group = 7.38 ± 3.93 months; t = 0.493, p = .629). All of the above values are expressed as mean ± standard deviation. Among the nine participants with FFI, two came from the same pedigree, five of them came from five different pedigrees, and the other two patients had no clear family history of the condition. Only two patients in the CJD group had positive family history from different pedigrees.

Table 1.

Comparison of the demographic data and the results of PSG study

	FFI	CJD	HC	P-value
Number of cases	9	8	18	—
Age	57.44 ± 7.31	51.50 ± 12.59	53.65 ± 7.51	.340
Sex (male/female)	6/3	5/3	12/8	.943
Total sleep time (min)	248.61 ± 123.43	278.56 ± 160.25	400.20 ± 46.89	.003*
Sleep efficiency (%)	41.44 ± 20.57	51.48 ± 30.80	83.27 ± 6.84	<.001†
Sleep stage (%)
N1	21.16 ± 26.56	22.58 ± 17.97	13.31 ± 5.39	.534
N2	48.53 ± 18.30	50.73 ± 28.06	59.79 ± 4.81	.487
N3	23.90 ± 28.17	8.10 ± 9.18	6.83 ± 5.93	.480
REM	7.51 ± 7.75	6.10 ± 7.64	20.09 ± 4.62	<.001†
Breathing rates (bpm)
Wake	22.33 ± 4.82	23.63 ± 4.57	17.30 ± 3.16	.001†
N2	19.75 ± 5.60	17.57 ± 4.04	15.79 ± 2.53	.194
N3	18.11 ± 3.14	18.88 ± 4.36	15.84 ± 2.34	.076
REM	22.75 ± 4.77	21.80 ± 3.83	16.95 ± 2.30	.004†

Each value is expressed as mean ± standard deviation or the proportion of positive symptoms in total cases. Fisher’s exact test was used for categorical variables; the Kruskal–Wallis test or one-way analysis of variance were used for continuous variables. bpm: breaths per minute; FFI: fatal familial insomnia, CJD: Creutzfeldt–Jakob disease, HC: healthy control, REM: rapid eye movement.

*Only FFI and HC were statistically different among three groups.

^†FFI and CJD were statistically different compared with HC, while there was no difference between FFI and CJD.

All patients in the FFI and CJD groups showed sleep-related and neuropsychiatric symptoms. Although all patients with FFI had significant progressive autonomic symptoms, such as hyperhidrosis, hypertension, tachycardia, and weight loss (Supplementary Table 1), no patient with CJD reported definite complaints of dysautonomic symptoms. Further, although none of the patients in the FFI group had cortical or basal ganglia hyperintensity on diffusion-weighted imaging and fluid-attenuated inversion recovery imaging, five patients showed signs of thalamic hypometabolism on brain 18F-fluorodeoxyglucose-positron emission tomography/magnetic resonance imaging (18F-FDG PET/MRI). The detailed demographic data, clinical features, and auxiliary examinations of patients in the FFI group are shown in Supplementary Table 1.

During the nocturnal video-PSG, stage N1 and N2 sleep was observed in all patients with FFI and CJD; stage N3 sleep was observed in eight patients with FFI and four with CJD, and REM sleep was observed in five patients with FFI and four with CJD. No sleep spindles were found in most patients with FFI or CJD. A few sleep spindles with low amplitude were observed in one patient with CJD. The total sleep time, sleep efficiency, sleep stage distribution and respiratory rate in each sleep stage are shown in Table 1. Examples of sleep stages in case 5 of the FFI group are shown in Supplementary Figure 1.

HR and HRV

Heart rate

Results of the mixed-model repeated-measures analysis of ­variance showed that compared with the CJD and HC groups, the FFI group showed higher mean HR during wakefulness (FFI group = 95.68 ± 10.31 bpm, CJD group = 74.92 ± 20.69 bpm, HC group = 70.78 ± 11.68 bpm; F = 14.494, p < .001), stage N2 (FFI group = 90.48 ± 9.12 bpm, CJD group = 64.23 ± 17.26 bpm, HC group = 65.87 ± 9.25 bpm; F = 15.903, p < .001), stage N3 (FFI group = 91.49 ± 9.82 bpm, CJD group = 63.22 ± 21.00 bpm, HC group = 64.06 ± 8.92 bpm; F = 15.786, p < .001), and REM sleep (FFI group = 91.61 ± 12.71 bpm, CJD group = 60.63 ± 17.88 bpm, HC group = 67.98 ± 8.68 bpm; F = 14.256, p < .001) periods. All of the above values are expressed as mean ± standard deviation. Meanwhile, there was no significant difference in the mean HR between the CJD and HC groups. In addition, in the HC group, the mean HR decreased progressively from wakefulness to N2 and N3 stages and increased again in REM, whereas there was no such regular fluctuation in the FFI and CJD groups (Figure 1). All data were obtained from the lead II electrocardiography of nocturnal video-PSG.

Figure 1.

Mean heart rates during wakefulness, N2 and N3 stages, and REM sleep in the three groups. *p < .05, **p < .01.

HRV indicators related to parasympathetic function

RMSSD and SD1 were lower in the FFI group, while there were no differences between the CJD group and the HC group during wakefulness (F = 10.425, p < .001; F = 10.426, p < .001), stage N2 (F = 10.707, p < .001; F = 10.707, p < .001), stage N3 (F = 9.016, p < .001; F = 9.014, p < .001), and REM sleep periods (F = 6.120, p = .003; F = 6.118, p = .003) (Figure 2, A and C). Compared with the HC group, the FFI and CJD groups had lower PSD of the HF component during wakefulness (F = 7.212, p = .001), stage N2 (F = 8.716, p < .001) and stage N3 (F = 7.865, p = .001). In REM sleep, significant differences were only observed between the FFI group and the HC group (F = 4.989, p = .008) (Figure 2B). All data were obtained from the lead II electrocardiography of nocturnal video-PSG. The results of other common HRV indicators related to parasympathetic function are shown in Supplementary Table 2.

Figure 2.

Dynamics of RMSSD (A), high-frequency (B), and SD1 (C) during wakefulness, non-REM sleep stages N2 and N3, and REM sleep in the FFI, CJD, and healthy control groups. *p < .05, **p < .01.

HRV indicators related to autonomic balance

Compared with the HC group, the FFI and CJD groups had higher LF/HF ratio during wakefulness (F = 4.135, p = .018), stage N2 (F = 6.514, p = .002) and stage N3 (F = 8.104, p < .001). In REM sleep, significant differences were only observed between the FFI group and the HC group (F = 8.036, p = .001) (Figure 3A). The SD1/SD2 ratio was lower in the FFI groups than in the HC group during wakefulness (F = 4.429, p = .014) and stage N2 (F = 3.023, p = .052). Meanwhile, the SD1/SD2 ratio was not significantly different among the three groups during stage N3 and REM sleep (Figure 3B). However, CD2 was lower in the FFI group than in the CJD and HC groups during all wakefulness and sleep periods (Figure 3C). All data were obtained from the lead II electrocardiography of nocturnal video-PSG. The results of other common HRV indicators related to autonomic balance are shown in Supplementary Table 2.

Figure 3.

Dynamics of the low-frequency/high-frequency ratio (A), SD1/SD2 ratio (B), and correlation dimension (C) through wakefulness and stages N2, N3, and REM sleep in the FFI, CJD, and healthy control groups; *p < .05, **p < .01.

Discussion

This study found that compared to the HC and CJD groups, the FFI group had significantly higher HR and a lower amplitude of HR oscillations throughout wakefulness and sleep stages. LF/HF, SD1/SD2, and CD2, which reflect the balance of cardiac autonomic nerve function, and RMSSD, HF, and SD1 which reflect cardiac parasympathetic function, were all significantly different in the FFI group compared to the HC group. RMSSD, SD1, and CD2 in the FFI group were significantly lower than in the CJD group. This was a pilot study to present evidence that cardiovascular dysautonomia in patients with FFI is possibly partly attributed to reduced parasympathetic activation.

The unbalanced autonomic function in patients with FFI is an established clinical feature of the disease [5]. In addition to the disputed LF/HF ratio, the SD1/SD2 ratio and CD2 were also used to measure the balance between SNS and PNS activity in this study. Both the LF/HF and the SD1/SD2 ratios were significantly different during wakefulness and some sleep stages between the FFI and HC groups, whereas there was no difference between the FFI and CJD groups. However, the FFI group had lower CD2 than the CJD and HC groups during wakefulness and sleep stages, indicating a more marked alteration of the autonomic nervous system in FFI and also supporting the differential/unbalanced involvement of the SNS and PNS in FFI [28]. This finding could indicate that although CJD has subclinical autonomic function alterations, these are less severe than those in FFI and that CD2 can detect the difference between the two diseases. Combined with the significant increase in HR and no rhythmic changes in patients with FFI, our findings are consistent with previous norepinephrine studies indicating that the heart-rate axis of balance changed toward sympathetic activation [5–7, 29]. As early as 1991, researchers identified autonomic imbalances and sympathetic activation in patients with FFI refected in elevated plasma noradrenaline levels and blood pressure [5]. Since then, more studies have found abnormal increases of plasma norepinephrine levels and loss of normal secretion rhythms in FFI [6, 7, 29]. RMSSD, HF, and SD1, as indicators of vagally mediated changes, were significantly lower during the wakefulness and sleep stages in the FFI group than in the HC group, possibly indicating decreased PNS activation. Unlike the HF band, RMSSD and SD1 could also reflect a significant difference in PNS function between the FFI and CJD groups.

To the best of our knowledge, this study shows the first objective evidence of a parasympathetic hypofunction in FFI. Many indicators of autonomic function, such as the blood pressure, HR, and plasma adrenaline levels in the resting state or after injection of norepinephrine, can only reflect unbalanced autonomic function and highlight sympathetic hyperactivity [30]. In contrast, indicators of HRV (e.g., RMSSD and HF) can reliably reflect parasympathetic function. On the one hand, HRV can detect whether there is a moderate reduction in the mostly maintained parasympathetic function in FFI in the general population [5]. On the other hand, HRV does not only reflect the integrity of the baroreflex pathway but also the integrity of the entire PNS including the parasympathetic nerve center and its connecting fibers [5, 7]. Given that the majority of patients with FFI in this study were in the middle and advanced stages of the disease, the possibility of FFI parasympathetic hypofunctioning occurring exclusively in the advanced stage of the disease cannot be ruled out, as reported previously [6]. The nonparalleled loss of the circadian rhythm of blood pressure and then of HR reflected disease progression and confirmed the preservation of the parasympathetic function in patients with FFI, at least in the early stages of the disease. This is considering that both circadian rhythm of HR and blood pressure will completely disappear at its preterminal stages [6].

In our study, five patients with FFI displayed hypometabolism of the thalamus on a brain 18F-FDG PET/MRI scan, which has been described in detail in our previous studies, whereas none of the patients with CJD did [31]. This finding supported prior postmortem findings that specific thalamic injury was the primary cause of sympathetic overactivity, which may be also explain the decreased parasympathetic function in FFI [1, 32]. Ter Horst and Postema [33] used the retrograde transneuronal viral labeling technique to selectively identify the parasympathetic neurons in the forebrain involved in the regulation of heart activity in rats and found that the parasympathetic pathways were similar to the sympathetic pathways. More importantly, several studies have shown that the thalamus is involved in the regulation of the parasympathetic function index of HRV [34, 35]. One study used a combined HRV–fMRI approach to study the neural correlates of exercise-induced cardiovagal outflow in humans and found that HF power negatively correlated with fMRI activity in the mediodorsal thalamus [35]. Previous findings also showed that the complexity index of HRV was lower in patients with intracerebral hemorrhage of the thalamus [34]. Thus, we considered that degeneration of the thalamus might also contribute to the decreased parasympathetic activity in patients with FFI, and further analysis of the correlation between imaging and objective parasympathetic function indicators with a larger sample is needed to verify this notion.

The new clinical diagnostic criteria for FFI proposed by our team and published in 2022 better refined the clinical hallmarks of the disease, but early recognition of FFI and differentiation from CJD remains difficult [3]. Patients with FFI are easily misdiagnosed as having CJD when they present with rapid progressing dementia as a major complaint or have no positive family history [36]. Recent research suggests that patients with CJD may have FFI-like sleep abnormalities as evidenced by PSG and thalamic hypometabolism on PET imaging, making dysautonomia an essential factor in determining the difference between FFI and CJD [13, 37]. Although most clinicians have concentrated on progressive sympathetic symptoms in patients with possible prion disease, these symptoms are often ignored by patients and their families [4]. The current study found significant differences between the FFI and CJD groups with respect to RMSSD and SD1, which measure parasympathetic function, and correlation dimension, which assess the autonomic function balance. HRV may be a possible supportive feature for the clinical diagnosis of suspected FFI because FFI has more severe unbalanced autonomic nerve function and possibly lower parasympathetic activity than does CJD.

The limitations of this study mainly lie in the small sample size, which could have led to a relatively low statistical power of calculations. Subsequent studies with bigger sample sizes should be conducted. In addition, because the interpretation of the LF, the LF/HF ratio and SD2 is still controversial, and the LF, HF, and SD1/SD2 ratio are possibly influenced by respiration, baroreceptor sensitivity, disease severity, and an insufficiently long monitoring period, further research is needed to clarify the underlying biological mechanism of the HRV indices. Moreover, this study also lacked additional markers used to determine autonomic nerve function in patients, such as norepinephrine levels and the cold water immersion test.

Conclusions

In conclusion, HRV as a convenient and noninvasive method for evaluating autonomic nerve function, is helpful to exploring the value of autonomic nerve dysfunction in the diagnosis of FFI, and is expected to be used in the differential diagnosis of patients with FFI and CJD. The causes of cardiovascular dysautonomia in FFI patients may involve the role of both PNS and SNS, which requires further prospective studies.

Funding

This work was supported by the Beijing Municipal Natural Science Foundation (grant no. 7202060), the National Natural Science Foundation of China (no. 81971011), and the Ministry of Science and Technology of the People’s Republic of China (grant no. 2019YFC0118600).

Disclosure Statement

None declared.

Data Availability

The datasets generated and/or analyzed during the current study are not publicly available for privacy reasons, but data summaries are available from the corresponding author on reasonable request.