Abstract
Background:
Earlier detection of parkinsonism, specifically during its prodromal stage, may be key to preventing its progression. Previous studies have produced contradictory results on the association between sleep symptoms and prodromal parkinsonism.
Objective:
We conducted a prospective study within the Canadian Longitudinal Study on Aging (CLSA) to determine whether self-reported symptoms of insomnia, somnolence, apnea, and restless legs syndrome predate the diagnosis of parkinsonism after three years of follow-up.
Methods:
At baseline, amongst other information, participants completed a questionnaire for difficulty initiating or maintaining sleep, daytime somnolence, snoring or stopping breathing during sleep, and symptoms of restless legs syndrome. After 3 years of follow-up, baseline responses from participants who self-reported a new diagnosis of parkinsonism (cases) were compared to those who did not (controls). For each case, 10 controls were individually matched by age, sex, education, BMI, caffeine, smoking, and alcohol. Binary unconditional logistic regression models were used to estimate the association between sleep symptoms and new-onset parkinsonism, adjusting for age, sex, education, BMI, smoking, alcohol, and caffeine.
Results:
We identified 58 incident-parkinsonism cases and 580 matched controls (65.5%male, mean age = 69.60, SD = 8.0). Baseline symptoms of sleep-onset insomnia (12.1%vs. 13.0%, Adjusted OR[95%CI] = 0.87[0.32,2.33]), sleep-maintenance insomnia (24.1%vs. 20.2%, AOR = 1.01[0.46,2.20]), daytime somnolence (8.6%vs. 7.4%, AOR = 1.11[0.37,3.39]), obstructive sleep apnea (27.3%vs. 26.2%, AOR = 0.84[0.40,1.79]), and restless leg syndrome (20.6%vs. 9.9%, AOR = 1.34[0.42,4.25]) were similar among those who developed parkinsonism and those who did not.
Conclusion:
Symptoms of insomnia, somnolence, apnea, and restless legs did not predate a new diagnosis of parkinsonism over 3 years.
INTRODUCTION
Earlier detection of parkinsonism, specifically during the prodromal stage, may be key to preventing its progression. However, identification of prodromal parkinsonism and Parkinson’s disease (PD) is challenging, and diagnosis is often delayed until the motor features are manifest.
Sleep disorders (e.g., insomnia, daytime somnolence), parasomnia (e.g., rapid eye movement sleep behavior disorder (RBD), restless leg syndrome (RLS)), and breathing-related sleep disorders (e.g., obstructive sleep apnea (OSA)) are commonly reported in PD [1–3] and are frequently reported at the early stages of the disease [4]. However, except for RBD, which is a well-established predictor of parkinsonism [5–7], there is limited and often-contradictory evidence on whether other sleep disorders are prodromal markers of PD. Some studies reported a higher prevalence of PD in patients with insomnia [8], daytime somnolence [9], OSA [10–12], and RLS [13]. On the contrary, other studies showed no evidence of an association between sleep problems and PD/parkinsonism [14–18].
This study aims to determine whether sleep symptoms predate the diagnosis of PD/parkinsonism in a cohort of the Canadian population. The Canadian Longitudinal Study on Aging (CLSA) is a large population-based cohort of participants aged 45–85, at enrolment. At baseline, amongst other information, subjects completed screenings for sleep symptoms. We evaluated the development of parkinsonism after 3 years of follow-up and conducted a prospective nested case-control study within the CLSA cohort to explore if sleep symptoms predated parkinsonism diagnosis.
MATERIALS AND METHODS
Study population
Data were taken from the Comprehensive cohort of 30,097 participants of the CLSA, selected from the general Canadian population aged 45–85 years. Participants were selected randomly from 11 data collection sites from 7 Canadian provinces. Persons living on federal First Nations reserves, full-time members of the Canadian Armed Forces, residents of institutions, and those with inability to respond in English or French, or cognitive impairment were not recruited by CLSA. Data were collected during in-home visits, data collection site visits, and telephone interviews [19]. The baseline data were collected during 2012–2015, and the first follow-up data were collected in 2015–2018. 2,888 subjects were lost to follow-up, resulting in 27,209 individuals for the first follow-up. Full details of the study can be found elsewhere [19, 20].
The CLSA protocol was reviewed and approved by 13 research ethics boards across Canada. Signed informed consent was obtained from all participants [19].
Cases
At follow up, a diagnosis of parkinsonism was provided by self-report according to response to the question “Has a doctor ever told you that you had Parkinsonism or Parkinson’s Disease?”. To ensure accuracy of diagnosis, we additionally required participants to either report parkinsonism symptoms (i.e.,≥3 parkinsonism symptoms at follow-up according to a validated screening questionnaire [21]), or to be taking medication for parkinsonism. Participants with prior diagnosis of PD/parkinsonism at baseline were excluded from the primary analysis.
Controls
For each case, 10 controls free of PD/parkinsonism at baseline and follow-up were selected from the cohort, matched by baseline age, sex, education, caffeine, smoking, alcohol consumption, and BMI to each case. Further details of parkinsonism diagnosis and matching variables can be found in the Supplementary Material.
Sleep symptoms
At baseline, the following sleep symptoms were queried and coded as binary (yes/no): Difficulty initiating sleep was queried as: “Over the last month, how often did it take you more than 30 minutes to fall asleep?” [22]. Those reporting these symptoms with a frequency≥3 nights per week were considered as having symptoms of sleep-onset insomnia. Difficulty maintaining sleep was queried as: “Over the last month, how often did you wake in the middle of the night or too early in the morning and found it difficult to fall asleep again?” [22]. Those reporting these symptoms with a frequency≥3 nights per week were considered as having symptoms of sleep-maintenance insomnia. Difficulty staying awake was queried as: “Over the last month, how often do you find it difficult to stay awake during your normal waking hours when you want to?” [22]. Those reporting these symptoms with a frequency≥3 nights per week were considered as having symptoms of daytime somnolence. Possible obstructive sleep apnea was screened according to the STOP questionnaire [23]:≥2 self-reported symptoms of snoring, daytime somnolence, being observed to stop breathing, or hypertension were considered as indicating high risk of sleep apnea. Restless leg syndrome was defined according to the 4-item minimal criteria [24]: uncomfortable feeling in legs and urge to move legs while sitting or lying down, that gets worse during evenings/nights and relieves by movement/activity ≥3 times per week. Overall dissatisfaction with sleep pattern, according to their response to the question: “How satisfied or dissatisfied are you with your current sleep pattern?” [25]. This question was queried in 5-Likert criteria. Participants were categorized into two groups: dissatisfied (those who answered: very dissatisfied or dissatisfied) or not dissatisfied (those who answered: neutral, satisfied or very satisfied). Number of hours of actual nighttime sleep during the past month (numerical variable).
Secondary/sensitivity analyses
The primary analysis was whether the presence of each of the cardinal sleep problems predated parkinsonism diagnosis. In addition, we performed several sensitivity and secondary analyses, as below. Duration of symptoms (secondary analysis) –We hypothesized that sleep symptoms of a more recent onset may be more likely due to prodromal neurodegeneration. Therefore, we stratified results according to the duration of sleep symptoms (<or ≥10 years). Prevalent parkinsonism (secondary analysis) –For comparison purposes (e.g., to ensure the validity of sleep questionnaire) and to evaluate the baseline association between parkinsonism and sleep symptoms, we compared participants with prevalent parkinsonism (i.e., those reporting parkinsonism at baseline) with those who did not report parkinsonism at baseline in the entire cohort. Incident parkinsonism in the entire cohort (sensitivity and secondary analysis) –To show the result of the entire cohort and not just the matched controls, we also compared incident parkinsonism (new-onset cases at follow-up) with controls in the entire cohort (parkinsonism-free participants at baseline and follow-up). Sex- and age-stratified analysis (secondary analysis) –We evaluated the relationship between PD/parkinsonism and different sleep problems stratified according to sex (male/female) and age (≤or > 70 years). Criteria for a defined sleep-wake disorder (e.g., in DSM or AASM guidelines) generally require that the sleep symptoms have a direct impact on the quality of life and/or be distressing to the patient [26]. Our primary research question centered around the presence or absence of sleep symptoms per se (i.e., independent of their perceived impact on quality of life); however, we also conducted a secondary analysis assessing the presence of sleep onset/maintenance insomnia disorder according to DSM-V criteria, which was defined as dissatisfaction with sleep pattern and difficulty initiating/maintaining sleep with a frequency ≥3 nights per week, and duration ≥3 months, that significantly interfered with daily function. Interference with daily function was queried as “To what extent do you consider your problem falling/staying asleep to interfere with your daily functioning (for example, from daytime fatigue, ability to function at work/daily chores, concentration, memory, mood, etc.)”, with a severity of Somewhat, Much, or Very much being coded as daily function interference. Assessing possible vascular parkinsonism confound (secondary analysis) –For this analysis, we divided the parkinsonism screening questions [21] into two domains: 1) symptoms that are common/overlapping in both idiopathic PD and vascular parkinsonism (e.g., trouble rising from a chair, poor balance, freezing, and shuffling feet), and 2) symptoms less commonly observed in vascular parkinsonism (e.g., small handwriting, soft voice, less expressive face, tremor, and difficulty buttoning buttons). We calculated the average scores in each domain, then stratified incident parkinsonism patients into two groups: PD with vascular-overlap symptoms (average score for possible vascular-PD symptoms greater than nonvascular symptoms), and PDs with nonvascular-overlap symptoms. Dividing incident parkinsonism cases into two groups: those who had parkinsonian symptoms [21] at baseline (baseline parkinsonism screening score ≥3), and those who reported no/few prodromal symptoms (baseline parkinsonism screening score < 3), and compare their previous exposure to sleep problems (secondary analysis).
Statistical analysis
Descriptive results were presented using the mean and standard deviation (SD) for continuous variables and frequency for categorical variables.
Binary logistic regression models were used to estimate the odds ratio (OR) of parkinsonism incidence and 95%confidence intervals (CI). The predictor variables of interest were the exposure to any of the sleep symptoms/problems. Crude OR and adjusted OR (AOR), including age, sex, education, BMI, caffeine, smoking, and alcohol in the model, are presented.
Loss to follow-up
As this was a prospective study looking at development of parkinsonism, only participants who provided both baseline and follow-up data could be included. The lost to follow-up cases are often different from those who provide follow-up data. To provide a better insight into these groups’ characteristics, we compared sociodemographic data, parkinsonism screening score, baseline prevalence of PD/parkinsonism, and sleep problems among those who had follow-up and those lost to follow-up in the entire cohort.
RESULTS
Among the 27,765 CLSA participants who provided follow-up data, 58 individuals previously free of parkinsonism reported PD/parkinsonism at follow-up, among which 52 (89.7%) were taking medication for PD, and 6 (10.3%) were not taking medication but had ≥3 parkinsonism symptoms. 26,516 were negative for parkinsonism at both baseline and follow-up. We selected 580 of these for the matched-control group, randomized case order when drawing matches, and took samples without replacement. In both groups, 65.5%of participants were male; the mean age of cases and controls were 69.7 (SD = 8) and 69.6 (SD = 8) (matching variables) (Table 1).
Baseline characteristics of incident parkinsonism cases and matched controls
None of the reported baseline sleep symptoms were found to be associated with the risk of developing parkinsonism after three years. The number of sleep hours was also similar among cases and the matched-control group (Table 2). Sleep onset/maintenance insomnia disorder (according to DSM-V criteria) was observed in 16.2%of incident parkinsonism cases and 11.0%of matched-controls, however the 95%confidence interval crossed 1 (AOR = 1.56 [0.61, 3.98]).
Baseline sleep symptoms among incident parkinsonism cases and matched controls
AOR, Adjusted for age, sex, education, BMI, smoking, alcohol, and caffeine consumption.
When sleep symptoms were divided according to duration, both long-term (≥10 years) and short-term (<10 years) insomnia symptoms were not clearly different between cases and controls. Similarly, long-term or short-term daytime somnolence and uncomfortable feeling or urge to move legs were not clearly different in both groups (Table 3).
Stratification according to duration of sleep symptoms
AOR, Adjusted for age, sex, education, BMI, smoking, alcohol, and caffeine consumption.
Stratification by age (≤or > 70 years) (Supplementary Table 1) or sex (Supplementary Table 2) did not change the results, except for ‘being observed stopped breathing in sleep’, which was higher among women reporting incident parkinsonism (21.1%vs. 7.1%, AOR = 4.68 [1.15, 19.05]) (Supplementary Table 2). No sleep symptoms were different between vascular-overlap or nonvascular-overlap incident parkinsonism cases and control group (Supplementary Table 3). Also, dividing incident cases into those with and without motor symptoms at baseline (on the parkinsonism screening questionnaire), revealed that exposure to sleep problems was similar between those with and without baseline motor symptoms (Supplementary Table 4).
In contrast to incident parkinsonism, we saw the expected evidence of sleep abnormalities in those with prevalent parkinsonism at baseline. Among the 114 prevalent parkinsonism patients, 18.4%had daytime somnolence vs. 8.8%of the controls (AOR = 2.37 [1.20, 4.65]). Uncomfortable feeling in legs and urge to move the legs were more frequent among prevalent parkinsonism patients compared to controls (52.6%vs. 33.3%, AOR = 2.07 [1.20, 3.58], and 50.4%vs. 32.6%, AOR = 1.87 [1.08, 3.22], respectively) Table 4).
Comparison of sleep symptoms at baseline among prevalent parkinsonism, incident parkinsonism (diagnosed at 3-year follow-up), and control groups, in the entire cohort
AOR, Adjusted for age, sex, education, BMI, smoking, alcohol, and caffeine consumption.
Loss to follow-up
Participants lost to follow-up were significantly older, less married/lived with partners, less educated, and more likely to smoke. Participants who did not provide follow-up data, were more likely to report PD/Parkinsonism at baseline (0.8%vs. 0.3%, OR = 2.27 [1.43, 3.63]). The prevalence of all studied sleep problems was significantly higher among participants with loss to follow-up (Supplementary Table 5).
DISCUSSION
Although PD patients often report different types of sleep disorders during the course of their disease, this population-based study did not find evidence that symptoms of insomnia, somnolence, apnea, or restless legs syndrome predated the diagnosis of PD/parkinsonism. Those with prevalent parkinsonism at the study baseline however, reported more somnolence and symptoms of uncomfortable feeling in legs and urge to move legs, compared to controls.
Insomnia and somnolence
Our findings revealed no evidence of an association between sleep symptoms and future diagnosis of parkinsonism over three years, which was similar to the findings of Abbott et al. [9]. By contrast, studies from the Taiwanese national cohort reported a 1.2 times higher risk of PD among patients diagnosed with chronic insomnia [8], and a higher rate of developing PD in patients taking zolpidem [27].
Similarly, whereas somnolence is a common side effect of PD medications and is common in late-stage PD, there is still controversy as to whether somnolence antedates PD. In a cohort study of 91,273 sleep disorder sufferers and 91,273 healthy controls, Gao et al. reported that sleepiness could be considered a predictive marker for PD [28]. Abbott et al., in their study on 3,078 Japanese-ancestry men aged 71–93 years, also reported a 3-times higher incidence of PD in patients with excessive daytime sleepiness [9]. Participants’ characteristics (older, single-sex, and only Japanese race) were the main difference between these two studies. However, our supplementary analysis only among men and only in participants > 70 years also did not find a link between daytime sleepiness and incident parkinsonism. Similarly, studies from the Parkinson’s Progression Markers Initiative (PPMI) cohort demonstrated no difference in somnolence among 423 untreated de novo PD patients compared to controls [29].
Apnea
Published evidence for apnea as a primary feature of PD is limited. In a Taiwanese study on 1,944 sleep apnea patients and 9,720 matched non-apnea controls, a 1.85-times (95%CI = 1.02,3.35) higher risk of PD was observed in the sleep apnea group after three years of follow-up, which was only observed in men, and patients older than 60 [10]. Similar results were obtained in another Taiwanese study of 16,730 OSA patients and 16,730 matched controls, and after the average 5.6-year follow-up period, OSA patients had a 1.37-fold (95%CI = 1.12,1.68) higher risk of developing PD [11]. However, contrary to the previously mentioned studies, Sheu et al. reported no statistically significant higher risk of PD among OSA male patients (adjusted hazard ratio = 1.72, 95%CI = 0.84,3.52), but showed 3.54 times (95%CI = 1.50,8.34) higher risk in female patients with OSA [30]. In a case-control study, Yong et al. did not find a difference in polysomnography-diagnosed OSA between 56 prevalent PD patients and 68 matched controls [17]. A meta-analysis of the results of studies with polysomnographic data presented a lower prevalence of OSA in PD patients, and they related this to lower BMI in PD patients [31]. In our study, we found no clear link between symptoms of apnea and PD; however, it is important to note that whereas the other sleep disorders in this study are defined by self-reported symptoms, diagnosis of apnea requires a polysomnogram. Therefore, our apnea findings should be viewed as preliminary.
A potential reason for conflicting results in apnea studies could be confounding by ‘vascular parkinsonism’, in which microinfarctions cause features that mimic idiopathic PD [1]. Participants with possible OSA did not have more of potential vascular parkinsonism symptoms (e.g., gait dysfunction). Details of this analysis are reported in the Supplementary Material and Supplementary Table 3.
Restless leg syndrome
Our results showed a similar prevalence of RLS among incident parkinsonism patients and controls. This is consistent with Ondo et al. who did not find evidence that early-onset RLS can predict the later development of PD. They evaluated the prevalence of RLS on 303 PD patients and reported that in 68%of the patients with RLS/PD, RLS symptoms occurred after PD symptoms [18]. Contrary to our results, Szatmari et al., in the national prospective cohort of 3.5 million US veterans selected 50,441 RLS patients and 50,441 matched-controls. RLS group had a 2.6 times (95%CI = 1.95,3.39) higher incidence of PD compared to the control group after 7.8 years of follow-up [13]. Moreover, Wong et al. found a 2.77-fold (95%CI = 1.08,7.11) increased risk of PD among men with RLS symptoms, present only in the first 4 years of follow-up [32]. Sohail et al. in their study on non-PD adults, found a positive link between sleep fragmentation and Lewy body pathology and substantia nigra neuron loss in brain autopsy, which can serve as marker of parkinsonism. Sleep fragmentation and sleep symptoms are not identical, although could manifest as insomnia, somnolence or RLS [33]. Of note, pain and reduced pain thresholds are well-documented features of PD; therefore, it remains unclear to what degree non-specific leg discomfort could confound any links between RLS and PD.
Sleep quality and duration
Sleep satisfaction and sleep duration were not different between incident PDs and controls. However, Lysen et al., in a prospective population-based Rotterdam Study, found that poor sleep quality and short sleep duration were associated with increased risk of PD, which was independent of baseline depression. They reported deterioration in sleep quality and shortening of sleep during the prodromal stage and over the two years prior to the diagnosis of Parkinsonism [34].
Idiopathic vs. Vascular parkinsonism
Certain symptoms of parkinsonism could be potentially confounded by ‘vascular parkinsonism’, which is a syndrome characterized by predominant gait impairment, generally due to extensive small white matter infarctions. Diagnosis of parkinsonism in this study was according to patient self-report; thus, differentiation between idiopathic parkinsonism and vascular parkinsonism could not be achieved entirely. To ensure the diagnosis of idiopathic parkinsonism, we conditioned the diagnosis by either taking medication for PD (89.7%) or ≥3 parkinsonism symptoms (10.3%). Also, we stratified incident parkinsonism patients into two groups according to the parkinsonism screening questionnaire: PDs with vascular-overlap symptoms and PDs with nonvascular-overlap symptoms (secondary analysis –part f). Sleep problems were similar between these groups and matched-controls, suggesting that our findings were less likely to be confounded by vascular parkinsonism.
Loss to follow-up
In most of the longitudinal studies, the loss to follow-up is not random [35], and similarly in our cohort, the probability of being lost to follow-up seem to be dependent of the outcome (i.e., parkinsonism), the exposure (i.e., sleep problems), and confounders (i.e., age). This implies that those with follow-up are more healthy. In general, non-random loss to follow-up tends to bias results towards lower observed odds ratio compared to the true odds ratio [35]. However, in this study, the percentage of loss to follow-up was < 10%, which is within the acceptable range (i.e., the observed OR is very close to the true OR) [35].
Strength and limitations
Some limitations of this study should be noted. First, despite the considerable size of the CLSA cohort, the number of incident parkinsonism cases was relatively small, limiting the power to assess small associations. In addition, the diagnosis of PD/parkinsonism and sleep problems were all self-reported and a physician assessment was not performed and there was no validation with medical records or objective sleep tests (e.g., polysomnography, accelerometry-derived measures). Nevertheless, we were able to use additional measures (i.e., the screening questionnaire [21]) included in the CLSA to help verify the diagnosis of PD/parkinsonism. Regarding the diagnosis of OSA, we could use only the STOP questionnaire rather than the full STOP-BANG criteria. Note also that unlike the other sleep conditions, a full diagnosis of apnea requires polysomnography; therefore, any results on apnea should be interpreted with caution. The diagnosis of RLS was performed according to the 4-item minimal criteria, which has the limitation of overdiagnosis of RLS due to RLS mimics. Finally, loss to follow-up can be a source of survival bias, also recall bias is likely when evaluating the duration of each sleep symptom. Further details of those lost to follow-up are presented in the Supplementary Material. On the other hand, the study had notable strengths, including population-based sampling, and prospective assessment of outcomes. Baseline assessments of exposure eliminates recall bias. Also, owing to the comprehensive nature of the CLSA database, we were able to match for diverse potential confounding variables.
Conclusion
Numerous sleep symptoms, including those of sleep onset/maintenance insomnia, daytime somnolence, obstructive sleep apnea, restless leg syndrome, dissatisfaction, and number of sleep hours did not predate the diagnosis of parkinsonism after 3 years of follow-up.
Footnotes
ACKNOWLEDGMENTS
This research was made possible using the data/biospecimens collected by the Canadian Longitudinal Study on Aging (CLSA). Funding for the Canadian Longitudinal Study on Aging (CLSA) is provided by the Government of Canada through the Canadian Institutes of Health Research (CIHR) under grant reference: LSA 94473 and the Canada Foundation for Innovation. This research has been conducted using the CLSA, Baseline Comprehensive Dataset version 4.1, and Follow-Up1 Dataset version 3.0, under Application Number 1906008.”
The CLSA is led by Drs. Parminder Raina, Christina Wolfson and Susan Kirkland.
DATA AVAILABILITY
Data are available from the Canadian Longitudinal Study on Aging webpage (
) for researchers who meet the criteria for access to de-identified CLSA data.
Furthermore, the opinions expressed in this manuscript are the author’s own and do not reflect the views of the Canadian Longitudinal Study on Aging.
CONFLICT OF INTEREST
S. Zolfaghari, Y. Yao, and A. Pelletier have nothing to disclose. RB. Postuma reports grants and personal fees from Fonds de la Recherche en Santä, grants from the Canadian Institutes of Health Research, grants from The Parkinson Society of Canada, grants from Weston-Garfield Foundation, grants from Michael J. Fox Foundation, grants from Webster Foundation, personal fees from Takeda, personal fees from Roche/Prothena, personal fees from Teva Neurosciences, personal fees from Novartis Canada, personal fees from Biogen, personal fees from Boehringer Ingelheim, personal fees from Theranexus, personal fees from GE HealthCare, personal fees from Jazz Pharmaceuticals, personal fees from Abbvie, personal fees from Jannsen, personal fees from Otsuko, personal fees from Phytopharmics, personal fees from Inception Sciences, other from Parkinson Canada, personal fees from Curasen, outside the submitted work. C. Wolfson reports grants from the Canadian Institutes of Health Research, the Multiple Sclerosis Society of Canada, the Public Health Agency of Canada, and the Canada Foundation for Innovation.