Introduction
Obstructive sleep apnea (OSA) is characterized by repetitive upper airway obstruction during sleep. OSA, defined by an apnea-hypopnea index ≥ 15 events per hour on polysomnography (PSG), is prevalent in 49.7% of men and 23.4% of women aged 40–85 years, and increases with age. Reference Heinzer, Vat and Marques-Vidal1 OSA remains largely underdiagnosed and undertreated in the general population (GP). Reference Kapur, Strohl, Redline, Iber, O'Connor and Nieto2–Reference Flemons, Douglas, Kuna, Rodenstein and Wheatley4 OSA contributes to poor outcomes like excessive daytime sleepiness, motor vehicle accidents, neurocognitive impairment and cardiometabolic diseases. Reference Knauert, Naik, Gillespie and Kryger5
Sleep disturbances are frequent in neurodegenerative diseases. Reference Miner and Kryger6 OSA is prevalent in 20-66% of individuals with Parkinson’s disease (PD), Reference Neto, Pereira, Sobreira, França Fernandes, Tumas and Eckeli7 38%–53% in Alzheimer’s disease (AD) Reference Gaeta, Benítez and Jorge8 and 30%–80% in stroke. Reference Liu, Lam, Chan, Chan, Ip and Lau9 OSA appears to have a bidirectional association with several neurological disorders. Reference Rosenzweig, Glasser, Polsek, Leschziner, Williams and Morrell10 It is an independent risk factor for worse neurocognitive outcomes. Reference Boulos11–Reference Gosselin, Baril, Osorio, Kaminska and Carrier13 and treatment can improve neurocognitive. Reference Costa, Lim and Thorpe14–Reference Kaminska, Mery and Lafontaine16 and functional Reference Ryan, Bayley, Green, Murray and Bradley17–Reference Aaronson, Hofman and van Bennekom19 outcomes. Hence OSA may be a modifiable risk factor in neurodegenerative conditions Reference Boulos11 but a large burden of untreated OSA remains.
PSG, the gold standard for diagnosis of OSA, is often difficult to obtain, costly and burdensome to individuals with neurodegenerative disease. Several screening instruments exist that might help identify patients at high risk for OSA (HROSA) for more targeted testing Reference Ramachandran and Josephs20 However, few tools have been validated specifically in neurological populations.
The primary aim for this study was to compare, in the Canadian Longitudinal Study on Aging (CLSA), the prevalence of HROSA using commonly used questionnaires and of self-reported OSA, in individuals with PD, AD, and stroke. Secondary aims were to compare the prevalence of HROSA and of self-reported OSA diagnosis between neurological conditions versus the general population (GP).
Methods
We conducted a cross-sectional study using the CLSA, a population-based, prospective research cohort of older Canadians that recruited 51,338 participants, aged 45–85 years, with no major cognitive impairment. Reference Raina, Wolfson and Kirkland21 The CLSA comprehensive cohort collected data through face-to-face and telephone interviews on demographic, social, medical, functional and other aspects relevant to aging at baseline with a plan for follow-up at three-year intervals. We used baseline (collected 2011–2015) and three-year follow-up data (collected 2015–2018). This study was approved by the Research Ethics Board of the McGill University Health Centre.
Stroke or TIA were ascertained with a positive answer to: “Has a doctor ever told you that you: have experienced a stroke or cerebrovascular accident (CVA); have experienced a mini-stroke or TIA; or suffer from the effects of a stroke, CVA, ministroke or TIA?”. Reference Raina, Wolfson and Kirkland21 PD (Parkinsonism) was identified based on an affirmative response to “Has a doctor ever told you that you had Parkinsonism or Parkinson’s Disease?” combined with either taking PD medications or a score ≥ 3 on the Tanner Questionnaire. Reference Raina, Wolfson and Kirkland21,Reference Oremus, Postuma and Griffith22 AD was identified with a positive answer to: “Has a doctor ever told you that you have dementia or Alzheimer’s disease?”. It should be noted that patients with major neurocognitive disorder are excluded from CLSA enrollment so AD patients would have mild impairment and not be representative of the whole AD population. Individuals were identified at baseline or at 3-year follow-up and were excluded from the general population.
Hypertension was defined as systolic blood pressure ≥ 140 mmHg or diastolic ≥ 90 mmHg (mean of 4 measures), history of hypertension or taking antihypertensive medications. Diabetes, depression, thyroid dysfunction, chronic obstructive pulmonary disease (COPD), chronic kidney disease (CKD) and heart disease were identified from the questionnaire item: ‘Has a doctor ever told you that you have…’. Head trauma was ascertained with the question: Have you suffered a head injury from any of the following…? Participants who answered “yes” for being a current smoker or using any type of tobacco products were categorized as smokers. Participants were categorized as regular alcohol drinkers if they drank ≥ 1 time per week in the past 12 months.
The STOP questionnaire (snoring, tiredness, observed apneas, hypertension) Reference Chung, Yegneswaran and Liao23 was included in the CLSA assessment. A score ≥ 2 is considered HROSA. The STOP-BAG Reference Waseem, Salama, Baltzan and Chung24 is based on the STOP-BANG Reference Nagappa, Liao and Wong25 and consists of the STOP with body mass index (BMI) > 35 kg/m2, age (>50 years) and gender (male). A score ≥ 3 indicates HROSA. The STOP-B28 is based on the STOP with BMI > 28 kg/m2. Reference Martins, Martinez, Cortes, Nascimento and Brendler26 The GOAL questionnaire consists of gender (males), obesity (≥30 kg/m2), age (≥50 years), and loud snoring Reference Duarte, Magalhães-da-Silveira, Oliveira, Silva, Mello and Gozal27 items, with a score ≥ 2 indicating HROSA. Reference Duarte, Magalhães-da-Silveira, Oliveira, Silva, Mello and Gozal27 Because sex and age are risk factors for the neurological conditions, and because hypersomnolence can be a feature of the disease process, screening scores in these conditions can be driven upwards by factors unrelated to apnea. Therefore, we analyzed specific symptoms (questionnaire items) individually. Participants who self-reported OSA under “Do you have any other long-term physical or mental condition that has been diagnosed by a health professional?” were categorized as having self-reported OSA.
Analyses
Patients with missing data were excluded. Descriptives statistics (mean, standard deviation, count, percentage) were used to report baseline characteristics of the study groups. We calculated the prevalence with 95% confidence intervals (CIs) of HROSA from each questionnaire and of individual OSA symptoms, for each of the groups (PD, AD, Stroke, GP). We compared the prevalence of HROSA and reported OSA within each group using Pearson chi-square test with False Discovery Rate corrections. Logistic regression was used to estimate odds ratios (ORs) for HROSA in neurological conditions versus the GP, unadjusted and adjusted for age, sex and BMI. We used analytic weights for better representation of the population in the regression analyses as provided by the CLSA. 28 Data were analyzed using SPSS 24.0 and R version 3.6. P < 0.05 was considered statistically significant.
Results
Baseline characteristics of participants included in our analyses are presented in Table 1. Among 30,097 participants included, 175 (0.58%) had PD, 125 (0.41%) had AD and 1791 (6.0%) had Stroke/TIA.
Table 1. Demographic characteristics of participants with and without neurological conditions in the Canadian longitudinal study on aging
PD = Parkinson’s Disease; AD = Alzheimer’s Disease; TIA = Transient Ischemic Attack; COPD = Chronic obstructive pulmonary disease; CKD = Chronic kidney disease; BMI = Body Mass Index.
Prevalence of symptoms and HROSA
The prevalence of snoring, sleepiness and observed apnea are shown in Table 2. Snoring was highly prevalent across all groups. Observed apnea was the second most prevalent variable in all groups. The combination of snoring and sleepiness presented a prevalence ranging from 3.5% to 6.5% across groups.
Table 2. Prevalence of OSA symptoms in neurological conditions and general population in the Canadian longitudinal study on aging
PD = Parkinson’s Disease; AD = Alzheimer’s Disease; TIA = Transient Ischemic Attack; N = total number of individuals in the population; CI = Confidence Interval.
We found a positive GOAL to have the highest prevalence in all groups (Table 3) and most notably in Stroke/TIA (81.1%) and PD (79.3%). Positive STOP-BAG and STOP-B28 occurred in similar proportions across all diagnostic groups. A positive STOP had the lowest prevalence of all questionnaires. OSA was self-reported in a much lower proportion than prevalence of HROSA by any questionnaire (Fig. 1).
Figure 1. Bar graph comparing the prevalence of high risk of OSA using different definitions in neurological conditions and the general population. The prevalence was statistically significantly different between all OSA measures (questionnaires or OSA diagnosis) within each study group, except for STOP-BAG vs. STOP-B28 in PD, AD and stroke. Differences within-group tested with Pearson chi-squared and post hoc false discovery rate correction.
Table 3. The prevalence of high risk of OSA using different definitions in neurological conditions and general population
PD = Parkinson’s Disease; AD = Alzheimer’s Disease; TIA = Transient Ischemic Attack; BMI = Body Mass Index; N = total number of individuals in the population; CI = Confidence Interval.
Neurological conditions vs. general population
To compare prevalence of symptoms, HROSA and OSA diagnosis between neurological disorders versus the GP, we performed logistic regression. In adjusted analyses, all OSA symptoms were more likely in individuals with stroke than the GP, including snoring with sleepiness. Participants with PD had higher odds of sleepiness than the GP. Those with AD had higher odds of observed apnea compared with the GP (Table 2). Table 4 shows ORs for HROSA in neurological disorders versus the GP. Individuals with stroke had higher adjusted odds of HROSA for each questionnaire. Individuals with PD showed higher odds for positive STOP-BAG and GOAL only. Those with AD had lower odds for positive STOP-B28 but higher for positive GOAL, compared with the GP. There were no significant differences between any neurological disorder and the GP with respect to self-reported OSA diagnosis.
Table 4. Logistic regression analysis comparing prevalence of high risk of OSA using different definitions in neurological conditions versus the general population
PD = Parkinson’s Disease; AD = Alzheimer’s Disease; TIA = Transient Ischemic Attack; OR = Odds ratio; OSA = Obstructive sleep apnea; CI = Confidence Interval.
1 OR values adjusted for age, sex and BMI
*Significant p-value at < 0.05.
Discussion
In this study, we found that different screening questionnaires result in a wide range of estimated prevalence of HROSA in older adults with and without neurological disorders, while OSA may be heavily underreported. Our findings indicate prevalence of questionnaire-based HROSA of 47%–81% in stroke, 37%–79% in PD and 31%–76% for AD (Table 3). Questionnaires in the somewhat younger GP were positive in 21%–61% of participants. Individuals with stroke were consistently more likely to have positive questionnaires than the GP. This corresponds to the high previously documented prevalence of OSA in stroke patients. Reference Liu, Lam, Chan, Chan, Ip and Lau9,Reference Hasan, Gordon and Wu29 Compared with the GP, individuals with PD were more likely to have a positive STOP-BAG and GOAL. Individuals with AD were more likely to have a positive GOAL, but less likely to have a positive STOP-B28.
The prevalence of OSA increases with age, estimated at more than 60% of people aged 65–99. Reference Laratta, Ayas, Povitz and Pendharkar30–Reference Ralls and Cutchen32 This is associated with age-related physiological changes such as decreased endurance and strength of head and neck muscles, hormonal, sleep, respiratory chemosensitivity alterations and comorbidities. Reference Ralls and Cutchen32 Several screening questionnaires have been developed to identify HROSA, which might be used to prioritize patients for sleep testing, or in epidemiological studies. Using the STOP with percentage of body fat, the estimated prevalence of OSA was 17.5% in the CLSA baseline cohort. Reference Thompson, Legault and Moullec33 However, only few studies have validated OSA screening tools in older populations or in neurodegenerative disorders. Reference Martins, Martinez, Cortes, Nascimento and Brendler26,Reference Godoy, Nucera, Colcher, de-Andrade and Alves34,Reference Gomes, Benedetti, Lafontaine, Kimoff, Robinson and Kaminska35 In the elderly, a lower BMI cutoff than used in middle-aged individuals has been found to be predictive of OSA, yielding an adapted tool, the STOP-B28. Reference Martins, Martinez, Cortes, Nascimento and Brendler26 In stroke patients, the STOP has been shown to have good characteristics with sensitivity 91% and specificity 93%. Reference Swartz, Cayley and Lanctôt36 The STOP-BAG had a sensitivity of 91% but lower specificity of 48%. Reference Katzan, Thompson, Uchino and Foldvary-Schaefer37 This suggests that the prevalence of HROSA ascertained by STOP, in individuals with stroke in our CLSA population, closely reflects the true prevalence of OSA in that group. OSA is known to be highly prevalent in stroke and an independent risk factor for stroke. Reference Xie, Zhu, Tian and Wang38 It may predate stroke, worsen during acute stages, and persist subsequently. Reference Young, Palta, Dempsey, Peppard, Nieto and Hla39 OSA is associated with recurrent stroke, cognitive decline and increased physical disability. Reference Liu, Lam, Chan, Chan, Ip and Lau9 Given the high prevalence of OSA and related adverse outcomes, it has been argued all stroke patients should be screened for OSA, irrespective of symptoms or other risk factors, and treated if OSA is found. Reference Boulos11 Conversely, prescreening with a questionnaire like STOP may help reduce testing burden and help efficient resources utilization.
In a previous study, Reference Gomes, Benedetti, Lafontaine, Kimoff, Robinson and Kaminska35 our group examined the diagnostic characteristics of OSA questionnaires in a PD clinical cohort. We found that GOAL ≥ 2 had the highest sensitivity (83%) but low specificity (35%), STOP-BAG ≥ 3 had sensitivity 74%, specificity 40%, STOP-B28 ≥ 2 sensitivity 76%, specificity 65% and STOP ≥ 2 sensitivity 65%, specificity 70%. These results are congruent with the current study, with the highest rate of HROSA identified by GOAL in each group, and the lowest with STOP. It is probable that GOAL and STOP-BAG are identifying more false positive cases given their lower specificity, especially given demographics of the population in question. Studies in PD have yielded inconsistent results with some studies finding that OSA is more common in PD and associated with its severity Reference Crosta, Desideri and Marini40 and others finding no association, Reference De Cock, Abouda and Leu41 possibly due to different OSA definitions, study populations or confounders such as medications. A recent meta-analysis found OSA to be present in close to 50% of PD patients, Reference Maggi, Giacobbe, Iannotta, Santangelo and Vitale42 and PD and non-PD risk factors were identified. The mechanisms contributing to the association between OSA and PD are not fully understood but there appears to be a causal bidirectional link. Reference Jeon, Hwang and Oh43 Treatment of OSA in PD has been found to be beneficial. Reference Kaminska, Mery and Lafontaine16,Reference Neikrug, Liu and Avanzino44 More research is needed to confirm OSA relationships and benefits of treatment but data to date suggest benefit to identify and treat OSA in PD when possible.
OSA is found with increased prevalence in patients with AD. Reference Emamian, Khazaie and Tahmasian45 Preexisting OSA is associated with increased risk of AD. Reference Osorio, Gumb and Pirraglia46 OSA treatment may improve outcomes. Reference Cooke, Ayalon and Palmer47,Reference Costa, Lim and Thorpe14 OSA is also a risk factor for hypertension, diabetes, depression and heart failure, all of which are common risk factors for AD. Reference Gaines, Vgontzas, Fernandez-Mendoza and Bixler48 The STOP-BANG has been evaluated in AD Reference Jorge, Benítez and Torres49 and found to have sensitivity 77% for at least moderate OSA but specificity of only 35%; modification of cutoffs for certain questions, e.g. age, BMI improved the characteristics of the questionnaire. This latter study and the one by Martins et al. Reference Martins, Martinez, Cortes, Nascimento and Brendler26 demonstrate that questionnaires developed for OSA screening of middle-aged individuals may not be well-suited for older individuals with or without neurodegenerative disorders. Moreover, cognitive decline may affect reliability of self-reported questionnaires. Currently, none of the commonly used questionnaires have been shown to have adequate diagnostic characteristics in AD, particularly in more advanced stages.
Participants with neurodegenerative conditions are more likely to be male and > 50 years old, which are both risk factors for OSA. Therefore, the GOAL and STOP-BAG, which include these two components, may not adequately distinguish those with OSA among individuals with neurodegenerative conditions (e.g. > 50% of PD patients would have a positive score based solely on sex and age). Snoring and BMI are best discriminatory factors for OSA in the GP. Reference Kolotkin, LaMonte, Walker, Cloward, Davidson and Crosby50,Reference Duarte and Magalhães-da-Silveira51 However, the threshold of 30 kg/m2 in GOAL and 35 kg/m2 in STOP-BAG may not be suitable for older individuals or those with neurological conditions. Reference Martins, Martinez, Cortes, Nascimento and Brendler26 We found all OSA symptoms to be more common in individuals with stroke compared with the GP (Table 5). Sleepiness was more common in individuals with PD. This is consistent with known high prevalence of sleepiness in PD, with OSA being one of multiple contributing causes. Reference Sobreira-Neto, Pena-Pereira and Sobreira52 Individuals with AD were more likely to have observed apneas compared with the GP. Though the confidence intervals are wide, the potential differences in symptoms further support the need for development and validation of specific tools for distinct patient populations.
Table 5. Logistic regression analysis comparing OSA symptoms in neurological conditions versus the general population
PD = Parkinson’s Disease; AD = Alzheimer’s Disease; TIA = Transient Ischemic Attack; OR = odds ratio; OSA = Obstructive sleep apnea; CI = Confidence Interval.
1 OR values adjusted for age, sex and BMI
*Significant p-value at < 0.05.
Given the significant impact of OSA on health and quality of life in aging individuals with neurological diseases, screening, diagnosis and treatment can have a meaningful impact on these patients, caregivers and the healthcare system. Previous studies estimated that over 80% of individuals of all ages with moderate-to-severe OSA remain undiagnosed and untreated. Reference Flemons, Douglas, Kuna, Rodenstein and Wheatley4 Only 3% of Canadians aged 18 years or older received a formal OSA diagnosis despite high rates of symptoms. 53 Our findings show that only a small proportion of older individuals in the CLSA cohort self-reported an OSA diagnosis despite having OSA symptoms (Fig. 1). Low diagnosis may have several causes such as poor symptom reporting to physicians, lack of access to medical care including specialist evaluation Reference Karunanayake, Rennie and Hagel54 and testing, lack of recognition of OSA as a possible diagnosis by physicians or perception that it is not worth pursuing in this population. Alternatively, a diagnosis may have been made but not reported by patients due to poor recall or understanding While not statistically significant, reported diagnosis of OSA in those with neurological disorders was lower than in the GP. We suspect that Canadians with neurological diagnoses might not be as frequently assessed for OSA while they might benefit most from treatment, this warrants further investigation.
Strengths and limitations
To date, our study is the first comparing different screening instruments for HROSA in neurological diseases, using a large population cohort that included the STOP assessment. Several limitations are noted. First, neurological conditions were self-reported. Second, there were no objective measures of OSA Reporting OSA symptoms and diagnosis depends on participant recall and could be underreported due to low perceived importance. However, this emphasizes the point that older individuals do not perceive OSA as a significant medical condition, despite the associated adverse outcomes There is also no information on the use of Continuous Positive Airway Pressure or other OSA therapy. Moreover, self-selection bias may have resulted in a study population not be fully representative of the Canadian population. Patients with moderate-severe dementia were excluded from the CLSA baseline evaluation, hence we cannot extrapolate results to advanced stages of the diseases. Participants in the CLSA were more educated, had higher household incomes, and were mostly White and Canadian-born, which may negatively influence the generalizability and external validity of results. Reference Raina, Wolfson and Kirkland55
Conclusion
In conclusion, we report, in a population-based cohort, that prevalence of HROSA in individuals with stroke, PD, AD and the GP, varies widely depending on screening tool. While all questionnaires suggest a higher prevalence of OSA in stroke/TIA than the GP, for PD and AD, there was wide variation of estimated prevalence depending on questionnaire. A self-reported diagnosis of OSA was found in a much smaller proportion of patients, suggesting that OSA is underdiagnosed or underreported in older adults. Strategies should be implemented to improve awareness of OSA in older Canadians and their healthcare providers with shared decision-making on testing and treatment.
Acknowledgements
We thank Dr Nadia Gosselin for her help with the CLSA dataset.
Author contributions
TG: design of the study, data preparation and analyses, interpretation of the data, manuscript writing, provided final approval for publication, and accepted accountability for all aspects of the work, AB: design of the study, help with data analyses, interpretation of the data, critical review, approval for publication and accepted accountability for all aspects of the work, RP: critical review, approval for publication and accepted accountability for all aspects of the work., DR: help with data, critical review, approval for publication and accepted accountability for all aspects of the work, MB: interpretation of the data, critical review, approval for publication and accepted accountability for all aspects of the work, RJK: interpretation of the data, critical review, approval for publication and accepted accountability for all aspects of the work, MK: study conceptualization, design and interpretation of the data, manuscript writing, critical review, approval for publication and accepted accountability for all aspects of the work.
Funding statement
CIHR Catalyst Grant: Analysis of Canadian Longitudinal Study in Aging (CLSA) Data (ACD-187256).
Competing interests
The author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the publication. Declaration for each author: TG: None, AB: None, RP: grants and personal fees from Fonds de la Recherche en Sante, grants from Canadian Institute of Health Research, grants from Michael J. Fox Foundation, grants from Webster Foundation, personal fees from Biogen, personal fees from Curasen, personal fees from Novartis, personal fees from Eisai, other from Parkinson Canada, grants from National Institute of Health, personal fees from International Parkinson and Movement Disorders Society, personal fees from Merck, personal fees from Vaxxinity, personal fees from Bristol Myers Squibb, personal fees from Clinilabs, personal fees from Ventus, personal fees from Korro, personal fees from Calico, outside the submitted work. DR: None, MB: research grants from Lady Davis Institute for Medical Research, Ministère des transports du Québec and the Réseau de recherche en sécurité routière, various speakers fees Eisai, Idorsia, GSK, Astra-Zeneca, Jazz Pharmaceuticals, Canadian Academy of Clinical Sleep Disorders Disciplines, FMOQ, FMSQ, CIUSSS Centre-Ouest de l’ile de Montréal, fiduaciary role in Canadian Academy of Clinical Sleep Disorders Disciplines, RJK: research grants from the Canadian Institutes of Health Research, The Multiple Sclerosis Society of Canada, Fonds de Recherche du Quebec – Santé, Research operating funds from Philips-Respironics Inc., Equipment support from ResMed Inc., and VitalAire Inc., Advisory board: Powell-Mansfield Inc., Eisai Inc., MK: Temple, Quebec Sleep Network, and various speaker fees, Research grants and in-kind support from Philips, the American Thoracic Society Foundation, Canadian Institutes for Health Research; Advisory Committee for Biron Soins du Sommeil; Advisory Board for Jazz Pharmaceuticals.
