Prevalence of common sleep disorders in a middle-aged community sample

Abstract

Study Objectives:

There is a paucity of contemporary prevalence estimates for common sleep disorders of insomnia, obstructive sleep apnea (OSA), and restless legs syndrome. We aimed to assess the prevalence of clinically significant common sleep disorders in a middle-aged community sample.

Methods:

Parents of participants in the community-based Raine Study underwent assessments between 2015 and 2017, including comprehensive questionnaires, anthropometric measures, and in-laboratory polysomnography. Clinically significant sleep disorders were defined as chronic insomnia using the Pittsburgh Sleep Symptom Questionnaire–Insomnia with duration criterion ≥ 3 months; OSA as apnea-hypopnea index ≥ 5 events/h with excessive sleepiness (Epworth Sleepiness Scale ≥ 11) or apnea-hypopnea index ≥ 15 events/h (even in the absence of symptoms); restless legs syndrome when participants endorsed the International Restless Legs Syndrome Study Group diagnostic criteria (2003) with symptoms ≥ 5 times/month involving moderate–severe distress.

Results:

At least 1 sleep-related assessment was completed by 1,005 (female = 586, 58.3%) middle-aged (45–65 years) participants, 72.5% of eligible Raine Study parents. The respective prevalences for clinically significant disease in females and males were as follows: OSA, 24.0% (95% confidence interval [CI]: 20.5–27.7) and 47.3% (95% CI: 42.2–53.4); insomnia, 15.8% (95% CI: 13.1–19.0) and 9.3% (95% CI: 6.8–12.4); restless legs syndrome, 3.7% (95% CI: 2.4–5.4) and 2.2% (95% CI: 1.1–3.9). At least 1 sleep disorder was present in 42.9% of those with complete data on all assessments (n = 895).

Conclusions:

Common sleep disorders are highly prevalent, to a clinically important extent, in an Australian community sample of middle-aged adults. Contemporary OSA prevalence is notably higher than previously reported and further work is needed to determine the communal impact of OSA.

Citation:

McArdle N, Reynolds AC, Hillman D, et al. Prevalence of common sleep disorders in a middle-aged community sample. J Clin Sleep Med. 2022;18(6):1503–1514.

BRIEF SUMMARY

Current Knowledge/Study Rationale: There is a lack of contemporary prevalence data regarding the most common sleep disorders: chronic insomnia, obstructive sleep apnea, and restless legs syndrome.

Study Impact: In this recent Australian community sample there was a very high prevalence of clinically significant sleep disorders in the middle-age adult years, with at least 1 sleep disorder present in approximately one-third of females and one-half of males. Clinically significant obstructive sleep apnea was the most common condition and further work is needed to establish the health impact of this high community prevalence. Such information is particularly important during middle age as this is a phase of high work-related productivity where the presence of these disorders is likely to have an exaggeratedly negative impact on community output.

INTRODUCTION

Insomnia, obstructive sleep apnea (OSA), and restless legs syndrome (RLS) are considered to be the most common sleep disorders in the general population. They are variously associated with reduced quality of life; adverse health sequelae, such as heart disease and depression; and increased mortality^1–4 A careful evaluation of the health impacts and the associated cost and resource implications of these disorders relies on accurate estimates of their prevalence in the community. Often, such population prevalence data are outdated and do not use recommended methodology or current definitions. To usefully guide health policy and health care planning, up-to-date prevalence data are needed, particularly when there is concern about changing prevalence rates.

The need for such data is especially relevant for common sleep disorders, with evidence that they are becoming more prevalent secondary to effects from Western lifestyles. For example, Peppard and colleagues⁵ modeled the impact of increased obesity on OSA prevalence over the last 2 decades and estimated relative increases of between 14% and 55%, depending on the sex and ages considered. Other contemporary changes, such as increased evening screen time and the emergence of a 24-hour society, may also predispose to increases in insomnia prevalence.^6,7

A recent evaluation of the financial costs associated with inadequate sleep in Australia⁸ was based on an online survey of a representative population sample.⁹ However, this survey contained selected questions about self-reported sleep disorder symptoms in place of validated survey instruments and no objective sleep study data were collected.⁹ While this study provides valuable data on sleep disorder prevalence, use of validated survey methods and objective sleep testing offer the prospect of greater accuracy.

Hence, there is a need for contemporary prevalence estimates for common sleep disorders in Australia, using both validated survey methodology and objective sleep study data. While there is a substantial literature on the overlap of OSA and insomnia, to our knowledge no studies have assessed the overlapping prevalence of the 3 most common sleep disorders (ie, insomnia, OSA, and RLS) using recommended methods. While we believe such information would be directly relevant in the Australian health care policy context, it is also likely to be applicable to other developed economies. The Raine Study completed comprehensive sleep assessments on a predominantly middle-aged sample of Western Australians between 2015 and 2017.¹⁰ The aim of this study was to utilize these data to provide a contemporary estimate of individual and overlapping prevalence of common sleep disorders in middle-aged (45–65 years) adults. We particularly focus on sex-specific prevalence rates, in view of the important role of sex on sleep disorder prevalence, and on clinically significant sleep disorders using current definitions.

METHODS

Study sample

We studied parents (Generation [Gen] 1) of participants (Gen 2) in the Raine Study, a longitudinal cohort study of the children of women who attended and gave birth to them at the major West Australian public obstetrics hospital between May 1989 and November 1991. These offspring (Gen 2) have been studied at regular intervals from in utero until young adulthood. Their parents (Gen 1) were recruited for comprehensive health and sleep evaluations between April 22, 2015, to June 16, 2017, at a time when the Gen 2 participants were an average of 26 years of age. An information pamphlet was posted to the Gen 1 parent cohort explaining the study aims and procedures and the areas of interest, which were respiratory, musculoskeletal, and sleep health. They were contacted by phone 2 weeks later and invited to take part in the ethically approved study (Human Research Ethics Committee, University of Western Australia, RA/4/1/7236). The participants provided informed consent prior to study procedures. A total of 1,098 Gen 1 participants were studied; however, this paper focuses only on the middle-aged subset (those aged 45–65 years, n = 1,022), as the number of participants outside this age range were limited.

Study procedures

The Raine Study Gen 1 participants underwent a comprehensive evaluation using survey methods, anthropometric measures, and level 1 in-laboratory polysomnography (PSG). Surveys contained detailed questions about lifestyle, medical conditions, and diagnostic features of sleep disorders and their associated symptoms using validated instruments (https://rainestudy.org.au/information-for-researchers/). From these evaluations, we established clinically significant sleep disorders in Gen 1 participants.

The prevalence of chronic insomnia was determined using the Pittsburgh Sleep Symptom Questionnaire–Insomnia (PSSQ-I).¹¹ The PSSQ-I incorporates Diagnostic and Statistical Manual of Mental Disorders (DSM), fourth edition, criteria to determine the presence of chronic insomnia, including a symptom duration requirement of at least 4 weeks. However, current definitions, as used in DSM-5, require a symptom duration ≥ 3 months. Therefore, we analyzed the PSSQ-I using a duration criterion of ≥ 3 months, instead of ≥ 4 weeks, to provide a prevalence estimate consistent with the current chronicity definition. We asked about RLS symptoms using the 2003 International Restless Legs Syndrome Study Group (IRLSSG) diagnostic criteria.¹² Specifically, we asked about the following: (1) an urge to move the legs when sitting or lying down, which was (2) accompanied by dysesthesia, (3) relieved by movement, and (4) worsened during the evening or night. A clinically significant RLS diagnosis was assigned when participants endorsed all 4 symptoms and had these symptoms 5 or more times per month involving moderate–severe distress, similar to the Epidemiology, Symptoms, and Treatment (REST) study criteria.¹³ Sleepiness was assessed by the Epworth Sleepiness Scale (ESS), which asks 8 questions about propensity to sleep in conditions of varying sleep-inducing potential, with a score range of 0– 24; scores ≥ 11 are considered to represent excessive sleepiness.¹⁴

PSG studies (Grael; Compumedics, Abbotsford, Australia) were conducted by trained sleep technologists in a bedroom-like environment at the Centre for Sleep Science, University of Western Australia. In particular, respiration was measured using both oronasal thermal airflow sensor (to score apnea) and nasal pressure transducer (to score hypopnea).¹⁵ Signals were analyzed according to the 2012 American Academy of Sleep Medicine (AASM) recommendations,¹⁵ and scorers were blinded to identifying or other sleep-related data. Of note, the 2012 AASM definition for hypopnea requires a ≥ 30% decrease in nasal flow from pre-event baseline associated with a ≥ 3% oxygen desaturation from pre-event baseline or an arousal. The presence of the cardinal OSA symptom of sleepiness is considered to be an important criterion for clinically significant disease likely to benefit from treatment.^14,16 For the purpose of this study, excessive daytime sleepiness was defined by ESS score ≥ 11. Hence, we defined clinically significant OSA as OSA of any severity—that is, apnea-hypopnea index (AHI) ≥ 5 events/h and excessive sleepiness (ESS ≥ 11) or AHI ≥ 15 events/h (even in the absence of symptoms). We present data at a periodic leg movement index (PLMI) cutoff of > 15 events/h.

Anthropometric assessments included height, measured using a stadiometer, and weight, measured using electronic scales, while wearing light clothing on the evening of the PSG.

Statistical methods

Continuous data are presented as mean ± standard deviation; categorical data are presented as numbers and percentages. Prevalence estimates and 95% confidence intervals (CIs) were calculated using frequencies and 2-sided binomial tests. Between-group comparisons were made using independent-sample t tests (continuous data), Pearson’s chi-square tests, or Fisher’s exact test (categorical data). For sleep disorders that had significant differences in unadjusted sex-stratified prevalence rates (P ≤ .10), Poisson regression models with robust error variances were used to investigate whether the associations between these disorders and sex remained after adjustment for body mass index (BMI) and education. Data analyses were conducted using IBM SPSS Statistics (version 27; IBM Corporation, Armonk, NY). A 2-tailed P value < .05 was considered statistically significant.

The Gen 1 sample from the Raine Study is a convenience sample of middle-aged adults. Due to differences in the age structure between this sample and the West Australian population, as determined by contemporaneous Census data,¹⁷ sex- and age-weighted averages were determined. Representativeness of the weighted sample was assessed by comparing key sociodemographic study sample variables with contemporaneous West Australian Census data¹⁷ and comparing obesity-related measures with National Health Survey data on middle-aged adults.¹⁸

RESULTS

Study participation

The study flow is presented in Figure 1. A total of 1,597 parents of the Gen 2 participants were eligible for the study, of whom 1,386 were in the in the “midlife” age range (45–65 years inclusive) on the date of first examination (April 2015). A total of 1,005 (72.5%) of the eligible middle-aged parents completed at least 1 sleep-related assessment. A total of 980 adults completed the general questionnaire, 941 attended a PSG, and 900 had analyzable PSG data. Of these 900 studies, 4 were excluded due to technical difficulties with breathing data. Complete data were available for overlapping prevalence estimates in 895 adults.

Figure 1. Study flow diagram.

PSG = polysomnography.

Participant characteristics

Sociodemographic characteristics of the participants are presented in Table 1. The mean (±standard deviation) age was 56.4 (±4.8) years and 601 (58.8%) were female. The majority of parents identified as Caucasian (91.8%), followed by Indian (3.0%) and Chinese (2.9%). Sleep architecture and sleep-related breathing data are provided in Table S1 ^{(786.1KB, pdf)} in the supplemental material. Briefly, study durations in the sleep laboratory environment were, on average (±standard deviation), 7.7 (±0.6) hours. Sex differences were observed for all sleep architecture variables, with the exception of stage N2 sleep (% of total sleep time) and rapid eye movement (REM) stage sleep (% of total sleep time). The average AHI during sleep was 14.7 ± 17.8 events/h (Table S1 ^{(786.1KB, pdf)} ), with males having an event frequency almost twice that of females. Average PLMI (events/h) was 16.2 events/h with no difference between the sexes.

Table 1.

Comparison of cohort participants with Western Australian Census data (2016 Australian Bureau of Statistics community profile for Western Australia) and Western Australian data from the Australian Health Survey (2011–2012).

	Raine Study Gen 1 Follow-up, Main Questionnaire Completion						Western Australian Population Census (2016), ABS Community Profile Dataa,b
	Unweighted			Weighted			All Individuals	Females	Males
	All Individuals	Females	Males	All Individuals	Females	Males
Characteristics of Complete Gen 1 Sample, n	1,022	601	421	1,033	614	419	457,190–1,091,870	213,305–506,003	243,885–585,867
Age (years)
45–49	118 [11.5]	87 [14.5]	31 [7.4]	272 [26.3]	161 [26.3]	111 [26.4]	172,520 [26.9]	85,943 [26.7]	86,850 [27.1]
50–54	275 [26.9]	173 [28.8]	102 [24.2]	246 [23.8]	151 [24.6]	95 [22.7]	162,438 [25.3]	80,893 [25.1]	81,546 [25.5]
55–59	380 [37.2]	217 [36.1]	163 [38.7]	228 [22.1]	132 [21.6]	96 [22.9]	149,899 [23.4]	75,781 [23.5]	74,118 [23.2]
60–65	249 [24.3]	124 [20.7]	125 [29.7]	287 [27.8]	170 [27.6]	117 [28.0]	156,716 [24.4]	79,621 [24.7]	77,098 [24.1]
Missing	0	0	0	0	0	0
General questionnaire, n	980	579	401	984	588	386
Marital status
Not married	281 [28.8]	195 [33.8]	86 [21.6]	303 [30.9]	209 [35.6]	94 [23.8]	198,204 [27.8]	113,165 [30.9]	85,037 [24.6]
Registered or de facto married	695 [71.2]	382 [66.2]	313 [78.4]	677 [69.1]	377 [64.4]	300 [76.2]	514,501 [72.2]	253,516 [69.1]	260,969 [75.4]
Missing or not applicable	4	2	2	4	2	2
Education completed
≥ Secondary (year 9 or above)c	965 [99.2]	570 [99.0]	395 [99.5]	970 [99.1]	579 [98.8]	391 [99.6]	815,969 [87.0]	410,405 [87.7]	405,564 [86.3]
Missing	7	3	4	5	2	3	71840	33485	38355
Qualifications
Postsecondary/equal to tertiary	364 [37.4]	209 [36.3]	155 [39.0]	334 [34.2]	198 [33.6]	137 [34.8]	154,164 [30.7]	80,779 [35.2]	73,385 [27.0]
Missing	7	3	4	5	2	3	815,777	40,380	41,197
Income levelsd,e
Low	181 [18.9]	142 [25.0]	39 [10.0]	182 [19.0]	139 [24.1]	43 [11.3]	241,931 [33.3]	164,355 [43.7]	77,576 [22.1]
Medium	272 [28.4]	189 [33.3]	83 [21.3]	281 [29.2]	191 [33.0]	90 [23.4]	241,620 [33.3]	133,698 [35.6]	107,922 [30.8]
High	421 [43.9]	175 [30.8]	246 [63.1]	407 [42.3]	176 [30.5]	231 [60.2]	242,948 [33.4]	77,911 [20.7]	165,037 [47.1]
Missing	22	11	11	23	11	13	714,441	33,418	38,023
Labor forcef
Employed	745 [76.4]	430 [74.4]	315 [79.1]	746 [76.1]	431 [73.3]	315 [80.2]	464,974 [61.2]	216,147 [56.2]	248,827 [66.3]
Unemployed/not in labor force	231 [23.7]	147 [25.6]	83 [20.9]	234 [23.9]	156 [26.7]	78 [19.8]	295,152 [38.8]	168,556 [43.8]	126,596 [33.7]
Missing	5	2	3	4	1	3	56,541	26,129	30,412
Occupationg
Professional/managerial	345 [46.7]	177 [41.5]	168 [53.8]	334 [45.2]	171 [40.0]	164 [52.3]	166,848 [36.5]	73,915 [34.7]	92,933 [38.1]
Clerical	123 [16.7]	110 [25.8]	13 [4.2]	122 [16.5]	109 [25.5]	14 [4.4]	70,421 [15.4]	55,335 [25.9]	15,086 [6.2]
Technical/trade/labor	93 [12.6]	26 [6.1]	67 [21.5]	95 [12.9]	27 [6.4]	68 [21.7]	105,321 [23.0]	27,928 [13.1]	77,393 [31.7]
Other	177 [24.0]	113 [26.5]	64 [20.5]	187 [25.3]	120 [28.1]	68 [21.6]	114,600 [25.1]	56,127 [26.3]	58,473 [24.0]
Missing	5	2	3	4	1	2	7,784	2,848	4,936
Work (hours per week)h
< 40	450 [60.6]	330 [76.9]	120 [38.3]	439 [59.1]	324 [75.4]	115 [36.6]	575,375 [52.7]	354,571 [65.9]	220,804 [35.6]
≥ 40	292 [39.4]	99 [23.1]	193 [61.7]	304 [40.9]	106 [24.6]	199 [63.4]	516,495 [47.3]	151,432 [28.1]	365,063 [58.9]
Missing	3	1	2	2	1	2	19,132	7,164	11,962
	Gen 1 Cohort, PSG Completion						Western Australian Population, Australian Health Survey (2011–2012), ABSa
	Unweighted			Weighted			All (n = 115,0500)	Females (n = 575,700)	Males (n = 574,800)
	All Analyzable PSG (n = 896)	Females (n = 527)	Males (n = 369)	All Analyzable PSG (n = 903)	Females (n = 537)	Males (n = 366)
Underweight (BMI < 18.5 kg/m2)	6 [0.7]	6 [1.1]	0 [0.0]	7 [0.7]	7 [1.2]	0 [0]
Normal (BMI 18.5–24.99)i	249 [27.8]	174 [33.0]	75 [20.3]	247 [27.4]	175 [32.6]	72 [19.7]	180,200 [28.0]	109,600 [35.1]	70,600 [21.3]
Overweight (BMI 25.00–29.99)	361 [40.3]	186 [35.3]	175 [47.4]	348 [38.6]	179 [33.3]	170 [46.4]	259,800 [40.3]	115,800 [37.0]	144,100 [43.5]
Obese (BMI ≥ 30.00)	279 [31.1]	161 [30.6]	118 [32.0]	300 [33.3]	177 [32.9]	124 [33.9]	203,900 [31.7]	87,200 [27.9]	116,800 [35.2]
Missingj	1 [0.1]	0 [0]	1 [0.3]	1	0	1	151,300	87,800	63,500

Data are presented as n [%] unless otherwise indicated. ^aCensus denominators are based on the denominator for ages 45–74 years in accordance with 10-year groups presented in ABS Census data by sex. Data for age were available by year. ^bCensus denominators varied between the ranges provided due to varying respondent numbers for census items. ^cRaine Gen 1 cohort data reflect any secondary education, while ABS Census Data include ≥ Year 9 education. ^dRaine Gen 1 income categories: low (≤ $31,999), medium ($31,200–$64,999), high (≥ $65,000); ABS income categories: low (≤ $25,999), medium ($26,000–$64,999), high (≥ $65,000). ^eDenominator includes respondents who reported no personal income, or who reported negative/nil income for the ABS Census Data. ^fDenominator includes “labor force not stated” responses for the ABS Census Data. ^gDenominator is only for participants who indicated they are active in the labor force. ^hData are for employed persons aged ≥ 15 years in ABS Census Data, with a denominator which includes workers who reported no hours in the week prior to Census night and any not stated responses. ⁱData from ABS Australian Health Survey (2011–2012) are grouped as underweight/normal, and represented in the normal row. ^jMissing represents “number not measured” for ABS Australian Health Survey (2011–2012) data. ABS = Australian Bureau of Statistics, BMI = body mass index, PSG = polysomnography.

Representativeness

Comparisons of sociodemographic and BMI characteristics with Western Australian Census data are shown in Table 1 (both unweighted and weighted for age and sex). A number of variables have significant proportional differences. In particular, Raine Study participants had a 12.2% higher proportion of individuals with at least a secondary education, 14.4% fewer in the low-income category, 15.2% more individuals were employed, and, in those employed, there was a 10.4% lower proportion of technical/trade occupations compared with the general population.

Prevalence estimates

Table 2 presents unadjusted weighted prevalence estimates for common sleep disorders for the total sample. Adjusted comparisons between females and males for each disorder are presented in Table 3.

Table 2.

Unadjusted prevalence of common sleep disorders in the cohort participants.

	All Participants, n [%, 95% CI]	Females, n [%, 95% CI]	Males, n [%, 95% CI]	P a	Missing Number [%]
Sleep questionnaires, n [%]	1016 [100]	598 [58.9]	418 [41.1]
Clinically significant chronic insomniab	134 [13.1, 11.2–15.4]	95 [15.8, 13.1–19.0]	39 [9.3, 6.8–12.4]	.001	42 [4.1]
Excessive sleepiness, ESS ≥ 11	130 [12.8, 10.8–15.0]	72 [12.0, 9.6–14.8]	58 [13.8, 10.8–17.4]	.45	46 [4.6]
Severe sleepiness, ESS ≥ 16	19 [1.9, 1.2–2.8]	11 [1.9, 1.0–3.2]	8 [1.9, 0.9–3.6]	.97	46 [4.6]
RLS diagnosisc	142 [14.0, 11.9–16.2]	93 [15.6, 12.8–18.6]	49 [11.8, 8.9–15.1]	.11	44 [4.3]
Clinically significant RLSd	31 [3.1, 2.1–4.2]	22 [3.7, 2.4–5.4]	9 [2.2, 1.1–3.9]	.23	44 [4.3]
PSG-based assessments,e n [%]	903 [100]	537 [59.5]	366 [40.5]
PSG-based leg movements
PLMI > 15 events/h	188 [20.8, 18.3–23.6]	102 [18.9, 15.8–22.5]	86 [23.6, 19.4–28.0]	.10	0 [0.0]
PLMI > 15 events/h and RLS diagnosis	41 [4.5, 3.3–6.0]	24 [4.4, 3.0–6.5]	17 [4.8, 2.8–7.2]	.77	39 [4.4]
PSG-based sleep-disordered breathing
OSA (AHI ≥ 5 events/h)	629 [69.6, 66.6–72.6]	313 [58.2, 54.1–62.4]	316 [86.2, 82.5–89.6]	< .001	0 [0]
Moderate-severe OSA (AHI ≥ 15 events/h)	264 [29.3, 26.3–32.3]	109 [20.3, 17.1–23.9]	155 [42.4, 37.4–47.5]	< .001	0 [0]
OSA with excessive sleepiness (AHI > 5 events/h and ESS ≥ 11)	84 [9.3, 7.5–11.3]	40 [7.5, 5.5–9.9]	44 [12.1, 9.0–15.6]	.026	6 [0.6]
Clinically significant OSA (AHI ≥ 5 events/h and ESS ≥ 11 or AHI > 15 events/h)	302 [33.5, 30.5–36.6]	128 [24.0, 20.5–27.7]	173 [47.3, 42.2–53.4]	< .001	2 [0.02]

Data used in these analyses are weighted prevalence data from the Raine study Gen 1 cohort. Data include number as a percentage of all data for the relevant questionnaire or assessment, including missing values. ^aP values express the difference between males and females using Pearson’s chi-square for binary variables. ^bChronic insomnia was defined using the PSSQ-I, with modification of duration criteria from ≥ 4 weeks to ≥ 3 months. ^cRestless legs syndrome diagnosis was defined according to IRLSSG criteria.^{12 d}Clinically significant RLS was defined according to IRLSSG criteria¹² and having symptoms 5 or more times per month involving moderate–severe distress. ^eSample size reflects participants who underwent in-laboratory PSG and had sufficient data to facilitate analysis. AHI = apnea-hypopnea index, CI = confidence interval, ESS = Epworth Sleepiness Scale, IRLSSG = International Restless Legs Syndrome Study Group, OSA = obstructive sleep apnea, PLMI = periodic leg movement index, PSG = polysomnography, PSSQ-I = Pittsburgh Sleep Symptom Questionnaire–Insomnia, RLS = restless legs syndrome.

Table 3.

Association of sleep disorders with sex, after adjustment for age, BMI, and education.

	Prevalence Ratio*	95% CI	P
Insomnia diagnosisa
Females	Reference (1.00)
Males	0.59	0.38–0.92	.019
PLMI > 15 events/h
Females	Reference (1.00)
Males	1.33	0.98–1.81	.07
OSA (AHI ≥ 5 events/h)
Females	Reference (1.00)
Males	1.50	1.37–1.64	<.001
Moderate–severe OSA (AHI ≥ 15 events/h)
Females	Reference (1.00)
Males	2.33	1.85–2.92	<.001
OSA with excessive sleepiness (AHI ≥ 5 events/h and ESS > 11)
Females	Reference (1.00)
Males	1.66	1.00–2.77	.05
Clinically significant OSA (AHI ≥ 5 events/h and ESS ≥ 11 or AHI ≥ 15 events/h)
Females	Reference (1.00)
Males	2.16	1.76–2.66	<.001

*Weighted Poisson regression models with robust error variances were used to determine prevalence ratios, 95% CIs, and P values for each sleep disorder by sex. Sleep disorders were modeled separately, with each model adjusted for BMI (continuous variable), age (continuous variable), and level of education (secondary school/TAFE or college/university/other). ^aInsomnia was defined using the PSSQ-I, with modification of duration criteria from ≥ 4 weeks to ≥ 3 months. AHI = apnea-hypopnea index, BMI = body mass index, CI = confidence interval, ESS = Epworth Sleepiness Scale, OSA = obstructive sleep apnea, PLMI = periodic leg movement index, PSSQ-I = Pittsburgh Sleep Symptom Questionnaire–Insomnia, TAFE = Technical and Further Education.

OSA

According to sleep breathing variables derived from PSG, OSA of any severity (AHI ≥ 5 events/h) was highly prevalent. Specifically, OSA of any severity occurred in more than half of females (58.2%, n = 313/537; 95% CI: 54.1–62.4%) and more than three-quarters of males (86.2%, n = 316/366; 95% CI: 82.5–89.6%). Moderate to severe OSA (AHI ≥ 15 events/h during sleep) occurred in one-fifth of females (20.3%, n = 109/537; 95% CI: 17.1–23.9%) and almost half of males (42.4%, n = 155/366; 95% CI: 37.4–47.5%). OSA of any severity with excessive sleepiness (AHI ≥ 5 events/h and ESS ≥ 11) was more common in males (12.1%, n = 44/366; 95% CI: 9.0–15.6%) than in females (7.5%, n = 40/537; 95% CI: 5.5–9.9%), as was clinically significant OSA, which occurred in 47.3% (n = 173/366; 95% CI: 42.2–53.4%) of males and 24.0% (n = 128/537; 95% CI: 20.5–27.7%) of females. Significant sex differences in weighted prevalence rates for OSA remained when adjusting for age, BMI, and education (Table 3).

Insomnia

Self-reported diagnosis of insomnia occurred in just over 1 in 7 participants overall. Weighted prevalence rates for females (15.8%, n = 95/598; 95% CI: 13.1–19.0%) were significantly higher than for males (9.3%, n = 39/418; 95% CI: 6.8–12.4%). The difference between sexes remained significant after adjustment for age, BMI, and education (Table 3).

RLS and periodic leg movements

The unadjusted, weighted prevalence of RLS as defined by “symptoms-only” was 14.0% (n = 142/1,016; 95% CI: 11.9–16.2%) and clinically significant RLS, at a frequency and severity likely to require medical treatment, occurred with a prevalence of 3.1% (n = 31/1,016; 95% CI: 2.1–4.2%). A PLMI > 15 events/h was observed in 1 in 5 participants (20.8%, n = 188/903; 95% CI: 18.3–23.6%). RLS defined by symptoms-only with PLMI > 15 events/h occurred in 4.5% of the sample and clinically significant RLS with PLMI > 15 events/h occurred in 1.1% of the sample. Prevalence rates for all RLS and PLMI variables were similar between females and males.

Overlapping sleep disorders

Figure 2 presents the overlap of clinically significant sleep disorders, based on participants who completed PSG and self-report measures. Although approximately 8% of the Gen 1 cohort were found to have isolated chronic insomnia, a further 4% of insomnia overlapped with clinically significant OSA (Table S2 ^{(786.1KB, pdf)} ). Similarly, while 0.7% had isolated clinically significant RLS, a further 1.3% of RLS overlapped with clinically significant OSA (Table S2 ^{(786.1KB, pdf)} ). Three overlapping disorders were rare. Overlaps between sleep disorders, using broader disease definitions such as mild asymptomatic OSA (AHI ≥ 5 events/h) and PLMI > 5 events/h, in line with previous sleep disorder reports from the Raine Study Gen 2, are shown in Figure S1 ^{(786.1KB, pdf)} in the supplemental material. Additional details on overlapping prevalence, including prevalence for females and males combined, using clinically significant and broader disease definitions can be found in Table S2 ^{(786.1KB, pdf)} and Table S3 ^{(786.1KB, pdf)} , respectively.

Figure 2. Prevalence and overlap of clinically significant sleep disorders in the subset of middle-aged adults with complete data on all sleep-related assessments (n = 895).

(A) Females. (B) Males. Each “person” represents a prevalence of 1%. Prevalence estimates are rounded to the nearest percent. Overlap between 2 disorders (ie, a person having multiple sleep disorders) is represented by a “person” shaded in the colors of each disorder. Clinically significant OSA was defined as OSA of any severity with excessive sleepiness (AHI ≥ 5 events/h and ESS ≥ 11) or AHI ≥ 15 events/h (even in the absence of symptoms), Clinically significant chronic insomnia was defined using the PSSQ-I, with modification of duration criteria from ≥ 4 weeks to ≥ 3 months. Clinically significant RLS was defined according to IRLSSG criteria when participants had symptoms 5 or more times per month involving moderate–severe distress. AHI = apnea-hypopnea index, ESS = Epworth Sleepiness Scale, IRLSSG = International Restless Legs Syndrome Study Group, OSA = obstructive sleep apnea, PSSQ-I = Pittsburgh Sleep Symptom Questionnaire–Insomnia, RLS = restless legs syndrome.

DISCUSSION

We estimated the prevalence of common sleep disorders in middle-aged adults in a community representative of those in a Western country, with a focus on clinically significant disease, likely to benefit from medical treatment. Contemporary gold-standard methods and current definitions were used to make accurate estimates. The prevalence of common sleep disorders was found to be very high in middle-aged individuals. OSA of any severity (AHI ≥ 5 events/h) was the most prevalent condition (69.6%); however, using conservative consensus definitions to define clinically significant disease, this estimate decreased (33.5%) but remained the most prevalent sleep disorder, followed by chronic insomnia (13.1%) and clinically significant RLS (3.1%). To our knowledge, this is the only study that provides contemporaneous high-quality prevalence data on all 3 common sleep disorders, allowing estimation of the degree of overlap between these disorders. When overlap was taken into account, the prevalence of the presence of “any” clinically significant sleep disorder in an individual was 42.9% (36.9% in females and 51.6% in males). We also confirm the well-described associations of sex with common sleep disorders. Common sleep disorders are associated with significant morbidity and health care costs and these data are important for informing health care providers in decisions about appropriate resourcing for their effective assessment and treatment.

Representativeness

In this article we report data on one or both biological parents of the Raine Study participants. The age structure of the participating parents differed from the general population (Table 1, unweighted data); hence, the study sample was weighted by age and sex to match the Western Australian Census data, and all subsequent analyses were conducted on the weighted sample. Comparisons of key sociodemographic characteristics in the Raine Study midlife participants with contemporaneous Western Australian Census data show a pattern consistent with many other epidemiological studies where participation rates are higher among those with greater education, employed, and with higher socioeconomic status.¹⁹ The likely implication of a higher socioeconomic status is that prevalence estimates are likely to be conservative and tend toward lower values than actual population prevalence for insomnia^20,21 and RLS.^22–24 However, there appears to be no significant independent effect of socioeconomic status on OSA prevalence.^25,26 Importantly, the prevalence of obesity, a key determinant of OSA, is closely matched between the Raine Study sample and the general population (Table 1).

There is a risk that people with sleep-related problems may have been more likely to participate in this study and complete sleep assessments, compared with nonparticipants. However, we believe this is unlikely to introduce a major bias, since the patient information sheet indicated that assessments were broader than sleep and designed to assess a number of health issues (eg, vision and respiratory health), reducing the risk of selection bias. Furthermore, the percentage of midlife participants who completed at least 1 sleep assessment was a high proportion of all Raine midlife participants (1,005/1,022 = 98.3%) and a reasonably high proportion of the eligible Raine midlife parents (1,005/1,386 = 72.5%). As indicated above, comparisons of sociodemographic characteristics between all Raine midlife participants and the general population suggest that our sleep disorder prevalence estimates are likely to be conservative.

Prevalence of OSA

OSA of any severity (AHI ≥ 5 events/h) was found in a very high percentage of the study sample—almost 6 in 10 females and 9 in 10 males. Most previous OSA community prevalence studies using PSG measurements were made using old technology and we are aware of only 2 studies using similar modern PSG diagnostic approaches, which include a sensitive nasal pressure signal to detect hypopneas. The more recent of these studies recruited representative Swiss community participants between 2009 and 2013, aged 35–75 years, with mild obesity levels (mean BMI of 25.1 kg/m² in females and 26.2 kg/m² in males).²⁷ Participants were assessed using home PSG monitoring with the same sensitive measuring techniques and contemporary scoring criteria (AASM 2012) as used in our study.¹⁵ Their prevalence estimates for any OSA were very similar to our estimate, occurring in 60.8% in females and 83.8% in males, despite participants having a slightly lower mean BMI (Gen 1 mean BMI: females, 28.6 ± 6.5; males, 28.7 ± 4.5 kg/m²). The similarity of the Australian and Swiss prevalences contrasts with those from Tufik and colleagues²⁸ who studied 1,042 representative Brazilians in 2007 using in-laboratory PSG and a nasal pressure signal. Any OSA was found in 30.5% of females and 46.5% of males, which is approximately half the rate, for each sex, found in the current study. However, that study used older, 2007 AASM, respiratory event scoring rules, which are stricter, and a broader age range (20–80 years) than the Australian and Swiss studies, which may have contributed to their lower prevalence estimates. Furthermore, their participants had less obesity, with 59.9% of their sample being overweight or obese compared with 71.9% of our sample.

Moderate–severe OSA (AHI ≥ 15 events/h) is considered likely to benefit from treatment.¹⁶ Even moderate–severe disease was highly prevalent in our sample, occurring in 20.3% of females and 42.4% of males, which is similar to the prevalence found in the Swiss sample of 23.4% in females and 49.7% in males.²⁷ In Brazil, moderate-to-severe OSA was present in 9.6% of females and 24.8% of males, again about half the rate, for each sex, found in the current study.²⁸ OSA of any severity with excessive sleepiness (ie, AHI ≥ 5 events/h and ESS ≥ 11) was found in 7.5% of females and 12.1% of males. Clinically significant OSA was present in approximately one-quarter of females and one-half of males in middle age.

Hence, in an Australian sample, we confirm the findings of other contemporary international studies of a concerningly high prevalence in the community of clinically significant OSA likely to benefit from treatment. Only 11.6% of participants with clinically significant OSA reported being recommended or prescribed continuous positive airway pressure at the time of the study (all underwent PSG without continuous positive airway pressure), highlighting the previously documented underdiagnosis and undertreatment of OSA in the community.^29,30

Prevalence of insomnia

We found chronic insomnia to be common in middle-aged adults, occurring in 13.1% of adults in middle-aged years: 15.8% of females and 9.3% of males. An older (2006–2008) Norwegian study using DSM-5 criteria estimated an insomnia prevalence of 7.9% for the general adult population aged ≥ 20 years.³¹ Their data for those between 40 and 69 years of age (ie, comparable to the Gen 1 age range) were 8.0% in females and 5.6% in males.³¹ A more recent study (2012) from Hong Kong estimated the prevalence of chronic insomnia using DSM-5 criteria for all adults > 18 years of age of 10.8%.³² Data were not provided separately by sex, but the Hong Kong estimate for the combined sex rate is similar to the overall prevalence of 13.1% observed in the Raine Study Gen 1 sample. Comparison with the Hong Kong study is complicated by the lack of data provided separately for the middle-age range sampled in the current study. Overall, our more recent prevalence estimate for chronic insomnia is more in keeping with the recent data from Hong Kong than older data from Norway, which raises the possibility that the prevalence of insomnia may have increased over time, perhaps partly attributable to societal lifestyle changes.

Prevalence of RLS

Current IRLSSG (2012) diagnostic criteria for RLS specify that the RLS symptoms are not solely accounted for by another medical or behavioral condition (eg, leg cramps or positional discomfort).³³ For simplicity (4 IRLSSG questions vs a more extensive exploration of this facet of assessment) and to facilitate comparisons with other studies,^13,34–36 we did not assess for possible diagnostic mimics of RLS here. As such, some other conditions may have been miscategorized as RLS in this study, leading to an overestimate of RLS prevalence.

Using the 2003 IRLSSG “symptom-only” criteria to “diagnose” RLS we found a prevalence of 14.0% in individuals aged 45–65 years. A 2012 review of 35 community-based studies using 1995 or 2003 IRLSSG symptom-only criteria reported RLS prevalences ranging from 3.9% to 14.3%.³⁵ Two of the largest European and North American community-based studies (Asian countries appear to have lower RLS prevalence rates) of those aged ≥ 18 years using 2003 IRLSSG criteria found a lower prevalence than our study, possibly due to their inclusion of younger participants, who tend to have a lower prevalence. In particular, the REST study of 15,391 participants in United States and Europe found a prevalence of 7.2%¹³ and a more recent study of 4,003 Turkish individuals found a prevalence of 7%.³⁶ Prevalence rates for the 45- to 65-year age range were not reported in those studies.

In the REST study expert opinion was used to define clinically significant RLS likely to warrant treatment by including symptom frequency and severity (“distress”) criteria.¹³ In that study the prevalence of clinically significant RLS was 2.7% in all adults (age ≥ 18 years) and approximately 3.3% for the age range of 40–69 years. The prevalence of clinically significant RLS in the REST study in middle age (3.3%) is similar to the Raine Study estimate of 3.1% using similar criteria.

Periodic leg movements are considered to be a supportive objective marker of RLS, with a PLMI > 5 events/h occurring in 80% and a PLMI > 10 events/h in 67.9% of clinical cases.³⁷ The few community-based studies with data on RLS and periodic leg movements also found these 2 conditions to be associated.^38,39 Of participants who had valid PSG and RLS symptoms data, we found a PLMI > 15 in 32.5% of those with symptom-only RLS diagnosis, compared with 19.1% of those without (χ²₁ = 11.68, P < .001) .

The clinical relevance of the PLMI, apart from its association with RLS, is unclear, but there is increasing evidence that periodic leg movements may be a risk marker for cardiovascular disease.³⁹ We found a PLMI > 15 events/h in 20.8% of the total group (mean age = 55.3 years). There are few comparable community-based studies reporting PLMI. The prevalence of PLMI ≥ 15 events/h was 26.3% in a community sample of Wisconsin employees aged 40–70 years (mean age = 56.1 years)³⁹ and 28.6% in a Swiss community sample of individuals aged 35–75 years (mean age = 58.4 years).³⁸ These prevalence rates align closely with our results, taking into account the known increase in PLMI with age.

Sex differences

We confirm consistent findings from previous studies that OSA is more prevalent⁴⁰ and insomnia less prevalent in males compared with females⁴¹ in middle-aged adults. These sex differences persisted after adjustment for important confounders of age, BMI, and education (Table 3). The finding that RLS was not different between sexes in our study differs from many,¹³ but not all,³⁶ other community-based studies, which report a higher prevalence in females. Our finding that PLMI > 15 events/h tended toward a higher prevalence in males in the Gen 1 study sample is consistent with previous reports.^38,39

Overlapping conditions

There is increasing interest in the overlap between sleep disorders, which may represent subtypes with distinct pathophysiological mechanisms and clinical consequences. For example, the overlap of insomnia and OSA is associated with greater impairment in daytime function and reduced quality of life compared with either condition alone.⁴² Furthermore, identifying this subtype appears to have treatment implications.⁴³ In a recent meta-analysis of published studies, insomnia was estimated to occur in 38% of individuals with OSA⁴⁴; however, none of those studies used DSM-5 insomnia criteria and current OSA scoring criteria. In our study, insomnia occurred in 13.0% of those with clinically significant OSA and the overlap of clinically significant OSA occurred in 33.1% of those with insomnia. We also found that clinically significant RLS was more common in association with clinically significant OSA (1.3%) than as an isolated condition (0.7%) (Table S2 ^{(786.1KB, pdf)} ); however, caution in interpretation is needed as case numbers are small.

Limitations

A potential limitation of these data is the use of a convenience sample of parents from an existing, longitudinal study. However, the children who are current Raine Study Gen 2 participants are broadly representative of the source population and it was anticipated that the parents would be similarly representative. Comparisons of our age- and sex-weighted data with Census data show characteristics of the Gen 1 Raine Study sample are consistent with a higher socioeconomic status than the general population, which is likely to mean that the Gen 1 prevalence estimates are underestimates of the true prevalence rather than an overestimate. The prevalence of sleep disorders varies considerably by age and the current report only relates to the middle-aged adult population because our sample had few younger or older adults. However, we believe such a focus on the middle-age range is necessary as this is a pivotal group/sector in the workforce, given their experience, potential for ongoing workforce participation, and opportunity for intervention. Furthermore, the effect of age on common sleep disorders is generally well established, allowing potential for modeling of prevalence at other ages using our middle-aged sample and our previously published Raine young-adults data (average age = 22 years).³⁴

CONCLUSIONS

Clinically significant sleep disorders are highly prevalent in this middle-aged Western community sample. The prevalence of clinically significant OSA is most notable, both as an isolated disorder and as an overlapping condition with insomnia and RLS. The consistently high contemporary OSA prevalence estimates are underappreciated and more work is needed to better understand the impact of this high prevalence on human health.

ABBREVIATIONS

AHI

apnea-hypopnea index

BMI

body mass index

DSM

Diagnostic and Statistical Manual of Mental Disorders

ESS

Epworth Sleepiness Scale

Gen

Generation (1 or 2 of Raine Study)

IRLSSG

International Restless Legs Syndrome Study Group

OSA

obstructive sleep apnea

PLMI

periodic leg movement index

PSG

polysomnography

PSSQ-I

Pittsburgh Sleep Symptom Questionnaire–Insomnia

RLS

restless legs syndrome

DISCLOSURE STATEMENT

All authors have seen and approved the manuscript. The lead author (N.M.) affirms that the manuscript is an honest, accurate, and transparent account of the study being reported and that no important aspects of the study have been omitted. Work for this study was performed at University of Western Australia, School of Health Sciences, Nedlands, Australia. The Raine Study has been supported by the National Health and Medical Research Council (NHMRC) over the last 29 years with additional funding for core management provided by the University of Western Australia (UWA); Raine Medical Research Foundation; Telethon Kids Institute; UWA Faculty of Medicine, Dentistry and Health Sciences; Women and Infants Research Foundation; Curtin University; Edith Cowan University; Murdoch University; and the University of Notre Dame Australia. This year 26 follow‐up was made possible by the Raine Study team and sleep study technicians, The Centre for Sleep Science UWA, and Safework Australia, and supported by the National Health and Medical Research Council (ID 1021858, ID 1027449, ID 1044840, and ID 1084947). Peter Eastwood is supported by a NHMRC Senior Research Fellowship (1136548). N.M., P.E., D.H., and K.M. have received income to their institution for sponsored trials from Oventus Pty Ltd (Brisbane, Australia), Nyxoah Pty Ltd (Mont-Saint Guibert, Belgium), and Zelda (now Zelira) Therapeutics Pty Ltd (Australia). The other authors report no conflicts of interest.