Genetic Ancestry for Sleep Research

Abstract

Recent evidence has highlighted the health inequalities in sleep behaviors and sleep disorders that adversely affect outcomes in select populations, including African-American and Hispanic-American subjects. Race-related sleep health inequalities are ascribed to differences in multilevel and interlinked health determinants, such as sociodemographic factors, health behaviors, and biology. African-American and Hispanic-American subjects are admixed populations whose genetic inheritance combines two or more ancestral populations originating from different continents. Racial inequalities in admixed populations can be parsed into relevant groups of mediating factors (environmental vs genetic) with the use of measures of genetic ancestry, including the proportion of an individual’s genetic makeup that comes from each of the major ancestral continental populations. This review describes sleep health inequalities in African-American and Hispanic-American subjects and considers the potential utility of ancestry studies to exploit these differences to gain insight into the genetic underpinnings of these phenotypes. The inclusion of genetic approaches in future studies of admixed populations will allow greater understanding of the potential biological basis of race-related sleep health inequalities.

Sleep science has evolved rapidly in recent decades, from refining methods to measure sleep in humans to gaining understanding of the pathophysiology and significant health impact of sleep disorders. This knowledge has led to the realization that significant health inequalities exist in sleep behaviors and sleep disorders. The World Health Organization defines health inequalities “as differences in health status or in the distribution of health determinants between different populations” that are attributable to biological or sociocultural (voluntary) variations.¹ A multitude of important health determinants exist in all populations with complex interrelationships, including socioeconomic status (SES), literacy, geographic location, age, sex, race, individual behavior, and biology.²

Accumulating evidence highlights the association of self-identified race with sleep health inequalities.^{3, 4, 5} Racial differences in health outcomes are often due to cultural, social, economic, and environmental inequalities that have persisted for generations. For example, neighborhood disadvantage has been found to be a risk factor for pediatric OSA,⁶ and acculturation has important effects on sleep duration.⁷ However, in certain cases, racial inequalities can also be due to heritable biological differences in risk that have a genetic underpinning. If the frequency of a genetic risk factor varies substantially across ancestral populations originating from different continents (eg, African, European, Native American subjects), this variability can be exploited through the study of admixed populations whose genetic inheritance combines two or more ancestral populations. By correlating phenotype with genetic inheritance from the high-risk ancestral population, the location of the predisposing genetic alleles can be more readily identified. Thus, genetic studies of admixed populations (those who have descended from more than one ancestral population) can be a powerful tool for understanding the genetic underpinnings of biological traits, including sleep-related phenotypes. To the extent that biology (rather than social, cultural, environmental, and economic differences) underlies racial inequalities, genetic studies of admixed populations may also provide insights into the causes of racial health inequalities.

The present review discusses the current evidence on racial health inequalities in common sleep disorders focusing on reports in admixed populations, in particular: African-American (AA) subjects, who have a mixture of African and European ancestry, and Hispanic-American (HA) subjects, who have a mixture of African, European, and Native American ancestry, compared with European-American (EA) subjects. We further outline how inclusion of genetic ancestry and techniques such as admixture mapping in association studies can improve scientific inference and identify novel genetic susceptibility loci contributing to variability in sleep phenotypes.

Inequalities in OSA

OSA is a chronic disease with a rising prevalence and considerable attendant morbidity.^{8, 9} Similar to other chronic diseases such as hypertension, OSA has significant heterogeneity among afflicted individuals in terms of pathophysiology and health outcomes.^{10, 11} An underlying factor that contributes to this heterogeneity is racial inequalities in prevalence, risk, and outcomes of OSA (summarized in Table 1).^{4, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41} Epidemiologic studies in the United States report that the prevalence of OSA is greater among AA and HA subjects compared with EA subjects. The relative risk of OSA in AA subjects is moderated by age, with the highest risk being noted in children and a resurgence of risk after the age of 65 years.^{4, 12, 14, 15, 19} The conflicting data regarding higher prevalence in middle-aged AA individuals may be due to the underrepresentation of minorities in larger prospective cohorts and the use of various methods to define OSA, from questionnaires to different quantitative measures.^{13, 18} An important replicated finding is that AA subjects have more severe OSA, based on quantitative metrics such as the apnea-hypopnea index and oximetry, after adjustment for obesity.^{16, 17, 24, 25} However, these studies were performed in clinical populations in which bias related to access to care and delays in diagnosis may have confounded the results.⁴² Although this inequality has not been specifically examined in population-based cohorts, a similar trend is suggested by the Multi-Ethnic Study of Atherosclerosis (MESA), in which the ORs for mild, moderate, and severe OSA in AA subjects compared with EA subjects showed a linear trend toward increasing risk from 1.03 to 1.35.⁴ In fact, a statistically significant increase in risk was observed only for symptomatic OSA regardless of severity, suggesting a higher burden of disease. An older meta-analysis indicates that the independent effect of AA race on OSA risk and severity is small (effect sizes of 0.13 and 0.10, respectively), but AA subjects with OSA have significantly shorter sleep duration (effect size of –0.30).²² Considering the chronic complex nature of OSA, with multiple fixed (age, sex, and craniofacial anatomy) and modifiable (obesity) risk factors, this effect size is likely to be clinically significant.^{43, 44}

Table 1.

Health Inequalities in Sleep Apnea

Study	Year	Study Design	Sample	Measurements	Results
Risk and severity of sleep apnea
Ancoli-Israel et al12	1995	Cross-sectional, n = 400	Aged > 65 y, population-based cohort	HSAT	AA race was associated with twofold increased risk of severe OSA
Kripke et al13	1997	Cross-sectional, n = 355	Aged 40-64 y, population-based cohort	Home oximetry	Prevalence of OSA was three times higher in HA, AA, and other (mostly Asian) races
Redline et al14	1997	Cross-sectional, n = 847	Aged 2-86 y, genetic-epidemiology cohort	HSAT	Prevalence of OSA was higher in younger AA subjects (aged < 25 y; adjusted OR, 1.8)
Redline et al15	1999	Cross-sectional, n= 399	Aged 2-18 years, genetic-epidemiology cohort	HSAT	AA children had higher adjusted prevalence of OSA (OR, 3.5)
Stepanski et al16	1999	Case-control OSA, n = 128 no OSA, n = 62	Aged < 17 y	PSG	AA children with similar BMI had more severe oxygen desaturation
Meetze et al17	2002	Retrospective cross-sectional, n = 280	Aged > 18 y, single-center, clinical cohort	PSG	AA women with OSA were younger and had higher BMI and prevalence of hypertension; AA men had more severe oxygen desaturation
Young et al18	2002	Cross-sectional, n = 5615	Aged > 40 y, Sleep Heart Health Study	Home PSG	Race was not associated with risk of OSA
Rosen et al19	2003	Cross-sectional, n = 850	Aged 8-11 y, population-based cohort	HSAT	AA children had 3.5-fold increased adjusted risk of moderate OSA
Goodwin et al20	2003	Cross-sectional, n = 239	Aged 6-11 y, community-based cohort	Home PSG	HA or EA race was not associated with OSA risk or severity
Scharf et al21	2004	Cross-sectional, n = 233	Retrospective clinical cohort	PSG	AA subjects had more severe OSA; this risk was mediated by income and BMI
Ruiter et al22	2010	Meta-analysis	Prevalence from pooled sample, n = 2,534,882 Severity from pooled sample, n = 6,182	Mixed methods	Higher prevalence (effect size: 0.13) and greater severity (effect size: 0.10) of OSA in AA subjects
Ramos et al23	2011	Prospective cohort study, n = 1,964	Elderly, aged 75 ± 9 y	Questionnaires	HA subjects had higher risk for snoring (OR, 3.6) and daytime sleepiness (OR, 2.8) but no difference for AA subjects
Pranathiageswaran et al24	2013	Prospective, observational study, n = 512	Adults, single-center, clinical cohort	PSG	Young and middle-aged AA men had higher AHI after adjustment for BMI
Weinstock et al25	2014	Cross-sectional, n = 464	Aged 5-10 y, with OSA	PSG	20% increase in AHI in AA children; adjusted for BMI
Ramos et al26	2014	Prospective, observational study, n = 176	Adults, single-center, hospitalized patients with acute stroke	Berlin Questionnaire	HA subjects were at higher risk (OR, 2.6); AA subjects had equivalent risk for OSA
Chen et al4	2015	Cross-sectional, n = 2,230	Aged 54-93 y, population-based cohort	Home PSG	AA subjects had higher odds of OSA; OR, 1.78, short sleep; OR, 4.95, poor sleep quality; OR, 1.57, and daytime sleepiness; OR, 1.89 after adjustment for sex and age. HA subjects had higher odds of OSA and short sleep
Mediators and outcomes of inequalities in sleep apnea
Cakirer et al27	2001	Cross-sectional, n = 364 EA and 165 AA	Adults, Cleveland Family Study	HSAT	Brachycephaly, measured by using anthropometric calipers, was associated with AHI in EA subjects but not in AA subjects
Ancoli-Israel et al28	2002	Prospective clinical cohort, n = 70 each of AA and EA	Aged 65-93 y	Home actigraphy plus respiratory test	AA subjects had higher blood pressure-dipping ratios (“nondipping”) after adjustment for AHI and BMI
Nieto et al29	2000	Cross-sectional, n = 6,132	Aged > 40 y, Sleep Heart Health Study	Home PSG	Risk of hypertension was associated with severity of OSA, but not race. 77% of the cohort was EA, with < 10% in other racial categories
Spilsbury et al30	2006	Cross-sectional n = 843	Aged 8-11 y	HSAT	AA subjects were more likely to have OSA (OR, 3.9); the OR was reduced to 1.9 (0.8-4.6) with adjustment for neighborhood disadvantage
Surani et al31	2009	Cross-sectional, n = 172	Single-center clinical cohort	PSG	HA subjects with OSA had a twofold increased prevalence of diabetes
Baldwin et al32	2010	Cross-sectional, n = 5,237	Age > 40 y, Sleep Heart Health Study	Home PSG	AA subjects with frequent snoring, insomnia symptoms, or EDS had poorer physical health. HA subjects with frequent snoring, insomnia symptoms, or EDS had poorer mental health
Baron et al33	2010	Cross-sectional, n = 5,173	Mean age 66 y	ESS	Risk of EDS was higher in AA subjects (OR, 1.8-2.0) with OSA
Alkhazna et al34	2011	Cross-sectional, n = 280	Adult, single-center clinical cohort	PSG	OSA severity did not vary according to race, but AA subjects had higher prevalence of hypertension
Fulop et al35	2012	Cross-sectional, n = 5,301	AA cohort	Modified Berlin Questionnaire	Risk of OSA in AA men and women was associated with co-morbid hypertension and diabetes after adjustment for BMI and neck and waist circumference
Sands-Lincoln et al36	2013	Cross-sectional, n = 4,418	Adults, 2007-2008 NHANES survey data	Questionnaire	OSA symptoms were associated with risk of hypertension in overweight AA subjects (OR, 4.7), overweight EA subjects (OR, 1.6), and obese HA subjects (OR, 2)
Redline et al37	2014	Cross-sectional, n = 14,440	Aged 45-74 y, Hispanic Community Health Study	HSAT	Overall prevalence in HA subjects was 25.8% for mild OSA; 9.8% for moderate OSA; and 3.9% for severe OSA. Those with moderate to severe OSA were at higher risk for hypertension (OR, 1.4) and diabetes (OR, 1.9)
Bakker et al38	2015	Cross-sectional, n = 2,151	Mean age 68.5 ± 9.2 y, population-based cohort	Home PSG	Moderate to severe OSA was associated with abnormal fasting glucose in AA subjects (OR, 2.14) and EA subjects (OR, 2.85), but not among HA subjects, after adjusting for age, sex, waist circumference, and sleep duration
Turner et al39	2016	Two cohorts, n = 943 AA subjects, n = 452; EA subjects, n = 491	Mean age 80 ± 7.8 y, population-based cohort	Berlin Questionnaire	OSA risk was not different but sleep quality was poorer in elderly AA subjects
Nagayoshi et al40	2016	Cross-sectional, n = 1,844	Mean age 68 y, population-based cohort	Home PSG	Severe OSA was associated with higher prevalence of peripheral arterial disease in AA subjects
Wang et al41	2017	Cross-sectional, n = 774	Mean age 7 ± 1.4 y, clinical cohort	PSG	Association between race and AHI was mediated by socioeconomic variables, including poverty

Comparison group is European-American (EA) subjects unless otherwise specified. AA = African-American; AHI = apnea-hypopnea index; EDS = excessive daytime sleepiness; ESS = Epworth Sleepiness Scale; HA = Hispanic-American; HSAT = home sleep apnea test; NHANES = National Health and Nutrition Examination Survey; PSG = polysomnography.

There are fewer comparative studies on OSA prevalence between HA and EA populations. A previous report from a community cohort indicated a threefold higher prevalence of moderate to severe OSA in HA subjects.¹³ In the MESA cohort, HA subjects had a higher risk of mild, moderate, and severe OSA (OR, 1.6, 1.9, and 2.1, respectively). In contrast with observations in AA subjects, the prevalence of symptomatic OSA was not significantly higher in HA subjects.⁴ A report from the large Hispanic Community Health Study found that the prevalence of OSA (approximately 26%, 10%, and 4% for mild, moderate, and severe disease) was at least as high as studies from predominantly EA cohorts.^{8, 37} This study used home sleep apnea testing to assess OSA and included participants of seven Hispanic/Latino backgrounds; significant differences in the prevalence of moderate to severe OSA were found in men across diverse Hispanic backgrounds even after adjustment for age and obesity. Elderly and post-stroke HA subjects are at higher risk for OSA compared with their EA counterparts.^{23, 26} In contrast, data from a community pediatric cohort found no difference in the prevalence of OSA between HA and EA children.²⁰

Consistent with the inequalities in OSA discussed earlier, poorer health outcomes have been reported in AA and HA subjects compared with EA subjects with OSA. AA individuals with OSA symptoms report higher rates of excessive daytime sleepiness as well as poorer sleep quality and physical health.^{32, 33, 39} Elderly AA individuals have higher rates of abnormal 24-h blood pressure profiles (nocturnal nondipping blood pressure), independent of obesity and severity of OSA.²⁸ Observational data from a clinical cohort and a general population survey suggest that the risk of hypertension is higher in AA subjects.^{34, 36} In contrast, the risk of hypertension in OSA was not found to vary significantly according to race in the Sleep Heart Health Study, a younger community-based, predominantly EA cohort.²⁹ Severe OSA is associated with risk of peripheral arterial disease only in AA subjects.⁴⁰ Within AA subjects, OSA is independently associated with a higher prevalence of hypertension and diabetes and in women with cellular senescence (telomere shortening).^{35, 45} Moderate to severe OSA has been associated with abnormal fasting glucose levels in AA and EA populations but not in HA subjects.³⁸ However, HA subjects with moderate to severe OSA report poorer mental health and are at increased risk for diabetes.^{31, 32, 37}

The mechanisms underlying these racial inequalities in OSA are not well understood. Obesity is the strongest risk factor for OSA and is rooted in both environmental and genetic factors, varying significantly according to race.^{46, 47} Although the association of genetic risk loci for obesity with OSA traits remain to be described, it is plausible that racial inequalities in OSA are mediated in part by genetic and environmental factors that drive higher rates of obesity in AA and HA subjects. The majority of studies in OSA adjust for BMI, but although the heritability estimates for BMI in AA populations are similar to EA populations, the role of fat distribution (parapharyngeal and abdominal fat) remains unclear.^{48, 49} There are two reports identifying novel OSA genetic risk loci with genome-wide significance in HA subjects,⁵⁰ and another quantitative trait locus for non-rapid eye movement sleep apnea-hypopnea index in men of diverse racial backgrounds.⁵¹ In contrast, genome-wide association studies (GWAS) in EA and AA populations have not yet identified any OSA genetic loci meeting genome-wide significance criteria, although the sample sizes used have been smaller, and this topic remains an area of active investigation. Linkage analyses and candidate gene studies have reported differences in OSA risk loci between AA and EA subjects, but the sample sizes for replication have been underpowered.^{48, 49, 52, 53}

There are few studies examining the role of craniofacial characteristics as an explanatory factor in racial OSA inequalities. Although brachycephaly is an important mediator of OSA risk in EA subjects, enlarged tongue and soft tissue are reported to mediate the propensity for OSA in AA subjects.^{27, 54, 55} Although AA subjects have smaller upper airway dimensions, the heritability of this trait in AA individuals is similar to EA individuals.⁴³ There are few data regarding racial differences in OSA phenotypes with respect to respiratory control, upper airway collapsibility, and arousal threshold or racial differences in the physiological responses to perturbations in OSA (chronic intermittent hypoxia, sleep fragmentation, and sympathetic activation). AA groups have a lower vital capacity and more hypoxemic attenuation of baroresponse in sleep than EA groups,^{14, 56} but these results have not been replicated, and thus the implications of these findings remain unclear.

In summary, the prevalence and severity of OSA is higher in AA subjects and likely in HA subjects, independent of obesity. Thus far, limited data indicate that the risk of adverse health outcomes such as daytime sleepiness and cardiometabolic outcomes may also be higher in these populations. It is important to note that SES, neighborhood disadvantage, and poverty have been shown to mediate the elevated risk of OSA in AA subjects, underscoring the importance of considering and controlling for psychosocial factors in future studies of racial inequalities.^{21, 30, 41}

Inequalities in Insomnia

Overall, minorities are at a higher risk for insomnia compared with EA subjects.^{57, 58} When self-reported insomnia is examined specifically in AA populations (and in HA populations in some studies), the prevalence is either lower than in EA subjects or largely explained by SES and psychosocial stressors.^{22, 59, 60, 61, 62, 63, 64, 65} The importance of SES as a mediator of sleep quality can also be gleaned from two studies conducted in college students, in which education and health between racial groups should be comparable. These studies used validated questionnaires and found equivalent or a lower prevalence of insomnia in AA groups.^{66, 67} On the other hand, physician-diagnosed insomnia rates, sleep quality according to validated questionnaires, and quantitative measures indicate poorer sleep quality, particularly in urban AA populations.^{60, 68, 69, 70, 71} Thus, racial inequalities related to insomnia reveal paradoxical findings, which highlight the role that complex mediators and cultural beliefs may play on self-reported symptoms (Table 2).^{22, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71} In addition to the consideration of the multilevel factors noted earlier and their interactions, future use of high throughput wearable technologies and genetic ancestry to add quantitative measures of sleep quality, bio-geographical ancestry to population surveillance, and clinical association studies will advance our understanding of racial inequalities in insomnia.

Table 2.

Health Inequalities in Insomnia

Study	Year	Study Design	Sample	Measurements	Results
Bixler et al57	2002	Cross-sectional, n = 1,741	Aged > 20 y, population-based cohort	Self-reported symptoms	Non-EA subjects (OR, 2.0) were significantly associated with risk of insomnia
Allen et al68	2008	Survey, n = 1,910	Aged > 65 y, rural population-based cohort	Adapted validated questionnaire	AA subjects reported similar frequency of insomnia
Mezick et al69	2008	Cross-sectional, n = 187	Aged 45-75 y, population-based cohort	PSQI, home PSG, and actigraphy	AA subjects had less sleep efficiency, SWS, and poorer sleep quality. This effect was mediated by SES
Ruiter et al22	2010	Meta-analysis	Pooled sample of > 8,000	Mixed methods	AA subjects were less likely to report insomnia symptoms (ES: –0.19 for WASO and –0.23 for sleep complaints)
Patel et al65	2010	Cross-sectional, n = 9,714	Aged > 18 y, Population-based sample	Self-reported symptoms	AA and HA subjects reported poor sleep quality (OR, 1.59 and 1.65, respectively), and this outcome was mediated by SES
Grandner et al59	2010	Survey, n = 159,856	Aged > 18 y, population-based sample	Self-reported symptoms	AA and HA women were less likely to report insomnia symptoms (OR, 0.74 for each)
Ram et al60	2010	Cross-sectional, n = 6,139	2005-2006 NHANES	Physician-diagnosed and self-reported symptoms	AA and HA subjects (≥ 1.5% vs 0.8%) had higher rates of physician-diagnosed insomnia but less self-reported WASO
Gaultney66	2010	Survey, n = 1,845	College students	Validated questionnaire	AA students reported less risk for insomnia and poor sleep practices
Pigeon et al70	2011	Cross-sectional clinical cohort, n = 92	Mean age 52 y	PSQI	AA subjects were more likely to experience sleep disturbance (OR, 2.4) after adjustment for education, depression, and chronic illness
Grandner et al61	2012	Cross-sectional, n = 7,148	Aged > 18 y, population-based sample	Self-reported symptoms	Perceived racial discrimination in health-care setting was associated with increased risk of sleep disturbance (OR, 1.60; P = .04) in AA subjects
Singareddy et al58	2012	Prospective with follow-up of 7.5 y, n = 1,246	Aged ≥ 20 y, population-based cohort	Self-reported symptoms	Non-EA subjects were more likely (OR, 2.8) to develop incident insomnia independent of SES and physical and mental health
Grandner et al62	2013	Cross-sectional, n = 4,081	2007-2008 NHANES	Self-reported symptoms	AA subjects were more likely to report prolonged sleep latency (adjusted OR, 1.6). AA and HA subjects were less likely to report insomnia symptoms (adjusted OR, 0.56-0.82)
Hicken et al63	2013	Cross-sectional survey, n = 3,105	Aged > 18 y, urban population-based sample	Self-reported symptoms	AA subjects reported more insomnia symptoms. This finding was mediated by SES and racism-related vigilance. Similar trends were observed in HA subjects
Petrov et al67	2014	Survey, n = 1,684	College students	Validated questionnaire	Prevalence of insomnia was not different in AA subjects
Slopen and Williams64	2014	Cross-sectional, n = 2,983	Age > 18 y, urban population-based sample	Self-reported symptoms	Discrimination mediates sleep duration and sleep difficulty in AA and HA subjects, independent of SES and other psychosocial stressors
Carnethon et al71	2016	Observational, n = 496	Aged 35-64 y, urban population	Actigraphy, PSQI	AA subjects had significantly lower sleep efficiency, greater sleep fragmentation, and poorer self-reported sleep quality adjusted for SES and health indicators

Comparison group is EA subjects unless otherwise specified. ES = effect size; PSQI = Pittsburgh Sleep Quality Index; SES = socioeconomic status; SWS = slow wave sleep; TST = total sleep time; WASO = wake after sleep onset. See Table 1 for expansion of other abbreviations.

Inequalities in Narcolepsy

Narcolepsy is a rare, central disorder of hypersomnolence characterized by excessive daytime sleepiness, cataplexy, hypnagogic hallucinations, sleep paralysis, and abnormal rapid eye movement sleep, and it is caused by a lack of hypocretin/orexin.⁷² Genetic studies of narcolepsy reveal certain commonalities and differences across racial populations, including AA and EA subjects (Table 3).^{72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89} In most, but not all studies,⁷³ AA subjects are more frequently DRB1*15 (DR2) negative than EA subjects.^{74, 75, 76} A number of studies have shown the human leukocyte antigen (HLA) DQB1*0602 rather than DRB1*15 (DR2) is the unifying, most strongly associated susceptibility allele across racial groups, including AA and EA groups.^{75, 76, 77, 78, 79} In AA groups, the association is stronger with DQB1*06:02 because linkage disequilibrium (the nonrandom pattern of association between alleles at different loci within a population) is not as high between DQB1*0602 and DRB1*1501, indicating greater independence of these alleles.^{76, 78} Significant effects of other HLA-DQ alleles also have been reported across racial groups. In AA subjects, narcolepsy is associated with DQB1*0602 haplotypes bearing distinct DRB1 alleles, most commonly DRB1*1503, DRB1*1501, DRB1* 1101, and DRB1*0806.⁸⁰ In addition, the DRB1*13, DQA1*0103, and DQB1*0603 haplotypes confer moderate protection in EA and AA subjects.⁷⁸

Table 3.

Health Inequalities in Narcolepsy

Study	Year	Study Design	Sample	Measurements	Results
Kramer et al73	1987	Cross-sectional, n = 14	Aged 36-74 y, patients with narcolepsy	Genotyping	100% of EA and AA subjects were positive for DRB1*15 (DR2), indicating association between DRB1*15 (DR2) and narcolepsy. No differences between races
Neely et al74	1987	Cross-sectional, n = 18 AA subjects with narcolepsy; n = 99 AA control subjects	Adult patients with narcolepsy from a sleep clinic in Chicago	Genotyping	33% of AA narcoleptic subjects did not have DRB1*15 (DR2), which was lower than historic EA subjects; 100% of AA and EA subjects had DQw1; 22% of AA subjects had DQw1 but without DR2; 100% of historic EA subjects had both DR2 and DQw1
Mignot et al75	1994	Retrospective, n = 47	Adult patients with narcolepsy and cataplexy from US and international university-based sleep clinics and laboratories	Genotyping	95% of EA subjects but only 11% of AA subjects were DRB1*1501 (DR2) positive; 68% of AA subjects and 0% of EA subjects carried DRB1*1503. DQB1*0602 was found in 96% of AA subjects and 95% of EA subjects. DQB1*0602 was a more sensitive marker for narcolepsy than DRB1*15 (DR2) in AA and EA subjects
Rogers et al76	1997	Retrospective, n = 188	Adult patients with narcolepsy and cataplexy, from Stanford database	Genotyping	67.2% of AA subjects were positive for DRB1*15, compared with 84.5% of EA subjects. In AA subjects, association was stronger with DQB1*06:02. DQB 1*0602 was a more sensitive marker for narcolepsy with cataplexy than DRB1*15 (DR2) in AA and EA subjects
Mignot et al77	1997	Cross-sectional, n = 509	Aged 18-68 y, patients with narcolepsy enrolled in clinical trial for modafinil	Genotyping	DQB1*0602 positivity was significantly higher in AA subjects. DQB1*0602 was a more sensitive marker for narcolepsy than DRB1*15 (DR2) in AA and EA subjects
Mignot et al78	2001	Retrospective population-based case-control study, n = 420 narcolepsy with cataplexy, n = 1,087 control subjects	Adult patients with narcolepsy and cataplexy, from US and international university-based sleep clinics and laboratories	Genotyping	DQB1*0602 positivity was significantly higher in AA subjects. In AA subjects, association was stronger with DQB1*0602. DRB1*13, DQA1*0103, and DQB1*0603 haplotypes conferred moderate protection in EA and AA subjects
Pelin et al79	1998	Cross-sectional, n = 669	Adult patients with narcolepsy from Stanford database and multicenter clinical trial of modafinil	Genotyping	Both EA and AA subjects with or without cataplexy who were homozygous for HLA-DQB1*0602 had relative risks 2-4 times higher compared with HLA-DQB1*0602 heterozygotes. No differences in severity with increasing allelic dosage in EA or AA subjects
Mignot et al80	1997	Retrospective, n = 58	Adults, non-DRB1*15 patients with narcolepsy and cataplexy, from Stanford database	Genotyping	In AA subjects, narcolepsy was associated with rare DQB1*0602 haplotypes bearing distinct DRB1 alleles, most commonly DRB1*1503, DRB1*1501, DRB1* 1101, and DRB1*0806
Hallmayer et al81	2009	Case-control GWAS, n = 807 cases, n = 1,074 control subjects; replication: n = 1,057 cases, n = 1,104 control subjects	Adults, clinical cohorts from various national and international sleep clinics and laboratories	Genotyping	Association of narcolepsy within the TCRA locus polymorphisms in three ethnic groups, including EA and AA subjects
Kornum et al82	2011	Case-control GWAS, n = 807 cases and n = 1,074 control subjects; n = 1,858 cases and n = 2,384 control subjects	Adults, clinical cohorts from various national and international sleep clinics and laboratories	Genotyping	Association of narcolepsy with an SNP in the P2RY11 gene in three ethnic groups, including EA and AA subjects
Holm et al83	2015	Case-control GWAS, EA: n = 807 cases, n = 1,074 control subjects; Chinese: n = 1,078 cases, n = 1,903 control subjects; AA: n = 249 cases, n = 1,048 control subjects	Adults, clinical cohorts from various national and international sleep clinics and laboratories	Genotyping	Association of narcolepsy with an SNP in the EIF3G gene in three ethnic groups, including EA and AA subjects
Longstreth et al84	2009	Retrospective population-based study, n = 425 narcolepsy	Aged > 18 y, positive for DQB1*0602 allele, King County, Washington	Questionnaires, interviews, genotyping	Higher prevalence of narcolepsy with cataplexy in AA subjects
Koepsell et al85	2010	Retrospective population-based case-control study, n = 45 narcolepsy with cataplexy, n = 95 control subjects	Aged 18-50 y, positive for the DQB1*0602 allele, King County, Washington	Questionnaires, interviews, genotyping	Higher prevalence of diagnosed narcolepsy cases in AA than EA subjects (OR, 8.1). Higher prevalence of diagnosed narcolepsy in households with lower educational attainment and lower annual income
Okun et al86	2002	Retrospective cross-sectional, n = 484	Mean age 43.1 y, diagnosis of narcolepsy-cataplexy	ESS, MSLT, PSG, genotyping, hypocretin-1 level	AA and EA subjects exhibited similar symptoms, age of onset, and disease severity. Minor differences in other variables: EA subjects had less severe sleep paralysis and more reports of cataplexy affecting jaw and arms than AA subjects. AA subjects had more reports of negative emotions triggering cataplexy and going blank due to sleep attacks
Cairns and Bogan87	2015	Retrospective cross-sectional, n = 3,059 suspected hypersomnia, n = 79,651 general sleep clinic sample	Aged > 18 y, repository of scored and physician-interpreted records	MSLT, PSG	AA subjects were 2.8-4 times as likely to have a PSG SOREMP, controlling for other significant variables
Kawai et al72	2015	Retrospective cross-sectional, n = 1,097	Children and adults, diagnosed with narcolepsy, Stanford Center for Narcolepsy Research database	ESS, MSLT, PSG, genotyping, hypocretin-1 level	Sex ratio, PSG, and measures did not differ between AA and EA subjects. ESS score was higher and age of onset of sleepiness was earlier in AA subjects. HLA-DQB1*0602 positivity was higher in AA subjects but CSF hypocretin-1 level was more frequently low (≤ 110 pg/ml) in AA subjects. In patients with low CSF hypocretin-1 levels, AA subjects were 4.5-fold more likely to be without cataplexy
Andlauer et al88	2012	Retrospective population-based case-control study, n = 171 narcolepsy with cataplexy, n = 170 control narcolepsy without cataplexy	Aged > 18 y, university-based sleep clinics and laboratories	ESS, MSLT, PSG, genotyping, hypocretin-1 level	Patients with low CSF hypocretin-1 level (≤ 110 pg/ml) were more likely to be AA vs EA: 1% of normal-CSF-hypocretin-1 vs 20% of low-CSF-hypocretin-1 cases were AA subjects (OR, 28)
Chambers and Belicki89	1988	Cross-sectional, n = 97	Mean age 21.6 y; Canadian university students	Sleep Disorders Questionnaire	Abuse/trauma group scored more negatively than control subjects for narcolepsy (mean: 25.8 vs 21.1; P < .01)

Comparison group is EA subjects unless otherwise specified. CSF = cerebrospinal fluid; GWAS = genome-wide association study; MSLT = multiple sleep latency test; SNP = single nucleotide polymorphism; SOREMP = sleep-onset rapid eye movement period. See Table 1 for expansion of other abbreviations.

Beyond the HLA genes, other risk genes have also been implicated in the development of narcolepsy in various racial/ethnic groups (Table 3). Associations between narcolepsy and polymorphisms in the TCRA locus,⁸¹ in the P2RY11 gene,⁸² and in the EIF3G gene⁸³ have been found in AA and EA populations.

A few studies conducted to date have investigated the prevalence and clinical phenotypic expression of narcolepsy according to race. The prevalence of DQB1*0602 in narcolepsy was much higher in AA subjects than in EA subjects,^{72, 80} although in normal healthy adult sleepers, the prevalence of this risk allele did not differ significantly between these racial groups.⁹⁰ Notably, heterozygous or homozygous DQB1*0602 allele status influenced risk and disease expression in AA and EA subjects.^{79, 91} Thus, frequency differences of the DQB1*0602 risk haplotype in racial groups indicate that environmental factors contribute to the development of narcolepsy and thus may explain differences in narcolepsy prevalence. Indeed, two studies found a higher prevalence of diagnosed narcolepsy with cataplexy in AA subjects compared with EA subjects.^{84, 85} In individuals with narcolepsy with cataplexy, AA and EA groups exhibited similar symptoms, age of onset, and symptom severity but had minor differences in severe sleep paralysis, reports of cataplexy affecting the jaw and arms and being triggered by negative emotions, and reports of going blank due to sleep attacks.⁸⁶ In general clinic and hypersomnia groups, AA subjects, compared with EA subjects, were 2.8 to 4.0 times as likely to have a sleep-onset rapid eye movement period on polysomnography, after controlling for other significant variables.⁸⁷ Another study found that sex ratio, polysomnographic findings, and multiple sleep latency test measures did not differ between AA and EA subjects; AA subjects with narcolepsy had higher DQB1*0602 positivity, earlier symptom onset, and more severe daytime sleepiness, however.⁷² AA subjects also had lower cerebrospinal fluid hypocretin-1 levels and lower rates of cataplexy.^{72, 88}

Beyond racial differences, there is a paucity of research for other health inequalities in narcolepsy. One study found a higher prevalence of diagnosed narcolepsy with cataplexy in households with lower educational attainment and lower annual income.⁸⁵ Another study found that students with a history of adverse childhood experiences had higher mean scores on subjective sleep disorder measures of narcolepsy than those without such experiences.⁸⁹ Longitudinal studies among diverse populations are needed to understand this association and to investigate potential racial inequalities in the strength of this association.⁹²

In summary, there are well-established racial differences in narcolepsy, with genetic underpinnings. Because both environmental and genetic factors underlie racial inequalities, genetic ancestry is an ideal method for future studies in AA groups and other admixed groups with narcolepsy; the goal is to parse the genetic aspects of race from the cultural, behavioral, and social aspects that may underlie observed phenotypic differences. Such parsing could facilitate symptom identification and diagnosis of narcolepsy.

Inequalities in Chronotype, Circadian Parameters, and Circadian Rhythm Sleep Disorders

Chronotype (also referred to as morningness-eveningness or diurnal preference) shows considerable interindividual variation and is the tendency to be an early “lark” (alert and preferring to be active early in the day) or a late “owl” (alert and preferring to be active later in the day). Several studies have found racial differences in chronotype and associated circadian parameters (Table 4).^{93, 94, 95, 96, 97, 98, 99, 100} Furthermore, chronotype differences have been found between HA subgroups⁹³ and between European and African subjects.^{94, 95} Using experimental studies, Eastman et al⁹⁷ found robust differences between AA and EA subjects in basic properties of the circadian clock, including endogenous or free-running circadian period and the magnitude of phase advances and delays, which can contribute to morningness-eveningness, with self-identification techniques.⁹⁶ Using genetic ancestry, differences in circadian period, chronotype, and responses to shifts in the sleep-wake cycle between these two racial groups have been reported.^{98, 99, 100} Of interest, circadian period correlated with percentage of African and European genetic ancestry, whereby longer circadian periods were associated with a greater percentage of European ancestry and a smaller percentage of African ancestry. There was also sex according to ancestry differences: EA women had a shorter circadian period than men, but there was no sex difference in circadian period between AA men and women. The aforementioned racial differences suggest there could be racial inequalities in disease risk, in response to jet lag or shiftwork and/or in the development of circadian rhythm sleep disorders (CRSDs).^{94, 95, 97, 100} In contrast to chronotype or circadian parameters, however, there are no published studies of racial inequalities in CRSDs, which are extreme clinical variants of chronotype and include advanced and delayed sleep phase disorders.

Table 4.

Health Inequalities in Chronotype and Circadian Parameters

Study	Year	Study Design	Sample	Measurements	Results
Knutson et al93	2017	Cross-sectional, n = 113,429	Aged 18-74 y, self-identified Hispanic/Latino background	Self-reported bedtimes and wake times. Chronotype defined as midpoint of sleep on weekends adjusted for sleep duration on weekdays	Significant differences between various HA groups, with Mexican-American subjects having earliest chronotype, bedtimes, and wake times
Malone et al94	2016	Cross-sectional, n = 439,933	Aged 40-69 y, UK Biobank study	Chronotype via single item in which participants rated themselves as definitely a morning person, more a morning than an evening person, more an evening than a morning person, definitely an evening person	Morning vs intermediate chronotype was 1.4 times more prevalent in self-identified African subjects than white subjects in the United Kingdom
Malone et al95	2017	Cross-sectional, n = 2,044	Aged 40-69 y, UK Biobank study	Chronotype via single item in which participants rated themselves as definitely a morning person, more a morning than an evening person, more an evening than a morning person, definitely an evening person	African subjects had a 62% greater odds of being morning chronotype than white subjects in the United Kingdom
Smith et al96	2009	Experimental, n = 60	29 male subjects, aged 18-45 y	Circadian period, circadian phase advances and delays via DLMO measurements	AA subjects had significantly shorter circadian periods. AA subjects had larger phase advances and smaller phase delays
Eastman et al97	2012	Experimental, n = 94	45 male subjects, aged 18-42 y	Circadian period via DLMO measurements. Chronotype via MEQ	AA subjects had significantly shorter circadian periods. MEQ scores did not differ between races
Eastman et al98	2015	Experimental, n = 36	19 male subjects, aged 21-43 y	Circadian period, circadian phase shifts via DLMO measurements. Chronotype via MEQ	AA subjects defined according to genetic ancestry had significantly shorter circadian periods. Longer circadian periods were associated with a greater percentage of European ancestry and a smaller percentage of African ancestry. EA subjects had larger phase shifts after 9-h advance than AA subjects but were more likely to phase delay. MEQ scores did not differ between races
Eastman et al99	2016	Experimental, n = 45	22 male subjects, aged 18-44 y	Circadian period, circadian phase shifts via DLMO measurements. Chronotype via MEQ	AA subjects defined according to genetic ancestry had significantly shorter circadian periods and an earlier chronotype (more morningness). EA subjects had larger phase delays than AA subjects after 9-h delay
Eastman et al100	2017	Experimental, n = 63	31 male subjects, aged 18-44 y	Circadian period, circadian phase shifts via DLMO measurements	AA subjects defined according to genetic ancestry had significantly shorter circadian periods. EA women had shorter circadian periods than men, but there was no sex difference in AA subjects

Comparison group is EA subjects unless otherwise specified. DLMO = dim light melatonin onset; MEQ = morningness-eveningness questionnaire. See Table 1 for expansion of other abbreviations.

The genetic underpinnings of chronotype and CRSDs have been well studied, although only a few studies have examined racial differences in core circadian clock genes. Three GWAS have identified genetic components of chronotype, although all used European ancestry populations,^{101, 102} and nearly all candidate gene studies of chronotype or advanced and delayed sleep phase disorders have used individuals of European or Asian ancestry, with little investigation of admixed groups such as AA or HA subjects.^{102, 103} Of note, two studies have found racial differences in the frequencies of polymorphisms of core clock genes associated with chronotype and with CRSDs.^{104, 105}

In summary, despite extensive knowledge of the genetics of chronotype and CRSDs, studies examining racial differences are limited but promising. Given such findings, a better understanding of the prevalence of circadian rhythm perturbations according to race and the extent to which genetic differences underlie racial differences in phenotypes such as chronotype and circadian period are critical areas of health inequalities research.

Genetic Ancestry and Admixture Mapping: Application in Understanding Sleep Health Inequality

Although outside the scope of the present review, the importance of social determinants of health in mediation of racial health inequalities must be emphasized. With respect to inequalities in health behaviors, polygenic traits, and chronic diseases, social determinants of health such as education, lifestyle, living and work situations, income, environmental pollution, public policy, discrimination, and psychosocial stress can have a profound impact and should be considered in research and in practice.¹⁰⁶ As an overview, a schematic conceptual model is presented (Fig 1) outlining the multilevel genesis of racial sleep health inequalities.

Figure 1.

Conceptual model for multilevel genesis of sleep health inequalities. The solid arrows indicate relationships supported by human studies. Dashed arrows indicate hypothetical or animal data-supported relationships. CIH = chronic intermittent hypoxia; CRSD = circadian rhythm sleep disorder.

DNA sequence is identical across > 99.4% of sites across the human genome, and genetic diversity ranges widely across human populations, being greatest among populations with recent African ancestry.¹⁰⁷ The genomes of continental populations since the migration of humans out of Africa have diverged primarily because of genetic drift or natural random fluctuations of allele frequencies, with some contribution from natural selection acting differentially across continental populations and from the emergence of new population-specific mutations.¹⁰⁸ Admixed populations such as AA and HA groups, with recent ancestry from two or three continental populations, bring together combinations of these diverse genomes, and therefore carry substantial genetic variation that can contribute to phenotypic traits and health inequalities.^{109, 110} Genetic ancestry in admixed AA and HA populations can be measured reliably using as few as 400 ancestry-informative markers (AIMs),¹¹¹ which are single nucleotide polymorphisms that vary widely in allele frequency across ancestral populations from different continents. Analysis of these markers in combination provides estimates of the proportion of ancestry from each contributing ancestral population, or average “global ancestry,” across the genome.^{112, 113} For example, AA populations typically have approximately 80% recent African ancestry and approximately 20% recent European ancestry, whereas African, European, and Amerindian ancestry proportions vary widely in HA subjects based on different admixture events and the number of generations since the admixture event.^{107, 114} As genome-wide genotypes become more readily available, global genetic ancestry is increasingly calculated by using genome-wide data, which provide improved accuracy to differentiate between closely related populations compared with specifically selected AIMs.¹¹¹

Genetic ancestry is a powerful tool to search for biological contributions to health inequalities in admixed populations.¹¹⁰ Association of global ancestry estimates with sleep disorder prevalence and severity, or with differential response to sleep medications, signals that ancestry-specific risk or protective genetic factors are present. Importantly, ancestry-directed studies searching for genetic factors need to adequately control for other risk and social factors such as SES.¹¹⁵ Furthermore, statistical power will depend on heritability of the trait, strength of ancestry-specific effects, and sample size. For example, African genetic ancestry estimated by using approximately 1,700 AIMs in 70 AA subjects was found to be associated with highly heritable indices of sleep depth (slow-wave sleep and delta power), although this study may have been underpowered to detect genetic ancestry effects in less heritable sleep duration and sleep efficiency traits.¹¹⁶

If prevalence and severity of a sleep trait vary by ancestry, admixture mapping can identify genomic regions that track with disease in one ancestral population.¹¹⁷ Admixture mapping relies on the concept of “local ancestry,” in which the ancestral origin of each chromosomal segment in the mosaic genome of an individual can be quantified on the basis of polymorphisms with highly differentiated allele frequencies in the ancestral populations.¹¹⁸ In a case only admixture mapping study design, affected individuals from an admixed population are scanned for regions of the genome that deviate in local ancestry from genome-wide averages. In a case-control study design or for quantitative traits, regression models are used to test the association of local ancestry with disease status or trait levels.¹¹⁷ Admixture mapping has been used successfully to identify genetic associations for several traits; for example, multiple sclerosis,¹¹⁹ chronic kidney disease,¹²⁰ and pharmacogenetics of relapse of acute lymphocytic leukemia.¹²¹ Advantages of admixture mapping include the following: (1) the requirement of low-density genomic coverage to identify genetic signals, leading to a lower multiple testing burden relative to GWAS; (2) the ability to aggregate evidence from multiple independently associated variants in a region of local ancestry, even if these variants are not directly genotyped; and (3) the opportunity to detect regions with functional variants that have undergone selection in one of the ancestral populations. However, admixture mapping requires follow-up genotyping and association testing to identify specific contributing genetic variants in the identified regions. Furthermore, analysis for some HA groups remains challenging because computational methods that detect three-way admixture are rare, and ancestral populations for HA subjects are often not well known or represented in public sequence datasets.^{122, 123}

Newer methods combine independent association evidence from admixture mapping with single variant association tests in admixture-informed GWAS to enhance the power to detect novel genetic signals in admixed populations.^{124, 125} They can also identify new gene-gene or gene-environment interactions that arise based on the new combination of genomes or environmental contexts in admixed populations.¹²⁶ These could be useful tools in the effort to identify biological contributors to sleep inequalities in AA and HA populations.

Beyond gene discovery, follow-up studies of newly discovered sleep disorder loci in AA and HA populations are expected to be useful to: (1) test if genetic effects for sleep disorders generalize or are consistent across different US populations¹²⁷; (2) fine-map or narrow the genomic interval in which causal variants lie¹²⁸; (3) estimate polygenic risk for sleep disorders or pharmacogenetic response to medications that may have implications for personalized screening, prevention, or therapy in AA and HA populations¹²⁹; and (4) investigate the impact of sleep genetic factors on related comorbidities that contribute to health inequality.

Conclusions

Substantial inequalities exist in a wide range of sleep phenotypes and sleep disorders that may contribute to overall health inequalities given the impact of poor sleep on a wide range of psychiatric, neurocognitive, metabolic, and cardiovascular health outcomes. Exploiting these inequalities in admixed populations such as AA and HA subjects may prove to be a powerful tool to identify underlying genetic variants predisposing to sleep disorders, thereby providing important insights into the molecular pathogenesis of these diseases. For admixture studies to be statistically robust in identifying causal genetic variants, the studies must control for cultural, socioeconomic, or other environmental differences and be adequately powered to detect differences in risk between ancestral populations due to differences in the prevalence of the causal genetic variants.

Our current understanding of OSA and insomnia both in terms of the magnitude of health inequalities and the extent to which these inequalities are due to genetic variants suggest that very large sample sizes will be required for ancestry studies to be helpful. In contrast, ancestry studies in narcolepsy and CRSDs/chronotype, because they putatively have a much larger genetic underpinning, may be more fruitful. Future studies using admixture-informed approaches hold great promise in identifying and ultimately addressing biological contributions to sleep health inequalities in these settings. Furthermore, because sleep health inequalities result in a disproportional impact on health in minority populations, it is vital that future genetic studies include these racial groups so that subsequent knowledge gained from such research can be applied directly to those populations who would derive the greatest benefit.