Q indicates quarter; REM, rapid eye movement.
eFigure 1. Flow chart illustrating participant selection
eFigure 2. Conditional Inference Survival Tree Model evaluating all sleep stages as predictors of survival in the MrOS study
eFigure 3. Variable Importance based on random survival forest classifier evaluating all sleep stages as predictors of survival in the MrOS study
eTable 1. Mortality Data for the Osteoporotic Fractures in Men Study (MrOS) and Wisconsin Sleep Cohort by REM Quartile
eTable 2. Mortality HRs from Cox Regression with Percentage REM Sleep as Binary Variable (< 15%) in the Osteoporotic Fractures in Men Study and Wisconsin Sleep Cohort
eTable 3. Sensitivity Analyses of Mortality Hazard Ratios from Cox Regression in the Osteoporotic Fractures in Men Study
Customize your JAMA Network experience by selecting one or more topics from the list below.
Leary EB, Watson KT, Ancoli-Israel S, et al. Association of Rapid Eye Movement Sleep With Mortality in Middle-aged and Older Adults. JAMA Neurol. 2020;77(10):1241–1251. doi:10.1001/jamaneurol.2020.2108
Is less rapid eye movement (REM) sleep associated with increased mortality?
In this cross-sectional study of 4050 individuals from 2 independent cohorts, lower amounts of REM sleep were associated with increased risk of all-cause mortality. There was a 13% higher mortality rate over 12.1 years for every 5% reduction in REM sleep in a cohort of 2675 older men, and the finding was replicated in a cohort of 1375 middle-aged men and women followed-up for 20.8 years.
Less REM sleep is associated with increased mortality risk.
Rapid eye movement (REM) sleep has been linked with health outcomes, but little is known about the relationship between REM sleep and mortality.
To investigate whether REM sleep is associated with greater risk of mortality in 2 independent cohorts and to explore whether another sleep stage could be driving the findings.
Design, Setting, and Participants
This multicenter population-based cross-sectional study used data from the Outcomes of Sleep Disorders in Older Men (MrOS) Sleep Study and Wisconsin Sleep Cohort (WSC). MrOS participants were recruited from December 2003 to March 2005, and WSC began in 1988. MrOS and WSC participants who had REM sleep and mortality data were included. Analysis began May 2018 and ended December 2019.
Main Outcomes and Measures
All-cause and cause-specific mortality confirmed with death certificates.
The MrOS cohort included 2675 individuals (2675 men [100%]; mean [SD] age, 76.3 [5.5] years) and was followed up for a median (interquartile range) of 12.1 (7.8-13.2) years. The WSC cohort included 1386 individuals (753 men [54.3%]; mean [SD] age, 51.5 [8.5] years) and was followed up for a median (interquartile range) of 20.8 (17.9-22.4) years. MrOS participants had a 13% higher mortality rate for every 5% reduction in REM sleep (percentage REM sleep SD = 6.6%) after adjusting for multiple demographic, sleep, and health covariates (age-adjusted hazard ratio, 1.12; fully adjusted hazard ratio, 1.13; 95% CI, 1.08-1.19). Results were similar for cardiovascular and other causes of death. Possible threshold effects were seen on the Kaplan-Meier curves, particularly for cancer; individuals with less than 15% REM sleep had a higher mortality rate compared with individuals with 15% or more for each mortality outcome with odds ratios ranging from 1.20 to 1.35. Findings were replicated in the WSC cohort despite younger age, inclusion of women, and longer follow-up (hazard ratio, 1.17; 95% CI, 1.03-1.34). A random forest model identified REM sleep as the most important sleep stage associated with survival.
Conclusions and Relevance
Decreased percentage REM sleep was associated with greater risk of all-cause, cardiovascular, and other noncancer-related mortality in 2 independent cohorts.
Sleep issues affect approximately 50 to 70 million individuals in the US,1 contributing to multisystemic medical consequences including cardiovascular,2,3 metabolic,4 psychiatric,5,6 impaired cognition,7,8 quality of life,9 and all-cause mortality.2 Numerous studies have reported on the association between sleep and mortality, most focusing on effects of self-reported sleep duration.3,10,11 Despite emerging evidence of a sleep-mortality association, determining the aspects of sleep driving the association has been a challenge given the multidimensionality of sleep.
As a result, little is known about how the proportion of time spent in rapid eye movement (REM) sleep and non-REM sleep stages N1, N2, and N3 relate to timing or cause of death. However, decreased REM sleep has been linked with poor mental and physical health outcomes.12-20 We postulated that less REM sleep is associated with increased mortality risk and tested this association using data from the Outcomes of Sleep Disorders in Older Men (MrOS) Sleep Study.21 To evaluate consistency and generalizability of the findings, we replicated the analysis in the Wisconsin Sleep Cohort (WSC).22,23
The Stanford University institutional review board ruled this project exempt because the data used were publicly available. Participants of the MrOS and WSC studies provided written informed consent.
MrOS24 is an observational, longitudinal cohort of 5994 community-dwelling men enrolled across 6 centers throughout the US.21,25 Individuals were 65 years or older, able to walk without assistance, and were without bilateral hip replacements.25 The MrOS Sleep Study ancillary project included 3135 men (59.2%) recruited from December 2003 to March 200526; inclusion details for the individuals analyzed are in eFigure 1A in the Supplement. All men provided written informed consent; each site received institutional review board approval.
Data from the WSC study were used for replication. The WSC started in 1988 and is an ongoing, longitudinal, population-based study of the causes, consequences, and natural history of sleep disorders. It was established from a sample of state agency employees in Wisconsin aged 30 to 60 years at enrollment.22,23 Rationale and design were previously published.22
Questionnaires were mailed every 5 years, and a subsample had a sleep study every 4 years. eFigure 1B in the Supplement has a flowchart of the WSC participants analyzed. Informed consent was obtained from all participants under a University of Wisconsin-Madison Health Sciences institutional review board.
In MrOS, an unattended, portable in-home baseline polysomnography was conducted at sleep visit 1 (Safiro; Compumedics).27 The primary exposure was REM sleep, evaluated as percentage of total sleep time, and total number of minutes spent in REM sleep during a single night. Standard polysomnography characteristics were evaluated based on previously published definitions28 (eMethods in the Supplement). Self-reported sleep data included Epworth Sleepiness Scale,29 Pittsburgh Sleep Quality Index,30 Functional Outcomes of Sleep Questionnaire,31 and positive airway pressure use for more than 3 months.
In the WSC study, nocturnal in-laboratory polysomnography was collected using Grass-Telefactor Heritage digital sleep systems. Polysomnography characteristics are described in the eMethods in the Supplement. Because the current analysis included the Epworth Sleepiness Scale (which was collected beginning in 1993), the polysomnography data used were the earliest available with Epworth Sleepiness Scale data (collected via mailed survey).
In MrOS, home sleep-wake patterns were estimated using wrist actigraphy (SleepWatch-O; Ambulatory Monitoring) recorded in proportional integration mode.32 Participants were asked to wear devices continuously on their nondominant wrist for 4 or more consecutive 24-hour periods and to complete sleep logs to track time in and out of bed plus when the device was removed. Sleep logs were used to edit data and set sleep intervals. Actigraphy scoring algorithms used in this study have been previously published.33
Actigraphy measures included mean total sleep time (minutes) while in bed, mean sleep latency (minutes), mean wake after sleep onset while in bed (minutes), mean nighttime sleep efficiency, and mean total sleep time (minutes) outside sleep interval. Data were averaged over the entire period the device was worn to best reflect usual sleep patterns. Actigraphy was not collected for the WSC study.
In MrOS, after baseline, participants were contacted every 4 months to determine vital status. Next of kin were contacted in cases of nonresponse. Reported deaths through 2018 were confirmed by centralized review of death certificates. Of 2872 men with sleep data, 31 (1.1%) were missing information on mortality data and 166 (5.8%) ended participation during the follow-up period and were excluded from analyses. Cause of death was categorized by International Classification of Diseases, Ninth Revision codes as cardiovascular (396.9-442, 966.71, and 785.51), cancer (141.9-208.0), and other (codes not in previous categories). For each analysis of cause-specific mortality, individuals who died of a different cause were censored.
In WSC, deaths through 2018 were identified by matching social security numbers with 2 death record sources: National Death Index and Wisconsin State Bureau of Health Information and Policy, Vital Records Section. Matches on social security number were verified with participants’ age and sex. Cause of death was categorized using the same International Classification of Diseases, Ninth Revision codes as MrOS.
In MrOS, education, race/ethnicity, body mass index (calculated as weight in kilograms divided by height in meters squared), neck and hip circumference, smoking status, weekly alcohol use, and daily caffeine intake were collected at baseline along with the 15-item Geriatric Depression Scale,34 Modified Mini-Mental State Examination35 (evaluated as continuous and binary variable with a score <77 indicating impairment36), and Physical Activity Scale for the Elderly Scale.37
Data on prescription and nonprescription medications were collected at baseline. Each medication was matched to its ingredient(s) using the Iowa Drug Information Service Drug Vocabulary.38 Current use of medications known to affect sleep (antidepressants, benzodiazepines) and sleep medications (nonbenzodiazepines, nonbarbiturate sedative hypnotics) were used.
Self-reported history of physician diagnosis identified hypertension, angina, stroke, heart attack, transient ischemic attack, congestive heart failure, type 2 diabetes, chronic obstructive pulmonary disease, osteoarthritis, and rheumatoid arthritis.
In WSC, body mass index was assessed during the polysomnography visit. Data on education, race/ethnicity, smoking habits, weekly alcohol, and daily caffeine use were obtained by questionnaire. Participants reported current use of antidepressants or sedatives and physician diagnosis of hypertension, coronary artery disease, or heart attack.
Analysis began May 2018 and ended December 2019. Missing data were considered missing at random and imputed using multivariate imputation by chained equations operationalized using the R, version 3.2.5 MICE package (R Foundation).39 Rate of loss to follow-up was less than 10%. Based on actigraphy data, individuals with mean total sleep time more than 8 hours were categorized as long sleepers and total sleep time less than 5 hours as short sleepers.
Cox proportional hazards models were used to assess associations between percentage of REM sleep and all-cause, cardiovascular, cancer, and other mortality. Results are reported as hazard ratios (HRs) with 95% CIs for every 5% decrease in REM sleep (similar to SD and more clinically relevant than a 1% change). Analyses were performed using SAS, version 9.4 (SAS Institute) and R studio version 1.1.463 (R Foundation).
Cox models were built in 2 steps using a combination of clinical knowledge and empirical covariate selection. Model 1 included covariates selected based on known associations including age, race/ethnicity, education, body mass index, smoking status, alcohol, caffeine, medication use (sleep, antidepressant, or benzodiazepine), and study site. Next, more than 60 variables were evaluated using 6-fold cross-validation to select covariates to add to model 1 for final models (model 2).40 At each fold, data from 1 site were withheld and an optimal model identified using the My.stepwise R package (R Foundation). The criterion for entry into the final model was inclusion in at least 3 of 6 optimal models. We chose not to withhold a random sample to gauge whether site was associated with variable selection. All proportional hazards assumptions were met. Potential interactions were explored using random survival forests41,42; potential threshold effects were evaluated using Kaplan-Meier survival curves.43
Sensitivity analyses were conducted to rule out alternative explanations. First, analyses were run after excluding individuals censored in the first 2 years. Next, anyone with an Apnea-Hypopnea Index score greater than 30 and/or using antidepressants, benzodiazepines, or sleep medications was excluded. Another analysis excluded depressed individuals (Geriatric Depression Scale score >4 or antidepressant use). To assess residual confounding from sleep duration, we performed 2 analyses excluding individuals using different total sleep time definitions (total sleep time <5 hours or >8 hours from either polysomnography or actigraphy based on previous MrOS publication26 and total sleep time <6 hours or >8 hours). Finally, models were run using total REM sleep minutes and dichotomized percentage of REM sleep (cut point = 15%; similar to lowest quartile threshold).
Conditional inference survival tree and random survival forest methods were used to explore which sleep stages may be driving significance because sleep stages were interdependent (they add up to 100%). Percentage times in each sleep stage were used as predictors of all-cause mortality in MrOS. Variable importance was calculated from random forest results using mean decreased accuracy. For more details, see the eMethods in the Supplement.
Cox models were repeated using the WSC data set after matching all possible covariates across both data sets. All-cause mortality model was stratified by sex.
The MrOS cohort included 2675 individuals (2675 men [100%] and 2448 white individuals [91.5%]) with a mean (SD) age of 76.3 (5.5) years at baseline and was followed up for a median (interquartile range) of 12.1 (7.8-13.2) years. The mean (SD) age was 86.6 (5.2) years at follow-up. Percentage of REM sleep ranged from 0% to 43.9% and was normally distributed. Overall mean (SD) percentage of REM sleep was 19.2% (6.6%) (mean [SD] time in REM sleep, 69.7 [28.6] minutes) with values increasing from 14.8% in the lowest quartile to 23.6% in the highest. Table 1 reports demographic, lifestyle, and health characteristics varying across REM sleep quartile. Those in the lowest quartile tended to be older with higher rates of antidepressant use, hypertension, heart attack, transient ischemic attack, and lower Physical Activity Scale for the Elderly scores. As expected, most sleep variables varied across quartiles (Table 2).
The WSC cohort included 1386 individuals (753 men [54.3%] and 1311 white individuals [94.6%]) with a mean (SD) age of 51.5 (8.5) years at baseline. Participants were followed up for a median (interquartile range) of 20.8 (17.9-22.4) years (mean [SD] age at follow-up, 70.2 [7.7] years). Minutes in REM sleep (mean [SD], 67.8 [28.9] min) and percentage of REM sleep range (0%-43.0%) were similar, but overall mean percentage of REM sleep (mean [SD], 17.6% [6.5%]) was lower compared with MrOS data, possibly owing to longer total sleep time during in-laboratory vs in-home studies.
WSC individuals were younger, had a mix of men and women, had more obesity (body mass index of 30.7 vs 27.2), consumed more alcohol (3.6 vs 1.9 drinks per week), were more likely to be current or never smokers, and had higher antidepressant and/or sedative use. Despite these differences, most measures had similar distributions across REM sleep quartiles compared with MrOS (Table 3).
In MrOS, 1404 deaths (52.5%) were reported over a median (interquartile range) follow-up of 12.1 (7.8-13.2) years. Regression analysis of percentage of REM sleep as a continuous variable showed a downward trend, reflected in the lowest quartile of percentage of REM sleep having the highest percentage of deaths for each mortality category (eTable 1 in the Supplement). A 13% higher all-cause mortality rate for every 5% reduction in REM sleep (hazard ratio [HR], 1.13; 95% CI, 1.08-1.19) was observed after adjusting for covariates (Table 4). The association persisted for cardiovascular disease–related mortality (HR, 1.11; 95% CI, 1.02-1.20) and other mortality (HR, 1.19; 95% CI, 1.11-1.28) but was not significant for cancer-related mortality (HR, 1.06; 95% CI, 0.96-1.17) (Table 4). Kaplan-Meier curves showed a possible threshold, particularly for cancer-related deaths (Figure). Individuals with less than 15% REM sleep had a higher mortality rate compared with individuals with 15% or more for all mortality definitions (HR range, 1.20-1.35) (eTable 2 in the Supplement).
Sensitivity analyses found the 1776 of 2675 individuals (66.4%) who slept 5 to 8 or 6 to 8 hours had larger effect sizes on all outcomes except cardiovascular disease–related mortality in the 6- to 8-hour group (HR, 1.00; 95% CI, 0.85-1.19). No substantial differences were found analyzing the 2591 of 2675 individuals (96.9%) who survived the first 2 years, 1684 of 2675 individuals (63.0%) without severe sleep apnea or medication use, 2291 of 2675 individuals (85.6%) without depression (eTable 3 in the Supplement), or when using absolute time in REM sleep (data not shown).
There were fewer deaths in the WSC sample (184 [13.3%]), which was expected given younger ages. As with MrOS, those in the lowest REM sleep quartile had the highest percentages of death (eTable 1 in the Supplement). The effect size for 5% reduction in REM sleep on risk of all-cause mortality (HR, 1.17; 95% CI, 1.03-1.34) and noncardiovascular disease, noncancer-related mortality (HR, 1.26; 95% CI, 1.01-1.58) were significant despite younger age, inclusion of both men and women, longer follow-up period, reduced samples size, and event frequency. Effect sizes and direction for cardiovascular disease–related mortality (HR, 1.13; 95% CI, 0.87-1.45) and cancer-related mortality (HR, 1.13; 95% CI, 0.91-1.40) were similar to MrOS, although the smaller sample size widened CIs (Table 4).
When stratified by sex, decreased percentage of REM sleep was associated with all-cause mortality in women for every 5% REM sleep reduction (HR, 1.34; 95% CI, 1.07-1.68) but was not statistically significant in men (HR, 1.09; 95% CI, 0.92-1.30), with these estimates providing modest statistical evidence for a difference (P for interaction = .08).
Individuals with less than 15% REM sleep had a higher mortality rate compared with individuals with 15% or more with odds ratios ranging from 1.36 to 1.78 for all mortality definitions except cardiovascular (HR, 1.00; 95% CI, 0.52-1.90) (eTable 2 in the Supplement).
The first and second nodes in the conditional survival tree were percentage of REM sleep with cut points of 15.4% (similar to the lowest quartile cut point) and 10.9% (eFigure 2 in the Supplement). Percentage of N1 sleep was the fifth node with a cut point of 13.6% (eFigure 3 in the Supplement). The random survival forest model identified percentage of REM sleep as the most important sleep stage for predicting survival (mean decrease in accuracy = 0.058). Percentage of N1 sleep was a distant second at 0.001 (eFigure 3 in the Supplement). Both techniques found percentage of REM sleep overwhelmingly important compared with other stages, implying that contributions from other stages were inconsequential. This is consistent with the Cox results where the REM sleep β coefficient (0.17) was substantially higher than the N2 sleep β coefficient (0.06).
Survival analysis of older, community-based men found an association between less REM sleep and increased mortality, which replicated in an independent data set of middle-aged men and women. Similar effect sizes (HR ranging from 1.13-1.19 per 5% REM sleep decrease) were observed in MrOS for all-cause, cardiovascular, cancer, and other mortality after adjusting for confounding demographic, sleep, and health-related covariates. These effect sizes are slightly larger than mortality risk resulting from aging 1 year (HR for age ranging from 1.11-1.16) based on MrOS data. Sensitivity analyses showed findings persisted in subgroups with sleep duration between 5 to 8 and 6 to 8 hours (except cancer), without depression, severe sleep apnea, and in those not using medications that may affect REM. In sum, decreased REM sleep was an indicator of mortality risk across a broad age range, which is consistent with the study evaluating REM sleep and mortality in the Sleep Heart Health Study.44 When stratified by sex, there was a higher rate of all-cause mortality in women compared with men.
Despite the different outcome measures, our findings are consistent with reports linking REM sleep to other age-related diseases and conditions. Song et al45 found that increased time in N1 sleep and less time in REM sleep were associated with worsening cognitive performance in MrOS. Smagula et al46 evaluated the association between sleep stages and comorbid depression and anxiety in MrOS and found men with clinically significant depressive symptoms spent more time in N2 sleep and less time in REM sleep. Suh et al47 found a short average cycle length (sequence of non-REM and REM sleep stages, both >2 minutes and not interrupted by >2 minutes of wake) were significantly associated with cognitive decline.48 While not directly associated with mortality, these studies support the value of REM sleep and the importance of evaluating sleep stages independently from sleep duration.
In contrast, the HypnoLaus study49 evaluated the association between different sleep stages and hypertension, diabetes, overweight/obesity, and metabolic syndrome. Although initially a higher prevalence of metabolic syndrome was observed in individuals with decreased REM sleep, after multivariate adjustment the study concluded that normal variations in sleep stages contribute little to metabolic syndrome and associated disorders.49
In MrOS, mean total sleep time actigraphy (in bed) was lower than expected. When combined with mean total sleep time out of bed, total daily total sleep time was similar to reports from a different, similar aged cohort.50 A meta-analysis10 suggested sleep duration is associated with obesity,51 hypertension,52 cardiovascular outcomes,53 and all-cause mortality.2 A 2017 meta-analysis linked sleep duration and mortality.54 Therefore, we performed sensitivity analyses limiting the population to participants who slept 5 to 8 and 6 to 8 hours per night and found no change in effect sizes except for cancer in the 6- to 8-hour group. Further, obstructive sleep apnea has been linked to mortality, stroke, and cardiovascular disease.55-58 We found the REM sleep and all-cause mortality association remained significant after excluding individuals with Apnea-Hypopnea Index score greater than 30 and/or using antidepressants, benzodiazepines, or sleep medications.
Dew et al59 found an association between mortality and sleep latency more than 30 minutes and sleep efficiency less than 80%. Wallace et al60 evaluated which sleep characteristics predicted mortality. Rhythmicity and continuity were the strongest; however, sleep stages were not evaluated.
Sun et al61 used machine learning to predict brain age from sleep electroencephalography. Results for healthy individuals correlated well with chronological age. However, individuals with significant neurologic or psychiatric disease, hypertension, or diabetes had higher brain age compared with true age. Percentage of REM sleep may be another important, easy to interpret aging biomarker. Algorithms are currently being developed to accurately measure percentage of REM sleep using consumer wearable devices, which will reduce barriers for evaluating REM sleep in the general population.
This study has many strengths. Parallel analyses were conducted in 2 instrumental, well-characterized, population-based sleep cohorts. Machine learning was used to strengthen the analysis, and numerous sensitivity analyses were conducted to control for potential biases.
Limitations include the possibility of unmeasured and residual confounding. To address this concern, we used clinical knowledge and empirical model building to select covariates for the final models. MrOS did not include women, and the population’s mean (SD) age at baseline was 76.4 (5.5) years. However, replication in the WSC expanded the generalizability of the findings to include middle-aged men and women (mean [SD] age, 51.5 [8.5] years). Both MrOS and WSC comprised community-dwelling volunteers and therefore may be healthier than the general population; however, we adjusted for comorbidities and do not believe this affects the associations presented. Replication provides generalizability across a larger age group and is associated with reduction, not an elimination, of the likelihood of reverse causality. Generalizability to other races/ethnicities is limited because more than 90% of both cohorts were white. REM sleep was quantified based on 2 night of polysomnography. Although it is possible the first-night effect biased our results, it is unlikely the effect would be differential with respect to mortality. Also, another study using the MrOS polysomnography protocol found no evidence of first-night effect.62
A robust association was found between percentage of REM sleep and mortality in 2 independent cohorts, which persisted across different causes of death and multiple sensitivity analyses. Given the complex underlying biologic functions, further studies are required to understand whether the relationship is causal. Accelerated brain aging may result in less REM sleep, making it a marker rather than a direct mortality risk factor; however, mechanistic studies are needed. Strategies to preserve REM sleep may influence clinical therapies and reduce mortality risk, particularly for adults with less than 15% REM.
Corresponding Author: Eileen B. Leary, PhD, RPSGT, Stanford Center for Sleep Sciences and Medicine, 3165 Porter Dr, MC 5480, Palo Alto, CA 94304 (firstname.lastname@example.org).
Accepted for Publication: March 6, 2020.
Published Online: July 6, 2020. doi:10.1001/jamaneurol.2020.2108
Correction: This article was corrected on August 17, 2020, to fix an error in the Abstract Results.
Author Contributions: Dr Leary had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Leary, Watson, Ancoli-Israel, Redline, Zou, Mignot, Stone.
Acquisition, analysis, or interpretation of data: Leary, Watson, Redline, Yaffe, Ravelo, Peppard, Goodman, Mignot, Stone.
Drafting of the manuscript: Leary, Watson, Mignot.
Critical revision of the manuscript for important intellectual content: Watson, Ancoli-Israel, Redline, Yaffe, Ravelo, Peppard, Zou, Goodman, Mignot, Stone.
Statistical analysis: Leary, Watson, Ravelo, Zou, Mignot.
Obtained funding: Redline, Peppard, Mignot, Stone.
Administrative, technical, or material support: Leary, Watson, Redline, Peppard.
Supervision: Watson, Ancoli-Israel, Mignot, Stone.
Conflict of Interest Disclosures: Dr Ancoli-Israel reports consulting for Eisai and Merck outside the submitted work. Dr Redline reports grants and personal fees from Jazz Pharmaceuticals, consulting fees from Respicardia, and personal fees from Eisai outside the submitted work. Dr Peppard reports grants from the National Institutes of Health during the conduct of the study. Dr Mignot is a consultant and a principal investigator for clinical trials involving Merck, Jazz Pharmaceuticals, and Takeda and owns stocks in Dreem, Inexia, Orexia, and Alerion. Dr Stone reports grants from the National Institutes of Health during the conduct of the study and grants from Merck outside the submitted work. No other disclosures were reported.
Funding/Support: The Osteoporotic Fractures in Men Study is supported by the National Institutes of Health. Support also came from the National Institute on Aging, the National Institute of Arthritis and Musculoskeletal and Skin Diseases, the National Center for Advancing Translational Sciences, and National Institutes of Health Roadmap for Medical Research (grants U01 AG027810, U01 AG042124, U01 AG042139, U01 AG042140, U01 AG042143, U01 AG042145, U01 AG042168, U01 AR066160, and UL1 TR000128). The National Heart, Lung, and Blood Institute provided funding for the Osteoporotic Fractures in Men Study Sleep ancillary study Outcomes of Sleep Disorders in Older Men (grants R01 HL071194, R01 HL070848, R01 HL070847, R01 HL070842, R01 HL070841, R01 HL070837, R01 HL070838, and R01 HL070839). The Wisconsin Sleep Cohort was supported by the National Institutes of Health (grants R01HL62252, RR03186, and R01AG14124). Dr Redline was partially supported by the National Heart, Lung, and Blood Institute (grant R35 HL135818).
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.