Association of Circadian Abnormalities in Older Adults With an Increased Risk of Developing Parkinson Disease

Key Points

Question  Are circadian abnormalities in older adults associated with an increased risk of developing Parkinson disease over time?

Findings  In this longitudinal study of 2930 community-dwelling older men without Parkinson disease at baseline, the risk of incident Parkinson disease increased significantly with decreasing circadian amplitude, mesor, or robustness. Participants in the lowest quartile for these measures had approximately 3 times the risk of developing Parkinson disease compared with those in the highest quartile.

Meaning  Circadian rhythm disruption in elderly individuals may represent an important prodromal feature for Parkinson disease, and future studies should test whether circadian disruption could also be a risk factor for Parkinson disease and whether strategies to improve circadian function affect the risk of Parkinson disease.

Importance  Disruption in circadian activity rhythms is very common in older adults, particularly among those with neurodegenerative diseases, including Parkinson disease (PD). However, whether circadian disruption could be a prodrome for PD is unclear.

Objective  To determine the association between rest-activity rhythm (RAR) and risk of incident PD and to explore whether this association is independent of nighttime sleep disturbances.

Design, Setting, and Participants  The ancillary sleep study of the longitudinal cohort Osteoporotic Fractures in Men Study (MrOS) was conducted from December 1, 2003, to March 31, 2005. Of the 3135 community-dwelling men enrolled in the MrOS sleep study, 3049 had technically adequate RAR data; of these, 119 were excluded for having prevalent PD or missing incident data, leaving 2930 men without PD at baseline. Data were analyzed from February 1 through August 31, 2019.

Exposures  Twenty four–hour RAR parameters (amplitude, mesor, robustness, and acrophase) generated by wrist actigraphy–extended cosinor analysis.

Main Outcomes and Measures  Incident PD based on physician diagnosis. Multivariable logistic regression was used to determine the association between quartiles of RAR parameters and risk of incident PD.

Results  Among the 2930 men included in the analysis (mean [SD] age, 76.3 [5.5] years), 78 (2.7%) developed PD during 11 years of follow-up. After accounting for all covariates, the risk of PD increased with decreasing circadian amplitude (strength of the rhythm) (odds ratio [OR] per 1-SD decrease, 1.77; 95% CI, 1.30-2.41), mesor (mean level of activity) (OR per 1-SD decrease, 1.64; 95% CI, 1.22-2.21), or robustness (how closely activity follows a cosine 24-hour pattern) (OR per 1-SD decrease, 1.54; 95% CI, 1.14-2.07) (P < .005 for trend). Those in the lowest quartile of amplitude, mesor, or robustness had approximately 3 times the risk of developing PD compared with those in the highest quartile of amplitude (OR, 3.11; 95% CI, 1.54-6.29), mesor (OR, 3.04; 95% CI, 1.54-6.01), and robustness (OR, 2.65; 95% CI, 1.24-5.66). The association remained after further adjustment for nighttime sleep disturbances and duration in the lowest compared with the highest quartile (OR for amplitude, 3.56 [95% CI, 1.68-7.56]; OR for mesor, 3.24 [95% CI, 1.52-6.92]; and OR for robustness, 3.34 [95% CI, 1.45-7.67]). These associations were somewhat attenuated, but the pattern remained similar after excluding PD cases developed within 2 years after baseline in the lowest compared with the highest quartile (OR for amplitude, 2.40 [95% CI, 1.15-5.00]; OR for mesor, 2.76 [95% CI, 1.35-5.67]; and OR for robustness, 2.33 [95% CI, 1.07-5.07]). Acrophase was not significantly associated with risk of PD.

Conclusions and Relevance  In this cohort study, reduced circadian rhythmicity was associated with an increased risk of incident PD, suggesting it may represent an important prodromal feature for PD. Future studies are needed to determine whether circadian disruption could also be a risk factor for PD and whether strategies to improve circadian function affect the risk of PD.

Parkinson disease (PD) represents one of the fastest-increasing number of cases among neurological disorders, characterized by early death of dopaminergic neurons in the substantia nigra.¹ Importantly, PD pathology involves much more extensive brainstem neurotransmitter systems far outside the basal ganglia and occurs years or decades before dopaminergic neuronal death.² The multiple neuronal groups and areas affected by PD pathology, such as cholinergic, serotoninergic, and noradrenergic neurons in the brainstem and orexinergic neurons in the posterolateral hypothalamus,^3-6 play a key role in the development of sleep-wake disturbances, which is the most common nonmotor symptom in patients with PD.⁷

Notably, much less is known about the association between PD and circadian rhythms, the key regulator of sleep-wake cycles. Circadian rhythms change with aging, including decreased amplitude and robustness, high fragmentation, and advanced circadian timing, all of which may further accelerate the aging process.⁸ In accordance with pathological changes in sleep- and wake-promoting brain regions in early PD, circadian disruption among patients with PD is often much more severe than in healthy older adults and may occur even in very early stages of the disease.^9-11 However, it is unknown whether circadian abnormalities might precede the development of PD.¹²

Evolving evidence indicates that daytime napping and nighttime sleep disorders, including insomnia and rapid eye movement (REM) sleep behavior disorder, might be prodromal features for PD.^6,13 Because nighttime sleep and daytime napping are both regulated in part by circadian rhythms, it is plausible that circadian dysfunction is a prodrome or a risk factor for PD and may be more directly involved in the disease process. No study to date, to our knowledge, has examined the association between 24-hour rest-activity rhythmicity, an important behavioral marker of circadian rhythms, and subsequent risk of PD in healthy older adults. Understanding whether circadian disruptions are a prodrome for PD independent of sleep disturbances may have implications for the early detection and management of PD. In this large prospective cohort study of community-dwelling older men, we aimed to determine the association between rest-activity rhythms (RARs) assessed objectively by wrist actigraphy and the risk of developing PD during an 11-year follow-up and to explore whether this association is independent of sleep disturbances.

The Osteoporotic Fractures in Men Study (MrOS) enrolled 5994 community-dwelling men 65 years or older from 2000 to 2002 at 6 clinical centers in the United States, including Birmingham, Alabama; Minneapolis, Minnesota; Palo Alto, California; the Monongahela Valley near Pittsburgh, Pennsylvania; Portland, Oregon; and San Diego, California.¹⁴ From 2003 to 2005, a total of 3135 men were recruited into the ancillary MrOS sleep study and received comprehensive sleep assessments. Of these participants, 3049 men (97.3%) had technically adequate RAR data collected from December 1, 2003, to March 31, 2005. Of these, 119 were excluded from the analysis (62 because they had PD at baseline and 57 because they did not have follow-up data), leaving 2930 in our analytic cohort. All men provided written informed consent, and the study was approved by the institutional review board at each clinic site. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

We examined RAR using a commercially available actigraph (SleepWatch-O; Ambulatory Monitoring, Inc), a small device worn on the wrist. Movement is measured by a piezoelectric linear accelerometer (sensitive to ≥0.003 g), which generates a voltage each time the actigraph is moved. These voltages are gathered continuously and summarized during 1-minute epochs. Activity data were collected in the proportional integration mode, which computes movement as counts per minute based on an area under the curve analysis that accounts for intensity and frequency of movement. Men were asked to wear the actigraphs continuously for a minimum of three 24-hour periods.

An extension to the traditional cosine model that has been used extensively in prior studies of the MrOS participants^15-17 was used to map the circadian activity rhythm to the activity data. Activity data often assume a shape more similar to a squared wave than a cosine curve, and this extension allows for this shape. Improvement in fit of the extended cosine model compared with the traditional cosine model was formally tested¹⁵ and was found to be an improvement in 96.3% of the men in our study sample. The RAR parameters of the extended cosine model were calculated using nonlinear least squares and included the following measurements. First, amplitude, a measure of the strength of the rhythm, is the peak-to-nadir difference in activity at the point of greatest activity of the fitted curve (measured in arbitrary units of activity [counts/min]). Second, mesor is the mean level of activity of the fitted curve (measured in arbitrary units of activity [counts/min]). Third, a pseudo-F statistic was used for goodness of extended cosine fit or robustness of the rhythm (higher pseudo-F values indicate stronger rhythms). Fourth, acrophase, timing of peak activity of the fitted curve, was measured in portions of hours (time of day). Amplitude, mesor, and robustness were examined as continuous variables and as quartiles. Acrophase was examined in terms of the deviation from the population mean of peak timing of activity. We identified 3 categories based on having a peak time of more than 1 SD above and below the population mean for the study population. Phase-advanced participants were defined as having an acrophase of 1 SD or less from the mean (1:04 pm), and phase-delayed participants were defined as having an acrophase of greater than 1 SD from the mean (3:29 pm).

Sleep variables were also computed from the actigraphy data. Sleep efficiency, a measure of sleep quality, was defined as the percentage of time asleep after “lights off.” Total sleep time was defined as the time asleep from lights off to final awakening. Mean values for sleep efficiency and total sleep time were calculated for all nights the participant wore the device to obtain a more representative characterization of usual sleep patterns.

During the follow-up (5 visit- or questionnaire-based contacts during approximately 11 years), participants were asked to report whether they ever had PD diagnosed by a physician. In addition, at visits, participants were asked to bring in all prescription and nonprescription medications taken in the past 30 days, and these were entered into an electronic database with each matched to its ingredient(s) based on the Iowa Drug Information Service Drug Vocabulary.¹⁸ Medication use for PD was derived from this database and defined as the use of carbidopa/levodopa, pramipexole dihydrochloride, ropinirole hydrochloride, rotigotine, benztropine mesylate, selegiline hydrochloride, rasagiline mesylate, amantadine hydrochloride, entacapone, rivastigmine tartrate, or trihexyphenidyl hydrochloride.¹³ Incident PD was defined as physician-diagnosed PD in the primary analysis and as having a physician diagnosis and PD medication use in sensitivity analyses to verify physician-diagnosed PD. Follow-up time was calculated per participant as the time to the most recent collection of data on physician-diagnosed PD.

All participants completed questionnaires at the time of the sleep visit (baseline for this analysis), which included items about demographics, medical history, physical activity, smoking, caffeine intake, and alcohol use. The Geriatric Depression Scale was used to assess depressive symptoms (scores range from 0 to 15, with higher scores indicating increased depressive symptoms).¹⁹ Level of physical activity was assessed using the Physical Activity Scale for the Elderly (scores range from 0 to 793, with higher scores indicating greater physical activity).²⁰ Use of sleep medication and benzodiazepines were derived from the medication database. Functional status was assessed by collecting information on 5 instrumental activities of daily living.^21,22 Caffeine intake was based on answers to questions regarding intake of caffeinated coffee, tea, and soda.²³ Body mass index was calculated as weight in kilograms divided by height in meters squared. A history of coronary heart disease was defined as a prior diagnosis of myocardial infarction, angina, or congestive heart failure. The Modified Mini-Mental State Examination was used to measure global cognitive function (scores range from 0 to 100, with higher scores indicating better cognitive functioning).²⁴ Severity of sleep-disordered breathing was measured with polysomnography and defined with the apnea-hypopnea index (number of apneas plus hypopneas per hour of sleep associated with a desaturation of ≥3%) as described previously.²⁵ Periodic limb movement index was also measured with polysomnography and was used to indicate the frequency of periodic limb movements in sleep.²⁶

Data were analyzed from February 1 through August 31, 2019. Characteristics of participants were compared by categories of RAR using χ² tests for categorical variables, analysis of variance for normally distributed continuous variables, and Kruskal-Wallis tests for continuous variables with skewed distributions. To determine the association between RAR and incident PD, we used logistic regression to estimate odds ratios (ORs) and 95% CIs. The RAR parameters were analyzed as continuous values with the OR expressed per 1-SD decrease and by quartiles. The highest quartile served as reference, and a test for trend across quartiles was also completed with quartile entered into the model as a single multilevel variable. Models for all analyses included age and clinic. Variables that were significantly related (2-sided P < .10) to at least 1 RAR parameter measure were included as potential confounders in the multivariable analyses. These covariates included demographic characteristics, educational level, depressive symptoms, body mass index, physical activity, benzodiazepine use, alcohol and caffeine intake, smoking, cognitive function, and a history of hypertension, stroke, coronary heart disease, or type 2 diabetes.

We performed a number of sensitivity analyses, including (1) adjusting further for sleep efficiency, total sleep time, apnea-hypopnea index, and periodic limb movement index to determine whether associations were independent of sleep indices; (2) a lag time of approximately 2 years, only including PD cases identified after the first follow-up visit following the measurement of RAR to ensure that the risk factor preceded the PD diagnosis; and (3) including only PD cases that had been identified by both physician diagnosis and PD medication use. In addition, we used likelihood ratio tests to examine whether the addition of RAR parameters improved the fit of the model adjusted for all covariates and sleep disturbances. All significance levels reported were 2 sided, and all analyses were conducted using SAS, version 9.4 (SAS Institute, Inc).

The 2930 men had a mean (SD) age of 76.3 (5.5) years and were mostly white (2636 [90.0%]). Figure 1 shows the activity level by time, with fitted curve for the RAR parameters, in participants with and without PD. Men in the lowest quartile of amplitude (Table 1) were older (mean [SD] age, 78.1 [5.9] vs 74.7 [4.8] years for quartile 4) and more likely to have an educational level of less than high school (53 of 732 [7.2%] vs 29 of 732 [4.0%] for quartile 3), less heavy consumption of alcohol (35 of 732 [4.8%] vs 38 of 732 [5.2%] for quartile 3) and caffeine (mean [SD], 211.0 [234.4] vs 267.5 [254.5] mg/d for quartile 4), lower physical activity score (mean [SD], 118.0 [66.7] vs 172.3 [71.3] for quartile 4), worse Modified Mini-Mental State Examination score (mean [SD], 91.7 [7.3] vs 93.3 [5.4] for quartile 2), higher body mass index (mean [SD], 27.9 [4.2] vs 26.7 [3.6] for quartile 4), higher Geriatric Depression Scale score (mean [SD], 2.4 [2.5] vs 1.4 [1.9] for quartiles 3 and 4), at least 1 impairment in instrumental activities of daily living (269 of 732 [36.7%] vs 84 of 733 [11.5%] for quartile 4), and more comorbidities, including stroke (40 of 732 [5.5%] vs 19 of 733 [2.6%] for quartile 4), hypertension (412 of 732 [56.3%] vs 323 of 733 [44.1%] for quartile 4), coronary heart disease (253 of 732 [34.6%] vs 161 of 733 [22.0%] for quartile 4), and type 2 diabetes (124 of 732 [16.9%] vs 72 of 733 [9.8%] for quartile 4). In addition, they had lower sleep efficiency (mean [SD], 74.5% [15.5%] vs 79.9% [9.9%] for quartile 3), shorter total sleep time (mean [SD], 6.22 [1.53] vs 6.53 [1.12] hours for quartile 2), and a higher apnea-hypopnea index (mean [SD], 19.7 [17.0] vs 15.4 [13.3] for quartile 4). Similar patterns of baseline characteristics were observed for those with a lower mesor or robustness and for those with a delayed circadian activity phase.

During a total follow-up of 11 years, we observed 78 incident cases of PD (2.7%). Figure 2 shows the multivariable-adjusted ORs of incident PD associated with each RAR measure. After adjustment for demographics, site, educational attainment, depressive symptoms, body mass index, physical activity, benzodiazepine use, alcohol and caffeine consumption, smoking, comorbidities, and baseline cognition, the risk of incident PD increased with decreasing circadian amplitude, mesor, or robustness. The risk of PD increased in association with every 1-SD decrease in circadian amplitude (OR, 1.77; 95% CI, 1.30-2.41), mesor (OR, 1.64; 95% CI, 1.22-2.21), or robustness (OR, 1.54; 95% CI, 1.14-2.07). Those in the lowest quartile of amplitude, mesor, or robustness had triple the risk of developing PD compared with those in the highest quartile of amplitude (OR, 3.11; 95% CI, 1.54-6.29), mesor (OR, 3.04; 95% CI, 1.54-6.01), or robustness (OR, 2.65; 95% CI, 1.24-5.66). We did not observe an association between acrophase and risk of PD (Figure 2).

These associations were strengthened after further adjustment for nighttime sleep including sleep efficiency, total sleep time, apnea-hypopnea index, and periodic limb movement index. The addition of RAR significantly improved the fit of the multivariable model containing nighttime sleep for the lowest vs the highest quartile for amplitude (OR, 3.56; 95% CI, 1.68-7.56), mesor (OR, 3.24; 95% CI, 1.52-6.92), and robustness (OR, 3.34; 95% CI, 1.45-7.67). After excluding cases with PD developed within the first 2 years after baseline, the association was somewhat attenuated, although the pattern of the results was consistent with the analysis before exclusion, and there remained a significantly increased risk of PD associated with decreasing circadian amplitude for the lowest vs the highest quartile (OR, 2.40; 95% CI, 1.15-5.00), mesor (OR, 2.76; 95% CI, 1.35-5.67), or robustness (OR, 2.33; 95% CI, 1.07-5.07) (Table 2). The study of only PD cases confirmed by both physician diagnosis and PD medication use (n = 58) showed similar results, with more than double the risk of PD observed for those in the lowest quartiles of amplitude (OR, 2.47; 95% CI, 1.12-5.41) and mesor (OR, 2.29; 95% CI, 1.06-4.94) compared with the highest quartiles.

In this study of community-dwelling older men without PD, decreased circadian amplitude, mesor, or robustness at baseline were consistently associated with a higher risk of developing PD during an 11-year follow-up. Men in the lowest quartile of these RAR parameters had triple the risk of PD compared with those in the highest quartile. This association was independent of several confounders, including nighttime sleep disturbances. We did not find an association between the measure of circadian phase (acrophase) and risk of PD. Reduced circadian rhythmicity might be an important prodromal feature in incident PD and help with the early detection of the disease.

To our knowledge, this study is the first to report a longitudinal association between circadian disruption and subsequent risk of PD in community-dwelling older adults. Prior cross-sectional studies suggested that patients with PD tend to have lower circadian amplitude and mesor but no major shift in circadian phases and that these abnormalities occur early in the disease process.^9-11,27-29 Recent findings from the Rotterdam study³⁰ suggested an association between poor sleep quality and increased risk of PD, particularly in the first 2 years of follow-up, that is attenuated over longer periods of time. To determine whether circadian rhythm abnormalities precede the development of PD, we excluded all cases with PD at study baseline and in secondary analysis introduced a time lag of 2 years. Although the association was somewhat attenuated in secondary analysis, there remained more than double the risk of incident PD in the lowest compared with the highest quartile of RAR parameters. This indicated that weakened circadian rhythmicity rather than shifts in circadian activity timing might be prodromal features of PD.

Interestingly, a prior study by Tranah et al³¹ suggested that older women with reduced circadian amplitude or robustness or delayed phase had a 57% or 83% increased risk of dementia or mild cognitive impairment during 4.9 years of follow-up. Although reduced circadian rhythmicity may be a prodrome for both incident dementia and PD, the association was much stronger for PD than for dementia. In addition, the present study had a long lag time of 11 years, highlighting the potential value of circadian rhythmicity as an early prognostic marker for PD. Furthermore, in line with our findings, prior studies¹² have mostly linked circadian phase shift with dementia but not with PD. Future research should investigate the value of different circadian parameters for estimating the risk of different neurodegenerative diseases.

Growing evidence from animal studies also have shown that changes in circadian-related neurons are already present during the asymptomatic stage of PD. According to the Braak staging model of PD,³² abnormal aggregates of α-synuclein protein begin in the medulla and olfactory bulb before ascending to brainstem structures that regulate the body’s circadian clock and eventually involve the substantial nigra and basal ganglia, when motor symptoms appear. This model has resulted in a wide prodromal time window during which ongoing pathology may lead to nonmotor symptoms, including circadian changes years or decades before clinical detection of PD. For example, in a transgenic PD mouse model, reduced amplitude of the electrical activity of neurons within the suprachiasmatic nucleus was found before the onset of motor symptoms.³³ In addition, degeneration of the serotonergic neurons, which are important arousal neurotransmitters in the midbrain, has been found in PD Braak stage II before substantia nigra degeneration.³² Loss of γ-aminobutyric acid–containing neurons in the ventral tegmental area and nucleus incertus might also result in reduced circadian amplitude in the prodromal stage of PD.^6,34

During the last decade, sleep-wake disturbances (eg, REM sleep behavior disorder and sleep-related movement disorders) have been increasingly recognized as prodromes for PD.^7,35-37 Of note, REM sleep behavior disorder is relatively rare and is challenging to screen for in the general population.^6,38 Meanwhile, circadian disruption can be easily noticed as excessive daytime naps, reduced daily activity, more nighttime sleep fragmentation, or irregular sleep timing and is thus a more pragmatic marker to consider for community-dwelling older adults. Moreover, the association between circadian disruption and PD remained robust even after accounting for other sleep indices, such as sleep efficiency, sleep apnea, and periodic limb movements in sleep; the addition of RAR parameters improved the fit of the multivariable model containing these sleep indices. These findings suggest that the role of circadian disruption as an early marker for PD was independent of and added additional value to traditional sleep indices.

Notably, we cannot rule out the possibility that circadian disruption itself is a risk factor for PD. For example, sleep-wake disturbances have been associated with increased levels of α-synuclein and PD pathology at autopsy.^39,40 However, in the present study, the association between RAR and the risk of PD was independent of sleep disturbances. Furthermore, mouse studies have suggested that the targeted deletion of the circadian clock gene could cause degenerationlike neuropathology without affecting the sleep-wake rhythms.⁴¹ Therefore, sleep-wake disturbances alone are unlikely to explain the link between circadian disturbances and development of PD. Circadian disruption might also lead to PD through the dysregulation of immune and protein homeostasis in the brain or increased oxidative stress.^41-45 Future studies are needed to identify more specific biological mechanisms linking circadian disruption and PD. If circadian dysfunction is proven to be a risk factor for PD, then strategies to strengthen circadian output and boost synchronization of the central and peripheral circadian clocks might provide clues for the prevention and management of PD. Such strategies should be tested in randomized trials.

This study has several strengths, including the long follow-up period combined with rigorous methods to assess 24-hour circadian activity rhythms. The study population was not selected on the basis of sleep disorders or PD diagnosis. Several limitations also warrant consideration. We used a similar approach to a previous study¹³ and relied on physician diagnosis to determine incident PD and thus might have missed or misclassified some cases. This misclassification could have led to an underestimation of the association. Notably, an external validation study⁴⁶ suggested that most of the cases identified using the current approach can be confirmed by the treating neurologists or by expert medical record review. To quantify the effect of using self-report of physician-diagnosed PD, we included only PD cases confirmed by both physician diagnosis and medication use in sensitivity analyses and found similar results. In this study, we did not have information on the exact time when PD cases were first diagnosed and thus could not compare the incident rate of PD. It is also difficult to determine the exact time window between circadian dysfunction and occurrence of PD, given that the disease has a long preclinical time window and often progresses gradually over time. However, the long follow-up period together with our robust findings from the 2-year time lag analysis suggest that these circadian abnormalities preceded the development of clinical PD. Finally, these findings might not be generalizable to women or to younger or more ethnically diverse populations.

We found for the first time, to our knowledge, a robust association between weakened circadian rhythmicity and increased risk of developing PD during an 11-year follow-up in community-dwelling older men. Our results suggest that the reduced amplitude and/or robustness of the rhythms rather than disrupted timing (acrophase) are most indicative of a subsequent risk of PD, independent of nighttime sleep disturbances. Markers of circadian rhythmicity might be valuable as a prodromal feature to help with the early detection of PD. Future studies are needed to explore underlying mechanisms and to determine whether circadian disruption itself might contribute to the development of PD. If confirmed to be a risk factor for PD, then circadian rhythmicity could be a promising intervention target and will open new opportunities for the prevention and management of PD.

Accepted for Publication: March 18, 2020.

Corresponding Author: Yue Leng, MD, PhD, Department of Psychiatry, University of California, San Francisco, 4150 Clement St, San Francisco, CA 94121 (yue.leng@ucsf.edu).

Published Online: June 15, 2020. doi:10.1001/jamaneurol.2020.1623

Author Contributions: Dr Leng and Ms Blackwell had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Leng, Ancoli-Israel, Stone, Yaffe.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Leng.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Blackwell, Stone.

Obtained funding: Leng, Cawthon, Stone.

Administrative, technical, or material support: Cawthon.

Supervision: Ancoli-Israel, Stone, Yaffe.

Conflict of Interest Disclosures: Dr Leng reported receiving grants from National Institute on Aging (NIA) during the conduct of the study and grants from Global Brain Health Institute, the Alzheimer’s Association, and the Alzheimer’s Society outside the submitted work and from the University of California, San Francisco, Weill Institute for Neurosciences during the conduct of the study. Ms Blackwell reported receiving grants from the National Institutes of Health (NIH) and Merck & Co during the conduct of the study. Dr Cawthon reported receiving grants from Abbott Laboratories and Nestle and personal fees from BioAge Labs, Inc, outside the submitted work. Dr Ancoli-Israel reported serving as a consultant on the scientific advisory board for Eisai Co, Ltd, and Merck & Co outside the submitted work. Dr Stone reported receiving grants from Merck & Co outside the submitted work. Dr Yaffe reported receiving grants from the NIA during the conduct of the study; receiving personal fees from the Data and Safety Monitoring Board and Alector, Inc, outside the submitted work; and serving as a member of the Beeson Scientific Advisory Board and the Global Council on Brain Health.

Funding/Support: This work was supported by grants U01 AG027810, U01 AG042124, U01 AG042139, U01 AG042140, U01 AG042143, U01 AG042145, U01 AG042168, U01 AR066160, UL1 TR000128, 1K99AG056598 (Dr Leng), and K24AG031155 (Dr Yaffe) from the NIA, the National Institute of Arthritis and Musculoskeletal and Skin Diseases, and the National Center for Advancing Translational Sciences; grants R01 HL071194, R01 HL070848, R01 HL070847, R01 HL070842, R01 HL070841, R01 HL070837, R01 HL070838, and R01 HL070839 for the NIH Roadmap for Medical Research in the ancillary study Outcomes of Sleep Disorders in Older Men of the Osteoporotic Fractures in Men Study from the National Heart, Lung, and Blood Institute; grants R21 AG051380 and R01 AG034157 from the NIA (data analysis); and the Weill Pilot Award (Dr Leng from the University of California, San Francisco Weill Institute for Neurosciences.

Role of the Funder/Sponsor: The sponsors 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.