Abstract
Introduction: Short and long self-reported sleep durations are associated with a higher risk of stroke, but the association between objective estimates of sleep and 24-h activity rhythms is less clear. We studied the association of actigraphy-estimated sleep and 24-h activity rhythms with the risk of stroke in a population-based cohort of middle-aged and elderly. Methods: We included 1,718 stroke-free participants (mean age 62.2 ± 9.3 years, 55.1% women) from the prospective, population-based Rotterdam Study. Actigraphy-estimated sleep (total sleep time, sleep efficiency, sleep onset latency, and wake after sleep onset) and 24-h activity rhythms (interdaily stability, intradaily variability, and onset of the least active 5 h) were measured during a median of 7 days (Q1–Q3: 6–7 days). The association of sleep and 24-h activity rhythms with risk of stroke was analyzed using Cox proportional hazards models. Results: During a mean follow-up of 12.2 years (SD: 3.2), 105 participants developed a stroke, of whom 81 had an ischemic event. Although there was no clear association between actigraphy-estimated sleep and the risk of stroke, a more fragmented 24-h activity rhythm was associated with a higher risk of stroke (hazard ratio [HR] per SD increase 1.28, 95% confidence interval [CI] 1.07–1.53). A less stable (HR per SD increase in stability 0.78, 95% CI: 0.63–0.97) and more fragmented (HR 1.28, 95% CI: 1.04–1.58) 24-h activity rhythm was also associated with a higher risk of ischemic stroke. Conclusions: Disturbed 24-h activity rhythms, but not sleep, are associated with a higher risk of stroke in middle-aged and elderly persons. This suggests that unstable and fragmented activity rhythms may play a more prominent role in the risk of stroke than sleep per se.
Introduction
Sleep-wake disturbances are increasingly acknowledged as potentially modifiable targets for cardiovascular diseases, including stroke [1, 2]. Several meta-analyses report an association of both a short and long self-reported sleep duration with a higher risk of stroke [3, 4]. Yet, research typically only focusses on sleep duration and is mostly based on self-reported estimates of sleep. Objective markers of sleep might further improve our understanding of the relation between sleep and stroke. In addition, characteristics of sleep other than sleep duration are also known to affect health [5], including cardiovascular health. Sleep-wake disturbances are, for example, linked to development of atherosclerosis and small vessel disease, potentially via increased sympathetic activity and systemic inflammation [6‒8]. However, studies investigating their link with stroke are scarce. Evidence from the Sleep Heart Health Study suggests that increased wake after sleep onset and poor sleep efficiency measured with polysomnography are also linked to a higher risk of stroke [9], but these findings have not been confirmed in other studies.
Moreover, the circadian rhythm is closely related to – but different from – sleep and is behaviorally reflected by the 24-h activity rhythm [10, 11]. Disturbances in these 24-h activity rhythms have been linked to a higher risk of mortality, including cardiovascular mortality and incident cardiovascular disease [12‒15]. However, there is a lack of prospective studies that investigate whether 24-h activity rhythms contribute to the risk of stroke specifically [16]. In this study, we determined the association of actigraphy-measured sleep and 24-h activity rhythms with the risk of stroke in a population-based sample of middle-aged and elderly persons.
Methods
Study Setting and Population
The current study was embedded within the population-based Rotterdam Study, an ongoing prospective cohort of middle-aged and elderly persons in Rotterdam, the Netherlands. This cohort aimed to investigate determinants and consequences of aging and age-related disease [17].
Between 2004 and 2007, 2,632 participants were invited to wear an actigraph to objectively estimate sleep and 24-h activity rhythms. Of these, 2,082 (79.1%) agreed (Fig. 1). Older aged persons (mean age 68.4 years vs. 62.3 years, p < 0.001) and women (60.0% vs. 54.9%, p = 0.038) were more likely to refuse participation in the actigraphy study. We excluded 301 (11.4%) participants with (i) fewer than 4 days of measurement, (ii) actigraph malfunctioning or loss, or (iii) measurements during the week of the change to or from daylight savings. Additionally, we excluded 58 (2.2%) participants who had experienced a (first-ever) stroke before the study baseline. Another 5 (0.2%) participants did not give informed consent for follow-up, resulting in a final sample of 1,718 participants (65.3%). Included participants were followed until the occurrence of a first-ever stroke, loss to follow-up, death, or January 1, 2020, whichever occurred first. Follow-up was virtually complete (i.e., 98.9% of potential person-years were observed).
The Rotterdam Study has been approved by the Medical Ethics Committee of the Erasmus MC (registration number MEC 02.1015) and by the Dutch Ministry of Health, Welfare, and Sport (Population Screening Act WBO, license number 1071272-159521-PG) and has been entered into the Netherlands National Trial Register (NTR; www.trialregister.nl) and the WHO International Clinical Trials Registry Platform (ICTRP; www.who.int/ictrp/network/primary/en/) under shared catalog number NTR6831. All participants provided written informed consent to participate in the study.
Sleep and 24-h Activity Rhythms
Participants wore an actigraph (Actiwatch model AW4, Cambridge technology Ltd) for a median of 7 days (Q1–Q3: 6–7 days) and filled out a sleep diary during the same time period. Additionally, they pressed a marker button on the actigraph denoting bedtime (“lights out” time) and getting-up time. Participants were instructed to wear the actigraph continuously on their non-dominant wrist and remove it only when performing water-based activities. Sleep and wakefulness during time in bed were determined with a validated algorithm using a threshold of 20 counts [18]. We defined sleep onset as the midpoint of the first immobile period lasting ≥10 min after bedtime with ≤1 epoch of movement. All sleep and 24-h rhythm parameters were winsorized at 4 SD. Four sleep parameters were calculated.
- a
Total sleep time (hours) was calculated as time scored as sleep between sleep onset and sleep end;
- b
Sleep efficiency (%) was defined as the ratio of total sleep time to time between bedtime and get-up time, expressed as a percentage;
- c
Sleep onset latency (minutes) was calculated as the time between bedtime and sleep onset;
- d
Wake after sleep onset (minutes) was calculated as the time scored as awake between sleep onset and sleep end.
Besides, the abovementioned sleep parameters, we in addition calculated three non-parametric indicators of the behavioral functioning of the circadian rhythm, e.g., the 24-h activity rhythm [11].
- e
Interdaily stability quantifies how similar activity profiles are across days and thus how stable the rhythm is. It is calculated as the ratio between the variance of the average 24-h pattern around the mean and the overall variance, with a higher ratio indicating a more stable rhythm [11];
- f
Intradaily variability quantifies the rate of shifting between activity and inactivity and reflects the fragmentation of the rhythm. It is calculated as the ratio of the mean squares of the difference between consecutive hours (first derivative) and overall variance, with a higher ratio indicating a more fragmented rhythm. A favorable 24-h activity rhythm is characterized by low fragmentation and high stability [11];
- g
Onset of least active 5 h is the average time of day when the least active five consecutive hours start, which reflects the phase of the rhythm.
Stroke Ascertainment
Stroke was defined according to the World Health Organization criteria, describing a syndrome of rapidly developing symptoms of focal or global cerebral dysfunction lasting 24 h or longer, or leading to death, with an apparent vascular cause [19, 20]. We assessed prevalent stroke at baseline during interview and subsequently verified these data with medical records. Stroke-free participants who consented to follow-up were then continuously monitored for incident stroke through linkage of the study database with files from general practitioners and nursing homes. Additional medical information (e.g., clinical notes and neuro-imaging reports) was obtained from hospital records, if available. Whenever a potential incident stroke occurred, the records pertaining to the stroke were reviewed by research physicians and subsequently verified by an experienced vascular neurologist. Strokes were classified as ischemic or hemorrhagic based on information from imaging records; the stroke was classified as undetermined if no imaging records were available. Subarachnoid hemorrhages were not considered stroke events.
Covariates
Based on the principles of confounder selection [21], the following variables were included as confounders: age, sex, education, paid employment, smoking, body mass index, prevalent diabetes, and possible sleep apnea. All variables were assessed during a home interview at baseline, except for body mass index which was measured at the research center at baseline. Education was categorized as primary, lower, intermediate, and higher education. Smoking was divided into current, past, or never. Paid employment was based on self-report. Body mass index was calculated using height and weight (kg/m2), measured on calibrated scales without shoes and heavy clothing. Diagnosis of diabetes mellitus was obtained by interview and continuous inspection of medical records. Participants were categorized as having possible sleep apnea when they reported breathing pauses at least 1–2 nights per week or when they reported both breathing pauses occasionally and snoring for at least 2 nights per week on the Pittsburgh sleep quality index [22, 23].
Statistical Analyses
Missing data occurred in education, paid employment, smoking, body mass index, prevalent diabetes (all <1.0%), and possible sleep apnea (20.0%). The mice algorithm [24] was used to impute missing data, using 25 iterations. Pooled results of 25 datasets are presented. We used Cox proportional hazards models to examine the association of sleep and 24-h activity rhythms with incident stroke. Hazard ratios (HR) and 95% confidence intervals (CI) were calculated. Each sleep and 24-h rhythm parameter was modeled individually and transformed into z-scores, to facilitate direct comparison of effect estimates. We additionally present estimates for unstandardized sleep parameters. We investigated non-linearity in associations for total sleep time and onset of the least active 5 h by modeling a quadratic term and comparing the model with and without the quadratic term with a log-likehood ratio test. All analyses were adjusted for age and sex (Model 1). Additionally, associations were adjusted for education, paid employment, smoking, body mass index, prevalent diabetes, and possible sleep apnea (Model 2). As a secondary analysis, we repeated all analyses using only the most common subtype of stroke, ischemic stroke, as outcome. All analyses were performed using R version 4.3.0 [25].
Results
Population Characteristics
Baseline characteristics of our study population of 1,718 participants are described in Table 1. Mean age at baseline was 62.2 years (SD: 9.3), and 55.1% were female. During an average follow-up of 12.2 years (SD: 3.17), 105 (6.1%) persons developed a stroke. Of these, 81 (4.7%) were classified as ischemic.
. | |
---|---|
mean (SD) or N (%) . | |
Age at baseline, mean (SD), years | 62.2 (9.31) |
Follow-up time, mean (SD), years | 12.2 (3.17) |
Female, n (%) | 946 (55.1) |
Education, n (%) | |
Primary | 141 (8.2) |
Low | 698 (40.6) |
Intermediate | 511 (29.7) |
High | 355 (20.7) |
Paid employment, n (%) | 593 (34.5) |
Smoking, n (%) | |
Current | 364 (21.2) |
Former | 848 (49.4) |
Never | 492 (28.6) |
Body mass index, mean (SD), kg/m2 | 27.9 (4.25) |
Prevalent diabetes, n (%) | 218 (12.7) |
Possible sleep apnea, n (%) | 167 (9.7) |
Duration actigraphy measurement, days | 7 [6–7]a |
Sleep duration, mean (SD), h | 6.1 (0.9) |
Sleep onset latency, mean (SD), min | 21.1 (16.4) |
Wake after sleep onset, mean (SD), min | 62.8 (25.9) |
Sleep efficiency, mean (SD), % | 74.5 (8.6) |
Interdaily stability, mean (SD), score | 0.77 (0.12) |
Intradaily variability, mean (SD), score | 0.43 (0.14) |
Onset of least active 5 h, mean (SD), h:min | 01:18 (1:07) |
Incident stroke, n (%) | 105 (6.1) |
Incident ischemic stroke, n (%) | 81 (5.4) |
. | |
---|---|
mean (SD) or N (%) . | |
Age at baseline, mean (SD), years | 62.2 (9.31) |
Follow-up time, mean (SD), years | 12.2 (3.17) |
Female, n (%) | 946 (55.1) |
Education, n (%) | |
Primary | 141 (8.2) |
Low | 698 (40.6) |
Intermediate | 511 (29.7) |
High | 355 (20.7) |
Paid employment, n (%) | 593 (34.5) |
Smoking, n (%) | |
Current | 364 (21.2) |
Former | 848 (49.4) |
Never | 492 (28.6) |
Body mass index, mean (SD), kg/m2 | 27.9 (4.25) |
Prevalent diabetes, n (%) | 218 (12.7) |
Possible sleep apnea, n (%) | 167 (9.7) |
Duration actigraphy measurement, days | 7 [6–7]a |
Sleep duration, mean (SD), h | 6.1 (0.9) |
Sleep onset latency, mean (SD), min | 21.1 (16.4) |
Wake after sleep onset, mean (SD), min | 62.8 (25.9) |
Sleep efficiency, mean (SD), % | 74.5 (8.6) |
Interdaily stability, mean (SD), score | 0.77 (0.12) |
Intradaily variability, mean (SD), score | 0.43 (0.14) |
Onset of least active 5 h, mean (SD), h:min | 01:18 (1:07) |
Incident stroke, n (%) | 105 (6.1) |
Incident ischemic stroke, n (%) | 81 (5.4) |
Data are presented as mean (SD) or N (%) unless otherwise indicated, and shown for non-imputed data. Missing data were <1.0% in all covariates except for possible sleep apnea (20.0% missing data).
aMedian [Q1–Q3].
Sleep
No actigraphy-based sleep parameters were significantly associated with the risk of stroke (Table 2). The largest effect size was observed for sleep efficiency (hazard ratio [HR] per standard deviation increase 1.22; 95% confidence interval [CI] 0.98–1.52). Estimates adjusted for age and sex (model 1) were not meaningfully different. We found no evidence for non-linearity after fitting quadratic terms for the association of sleep duration with the risk of stroke (p = 0.15). Effect estimates for sleep and the risk of ischemic stroke were similar (Table 3).
. | Incident stroke . | ||||
---|---|---|---|---|---|
105 cases/21,145 person-years . | |||||
standardized . | non-standardized . | p value . | |||
HR . | 95% CI . | HR . | 95% CI . | ||
Sleep duration (h) | 1.16 | 0.94; 1.43 | 1.18 | 0.94; 1.49 | 0.16 |
Sleep onset latency (min) | 0.87 | 0.70; 1.09 | 0.99 | 0.98; 1.01 | 0.23 |
Wake after sleep onset (min) | 0.97 | 0.79; 1.19 | 1.00 | 0.99; 1.01 | 0.76 |
Sleep efficiency (%) | 1.22 | 0.98; 1.52 | 1.02 | 1.00; 1.05 | 0.08 |
Interdaily stability | 0.83 | 0.68; 1.01 | - | - | 0.059 |
Intradaily variability | 1.28 | 1.07; 1.54 | - | - | 0.008 |
Onset of the least active 5 h | 0.99 | 0.82; 1.20 | 0.99 | 0.83; 1.18 | 0.93 |
. | Incident stroke . | ||||
---|---|---|---|---|---|
105 cases/21,145 person-years . | |||||
standardized . | non-standardized . | p value . | |||
HR . | 95% CI . | HR . | 95% CI . | ||
Sleep duration (h) | 1.16 | 0.94; 1.43 | 1.18 | 0.94; 1.49 | 0.16 |
Sleep onset latency (min) | 0.87 | 0.70; 1.09 | 0.99 | 0.98; 1.01 | 0.23 |
Wake after sleep onset (min) | 0.97 | 0.79; 1.19 | 1.00 | 0.99; 1.01 | 0.76 |
Sleep efficiency (%) | 1.22 | 0.98; 1.52 | 1.02 | 1.00; 1.05 | 0.08 |
Interdaily stability | 0.83 | 0.68; 1.01 | - | - | 0.059 |
Intradaily variability | 1.28 | 1.07; 1.54 | - | - | 0.008 |
Onset of the least active 5 h | 0.99 | 0.82; 1.20 | 0.99 | 0.83; 1.18 | 0.93 |
HR, hazard ratio. Statistically significant (p < 0.05) results are shown in bold. Standardized effect estimates reflect the hazard ratio per SD increase in the sleep or 24-h activity rhythm parameter. Non-standardized effect estimates reflect the hazard ratio per unit increase in the sleep or 24-h activity rhythm parameter. All estimates are adjusted for age, sex, education, paid employment, smoking, body mass index, prevalent diabetes, and possible sleep apnea.
. | Incident ischemic stroke . | ||||
---|---|---|---|---|---|
81 cases/21,145 person-years . | |||||
standardized . | non-standardized . | p value . | |||
HR . | 95% CI . | HR . | 95% CI . | ||
Sleep duration (h) | 1.14 | 0.90; 1.45 | 1.16 | 0.89; 1.51 | 0.27 |
Sleep onset latency (min) | 0.88 | 0.68; 1.13 | 0.99 | 0.98; 1.01 | 0.31 |
Wake after sleep onset (min) | 1.04 | 0.83; 1.30 | 1.00 | 0.99; 1.01 | 0.74 |
Sleep efficiency (%) | 1.24 | 0.96; 1.59 | 1.02 | 1.00; 1.06 | 0.10 |
Interdaily stability | 0.79 | 0.64; 0.98 | - | - | 0.03 |
Intradaily variability | 1.27 | 1.03; 1.57 | - | - | 0.024 |
Onset of the least active 5 h | 1.01 | 0.81; 1.27 | 1.01 | 0.83; 1.23 | 0.91 |
. | Incident ischemic stroke . | ||||
---|---|---|---|---|---|
81 cases/21,145 person-years . | |||||
standardized . | non-standardized . | p value . | |||
HR . | 95% CI . | HR . | 95% CI . | ||
Sleep duration (h) | 1.14 | 0.90; 1.45 | 1.16 | 0.89; 1.51 | 0.27 |
Sleep onset latency (min) | 0.88 | 0.68; 1.13 | 0.99 | 0.98; 1.01 | 0.31 |
Wake after sleep onset (min) | 1.04 | 0.83; 1.30 | 1.00 | 0.99; 1.01 | 0.74 |
Sleep efficiency (%) | 1.24 | 0.96; 1.59 | 1.02 | 1.00; 1.06 | 0.10 |
Interdaily stability | 0.79 | 0.64; 0.98 | - | - | 0.03 |
Intradaily variability | 1.27 | 1.03; 1.57 | - | - | 0.024 |
Onset of the least active 5 h | 1.01 | 0.81; 1.27 | 1.01 | 0.83; 1.23 | 0.91 |
HR, hazard ratio. Statistically significant results are shown in bold. Standardized effect estimates reflect the hazard ratio per SD increase in the sleep or 24-h activity rhythm parameter. Non-standardized effect estimates reflect the hazard ratio per unit increase in the sleep or 24-h activity rhythm parameter. All estimates are adjusted for age, sex, education, paid employment, smoking, body mass index, prevalent diabetes, and possible sleep apnea.
24-h Activity Rhythms
A lower interdaily stability was significantly associated with a higher risk of stroke when adjusted for age and sex (HR 0.81 per SD increase in interdaily stability; 95% CI: 0.67–0.99) (Table 2), but this association was no longer significant when adjusting for the covariates in model 2. A higher intradaily variability was associated with a larger risk of stroke (HR 1.28, 95% CI: 1.07–1.54]. Further, lower interdaily stability (HR 0.79, 95% CI: 0.64–0.98) and higher intradaily variability (HR 1.27, 95% CI: 1.03–1.57) were associated with a higher risk of ischemic stroke (Table 3). Onset of the least active 5 h was not associated with the risk of stroke and ischemic stroke, and there was no evidence for a non-linear relationship (p = 0.73).
Discussion
In this population-based sample of middle-aged and elderly adults, we observed that objectively measured sleep was not associated with the risk of stroke and its subtypes. In contrast, for the 24-h activity rhythms, we found that an unstable and fragmented rhythm was associated with a higher risk of stroke, in particular ischemic stroke.
A 24-h activity rhythm with low stability and high variability is characterized by changes in day-night patterns throughout the week and more transitions between longer periods of rest and activity. Our observation that those with unstable and fragmented rhythms at baseline have a higher risk of stroke and ischemic stroke extends previous literature that linked 24-h activity rhythm disturbances to a higher risk of cardiovascular disease [14, 26]. Moreover, cross-sectional studies have shown that people with 24-h activity rhythm disturbances more often have atherosclerosis and small vessel disease [6, 8]. On the one hand, disturbances in 24-h activity rhythms have been hypothesized to contribute to development of cerebrovascular pathology. Both experimental and observational studies have shown that fragmented sleep and circadian misalignment are associated with increased sympathetic activity, systemic inflammation, hypothalamic-pituitary-axis dysregulation, and increased blood pressure [27‒29], which may all predispose to stroke. On the other hand, structural brain changes could lead to both disturbed rhythms and stroke. For example, small vessel disease might decrease the volume and cell number of the suprachiasmatic nucleus, the master clock of the circadian rhythm, reducing the ability of the brain to maintain a stable and robust rhythm [6]. Although the underlying mechanism remains unclear, our results indicate that rhythm disturbances signal a higher risk of (ischemic) stroke.
No sleep parameters were found to be significantly associated with the risk of stroke. In a previous study that utilized one-night polysomnography to measure sleep, lower sleep efficiency, a longer wake after sleep onset, and shorter sleep duration were associated with stroke [9]. In contrast, we observe a non-significant association of higher sleep efficiency with the risk of stroke, although with a relatively large effect size. A higher sleep efficiency is often thought to reflect good sleep health, but it can also result from a short sleep opportunity with accompanied short sleep shortage. However, the latter is unlikely to explain our association, as we do not observe associations of short sleep with stroke. A high sleep efficiency could also reflect a heightened sleep propensity, i.e., a heightened sleep need, which has been linked to a higher risk of mortality [30]. In the literature, the j-shaped link between short and long sleep self-reported sleep duration and the risk of stroke has been described extensively [4]. However, a recent Mendelian randomization study did not find evidence of an association of genetic liability to short or long sleep duration with the risk of stroke [31], which is in line with our findings. The heterogeneity in the evidence might stem from differences between self-reported and objectively estimated sleep. Self-reported sleep reflects the experience rather than the physiological state of sleep, and several factors, including sex, age, poor cognition, and depressive symptoms, are associated with discrepancies between self-report and actigraphy-based sleep [32]. Moreover, it has been suggested that parameters of sleep should not be studied in isolation, as sleep is a multidimensional construct and specific combinations of symptoms, such as short sleep with insomnia symptoms, might confer the highest risk for cardiovascular disease [5, 33]. Further research, utilizing multidimensional approaches to sleep health, is needed to further disentangle the complex associations between sleep and stroke.
Some methodological aspects of our population-based cohort study limit our ability to differentiate whether disturbed 24-h activities are a cause or a consequence of cerebrovascular disease. First, we did not adjust for cardiovascular risk factors (including hypertension and physical inactivity) or psychosocial factors (such as depressive symptoms or stress) as these can both confound and mediate the link between 24-h rhythms and stroke. Second, our results may in part be explained by possible reversed causation by underlying cerebrovascular disease. Third, although actigraphy can validly estimate sleep it is not the gold standard for determining sleep and cannot provide in-depth insights in sleep with regards to, for example, sleep stages or power spectra. Actigraphy-measured 24-h activity rhythms need to be studied together with biological markers of circadian rhythms to more robustly measure circadian rhythm disruptions [34]. Moreover, future studies are needed to assess whether our findings extend to other populations beyond middle-aged and elderly adults in the Netherlands. Lastly, possible sleep apnea was assessed through self-report and had a high proportion of missing data. Although we carefully imputed missing data, our results are possibly confounded by sleep apnea unknown to the participant, which could be mitigated by confirming the diagnosis with polysomnography. Nevertheless, our robust associations suggest that behavioral measures of rest-activity rhythm could be of predictive and prognostic value for stroke and could be considered for use in risk stratification and personalized prevention strategies. Future studies, including intervention trials, could investigate whether improving stability and reducing fragmentation of the rhythm can reduce the risk of stroke.
Conclusion
In summary, our study shows that unstable and fragmented 24-h activity rhythms are associated with a higher risk of stroke and ischemic stroke in the middle-aged and elderly and might play a more important role than sleep only. Future studies are needed to establish whether disturbed activity rhythms might add to predictive and prognostic models for cerebrovascular disease.
Acknowledgments
The authors are grateful to the study participants, the staff from the Rotterdam Study, and the participating general practitioners and pharmacists.
Statement of Ethics
The Rotterdam Study has been approved by the Medical Ethics Committee of the Erasmus MC (registration number MEC 02.1015) and by the Dutch Ministry of Health, Welfare, and Sport (Population Screening Act WBO, license number 1071272-159521-PG) and has been entered into the Netherlands National Trial Register (NTR; http://www.trialregister.nl) and the WHO International Clinical Trials Registry Platform (ICTRP; http://www.who.int/ictrp/network/primary/en/) under shared catalog number NTR6831. All participants provided written informed consent to participate in the study.
Conflict of Interest Statement
The authors have no conflicts of interest to declare.
Funding Sources
The Rotterdam Study is funded by Erasmus Medical Center and Erasmus University, Rotterdam, the Netherlands Organization for Health Research and Development (ZonMw), the Research Institute for Diseases in the Elderly (RIDE), the Ministry of Education, Culture and Science, the Ministry for Health, Welfare and Sports, the European Commission (DG XII), and the Municipality of Rotterdam. The funder had no role in the design, data collection, data analysis, and reporting of this study.
Author Contributions
Study design and conception: S.J.W.H., A.I.L., and M.K.I. Data collection: S.J.W.H. and B.P.B. Analysis and primary interpretation of results and draft manuscript preparation: S.J.W.H. All authors reviewed the results and approved the final version of the manuscript.
Data Availability Statement
Data can be obtained upon request. Requests should be directed towards the management team of the Rotterdam Study (datamanagement.ergo@erasmusmc.nl), which has a protocol for approving data requests. Because of restrictions based on privacy regulations and informed consent of the participants, data cannot be made freely available in a public repository.