Association of sleep duration at age 50, 60, and 70 years with risk of multimorbidity in the UK: 25-year follow-up of the Whitehall II cohort study.

BACKGROUND: Sleep duration has been shown to be associated with individual chronic diseases but its association with multimorbidity, common in older adults, remains poorly understood. We examined whether sleep duration is associated with incidence of a first chronic disease, subsequent multimorbidity and mortality using data spanning 25 years. METHODS AND
FINDINGS: Data were drawn from the prospective Whitehall II cohort study, established in 1985 on 10,308 persons employed in the London offices of the British civil service. Self-reported sleep duration was measured 6 times between 1985 and 2016, and data on sleep duration was extracted at age 50 (mean age (standard deviation) = 50.6 (2.6)), 60 (60.3 (2.2)), and 70 (69.2 (1.9)). Incidence of multimorbidity was defined as having 2 or more of 13 chronic diseases, follow-up up to March 2019. Cox regression, separate analyses at each age, was used to examine associations of sleep duration at age 50, 60, and 70 with incident multimorbidity. Multistate models were used to examine the association of sleep duration at age 50 with onset of a first chronic disease, progression to incident multimorbidity, and death. Analyses were adjusted for sociodemographic, behavioral, and health-related factors. A total of 7,864 (32.5% women) participants free of multimorbidity had data on sleep duration at age 50; 544 (6.9%) reported sleeping ≤5 hours, 2,562 (32.6%) 6 hours, 3,589 (45.6%) 7 hours, 1,092 (13.9%) 8 hours, and 77 (1.0%) ≥9 hours. Compared to 7-hour sleep, sleep duration ≤5 hours was associated with higher multimorbidity risk (hazard ratio: 1.30, 95% confidence interval = 1.12 to 1.50; p < 0.001). This was also the case for short sleep duration at age 60 (1.32, 1.13 to 1.55; p < 0.001) and 70 (1.40, 1.16 to 1.68; p < 0.001). Sleep duration ≥9 hours at age 60 (1.54, 1.15 to 2.06; p = 0.003) and 70 (1.51, 1.10 to 2.08; p = 0.01) but not 50 (1.39, 0.98 to 1.96; p = 0.07) was associated with incident multimorbidity. Among 7,217 participants free of chronic disease at age 50 (mean follow-up = 25.2 years), 4,446 developed a first chronic disease, 2,297 progressed to multimorbidity, and 787 subsequently died. Compared to 7-hour sleep, sleeping ≤5 hours at age 50 was associated with an increased risk of a first chronic disease (1.20, 1.06 to 1.35; p = 0.003) and, among those who developed a first disease, with subsequent multimorbidity (1.21, 1.03 to 1.42; p = 0.02). Sleep duration ≥9 hours was not associated with these transitions. No association was found between sleep duration and mortality among those with existing chronic diseases. The study limitations include the small number of cases in the long sleep category, not allowing conclusions to be drawn for this category, the self-reported nature of sleep data, the potential for reverse causality that could arise from undiagnosed conditions at sleep measures, and the small proportion of non-white participants, limiting generalization of findings.
CONCLUSIONS: In this study, we observed short sleep duration to be associated with risk of chronic disease and subsequent multimorbidity but not with progression to death. There was no robust evidence of an increased risk of chronic disease among those with long sleep duration at age 50. Our findings suggest an association between short sleep duration and multimorbidity.

Introduction

Approximately one third of human life is devoted to sleep, emphasizing the vital role of sleep in several physiological functions essential for health. There is also consistent evidence of an association of sleep duration with chronic diseases, such as cardiovascular disease (CVD) and cancer [1,2], and with mortality [2-4], although there remain a number of outstanding questions regarding the nature of this association. First, multiple chronic conditions often coexist within the same individual, a condition known as multimorbidity [5-8], but the association of sleep duration with multimorbidity remains poorly understood due to paucity of research and cross-sectional nature of existing studies [9-13]. It is unclear how sleep duration affects trajectories from a healthy state, to 1 or more chronic diseases, and subsequent mortality. Second, current guidelines recommend 7 to 8 hours of sleep for older adults [14] but whether both short and long sleep duration carry risk for multimorbidity remains unclear. Several biological mechanisms have been proposed to explain the role of short sleep duration in disease onset [15,16] but the role of long sleep is less well understood [3,17]. The observed risk of chronic conditions among long sleepers could be due to preexisting health conditions [15,18] or, alternatively, reflect non-restorative sleep that then affects risk of subsequent disease [15,19]. Third, as people get older, their sleep habits and sleep structure change [20]; whether sleep duration in mid and later life differentially affects subsequent risk of multimorbidity has not been investigated.

The first objective of the present study was to examine the association between sleep duration at 50, 60, and 70 years of age and incident multimorbidity, using repeat data on sleep duration and continuous assessment of chronic diseases spanning over 25 years. A second objective was to determine whether sleep duration at age 50 shapes the natural course of chronic disease, from a healthy state, to a first chronic disease, multimorbidity, and death using multistate models to examine the association of sleep duration at age 50 with transitions between each of these health states. In these analyses, the focus is on sleep duration at age 50 as chronic conditions are less prevalent and reverse causation bias that could arise from underlying conditions affecting sleep duration is less likely. In additional analyses, we examined the association of sleep duration with the onset of multimorbidity and death in the subgroup of participants with 1 chronic condition to examine whether sleep pattern after onset of chronic conditions is associated with adverse health outcomes [15]. Finally, in post hoc analysis, the association between sleep disturbance at age 60 and 70, where we had data on these measures, and risk of incident multimorbidity was examined.

Methods

This study is reported following the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S1 STROBE Checklist).

Study population

The Whitehall II study is an ongoing cohort study established in 1985 among 10,308 British civil servants (6,895 men and 3,413 women, aged 35 to 55 years) [21]. Since baseline, follow-up clinical examinations have taken place approximately every 4 to 5 years, each wave taking 2 years to complete, with the last completed wave conducted in 2015 to 2016. Except for 10 individuals, all participants (99.9%) are linked to UK National Health Service (NHS) electronic health records. The NHS provides most of the health care in the United Kingdom, including in- and out-patient care, and record linkage is undertaken using a unique NHS identifier held by all UK residents. Data from linked records were updated on an annual basis, until March 31, 2019. Written informed consent from participants and research ethics approvals were renewed at each contact; the most recent approval was from the University College London Hospital Committee on the Ethics of Human Research, reference number 85/0938.

Sleep duration

Sleep duration was measured at 6 data collection waves, 1985 to 1988, 1997 to 1999, 2002 to 2004, 2007 to 2009, 2012 to 2013, and 2015 to 2016 using the question “How many hours of sleep do you have on an average week-night?” Response categories were: ≤5 hours, 6 hours, 7 hours, 8 hours, and ≥9 hours. For each participant, sleep duration at age 50, 60, and 70 was extracted across the data waves using data from the wave at which the participant’s age was the closest to the target age, allowing a ±5 year margin for each age of interest.

Trajectories of change in sleep duration between age 50 and 70 [22] among those with at least 2 out of 3 measures of sleep duration at age 50, 60, and 70 were defined using group-based trajectory modeling using the traj-command in Stata [23]. Groups were chosen using the best model fit (Bayesian Information Criterion values and average posterior probabilities) and meaningful interpretation of trajectories [24].

At the 2012 wave, when participants were 60 to 83 years, an accelerometer sub-study—a one-off addition to the main data collection—was undertaken on participants who attended the central London research clinic or were assessed at home if they resided in the South-Eastern regions of England. Wrist-worn accelerometers, the GENEActiv (Activinsights, Kimbolton, UK), were worn 24 hours over 9 consecutive days [25]. Sleep duration was estimated using a validated algorithm guided by a sleep log [26]; data from the first and last nights were removed leading to data over 7 nights. Usual daily sleep duration was estimated as the mean of sleep duration over 7 nights and for those with less than 7 nights of measurement, weighted average of sleep duration was calculated as: 5 × week night sleep duration + 2 × weekend night sleep duration)/7.

In post hoc analysis, we used data on sleep quality measured using the Jenkins sleep problems scale [27]. This measure was introduced to the study questionnaire in 1997 and repeated at following study waves, allowing us to extract data on sleep quality at age 60 and 70, but not age 50. Participants were asked how often in the past month they had experienced the following symptoms: (1) trouble falling asleep; (2) waking up several times per night; (3) trouble staying asleep (including waking far too early); and (4) disturbed or restless sleep. The following response categories were available: Not at all (scored 0), 1 to 3 days (scored 1), 4 to 7 days (scored 2), 8 to 14 days (scored 3), 15 to 21 days (scored 4), and 22 to 31 days (scored 5). The sum of these items was then used as a continuous scale to measure sleep problems. The score was further dichotomized to reflect low sleep disturbance (0 to 11) and high sleep disturbance (12 to 20) [28].

Multimorbidity

Multimorbidity was defined as the presence of 2 or more chronic diseases out of a predefined list of 13 chronic diseases that were chosen because they are prevalent across the adult lifecourse. Inclusion of at least 12 conditions is thought to accurately reflect multimorbidity [29] and our list was chosen from previous research on multimorbidity [8,30]. As in previous studies, risk factors such as hypertension and obesity were not included in the list [31,32]. We identified chronic diseases using data from the Whitehall clinical examinations and via linkage to electronic health records up to March 31, 2019 from the Hospital Episode Statistics (HES) database, the Mental Health Services Data Set (which in addition to in- and out-patient data also include records on care in the community), and the national cancer registry. The chronic diseases considered [33] were:

diabetes (ICD10: E10-E14, reported doctor-diagnosed diabetes, use of diabetes medication, or fasting glucose ≥ 7.0 mmol/l),

cancer (malignant neoplasms ICD10: C00-C97),

coronary heart disease (ICD10: I20-I25, 12-lead resting ECG recording) [34],

stroke (ICD10: I60-I64, MONICA-Ausburg stroke questionnaire) [34],

heart failure (ICD10: I50),

chronic obstructive pulmonary disease (ICD10: J41-J44),

chronic kidney disease (ICD10: N18),

liver disease (ICD10: K70-K74),

depression (ICD10: F32, F33, or use of antidepressants),

dementia (ICD10: F00-F03, F05.1, G30, G31) [35],

mental disorders, other than depression and dementia (ICD10: F06, F07, F09, F20-48 (excluding F32: depressive episode and F33: major depressive disorder, recurrent) and F60-69 (excluding F65: paraphilias and F66: other sexual disorders)) [36],

Parkinson’s disease (ICD10: G20), and

arthritis/rheumatoid arthritis (ICD10: M15-M19, M05, M06).

Mortality

Mortality was ascertained from linked records from the British national mortality register (National Health Services Central Registry) with follow-up until March 31, 2019.

Covariates

Covariates included sociodemographic, behavioral, and health-related factors. Sex, ethnicity, and education were drawn from the baseline examination in 1985 to 1988. Other covariates, available at each wave of data collection, were extracted concurrently to the measure of sleep duration at age 50, 60, and 70.

Sociodemographic factors included age, sex, ethnicity (response to a question using the categories “White,” “South Asian,” “Black,” and “Other” and categorized in the analysis as White and non-White, due to the small numbers in the latter group), education (primary school or less, lower secondary school, higher secondary school, university, higher degree; treated as continuous variable), occupational position (high, intermediate, and low, representing income and status at work), and marital status (married or cohabiting, other).

Health behaviors included cigarette smoking (never smoker, ex-smoker, current smoker), alcohol consumption in the previous week (none, 1 to 14 units per week, >14 units per week), time spent in moderate and vigorous physical activity (hours per week), and frequency of fruit and vegetable consumption (less than daily, once a day, twice or more a day).

Health-related factors included hypertension (systolic ≥140 or diastolic ≥90 mmHg or use of antihypertensive medication), body mass index (BMI, categorized as <18.5, 18.5 to 24.9, 25 to 29.9, and ≥30 kg/m2) calculated using height and weight measured at the clinical examination using standard clinical protocols, use of sleep medication, and prevalence of 1 of the 13 conditions considered in the definition of multimorbidity.

Statistical analysis

The analysis plan was developed prior to data analysis (S1 Text). The analyses referred to as post hoc analyses were in response to suggestions from reviewers.

Association between sleep duration at different ages and incident multimorbidity

The association of sleep duration at age 50, 60, and 70 with incident multimorbidity was examined in separated models. The analyses were undertaken using Cox proportional-hazards regression with age as the timescale in participants free from multimorbidity at the measurement of sleep duration. Data were censored at date of multimorbidity diagnosis, death to account for competing risk [37], or March 31, 2019, whichever came first. In analysis of sleep duration at age 50, age at the beginning of the follow-up was the age at clinical assessment closest to 50 years from which the sleep duration measure and covariates were drawn. A similar approach was used in analyses of sleep duration at age 60 and 70. The proportional hazards assumption was verified using Schoenfeld residuals. Analyses were first unadjusted (age as timescale; Model 1), then adjusted for sociodemographic measures (Model 2), and finally additionally for behavioral and health-related factors (Model 3).

To examine the robustness of our findings, we repeated the main analysis (1) in participants free from any of the 13 chronic diseases used to define multimorbidity, (2) excluding users of sleep medication and (3) examined the association between accelerometer-assessed sleep duration at mean age 69 (range = 60 to 83) years and incident multimorbidity. The covariates were drawn from the 2012 wave of data collection, concurrent to the accelerometer measure. Given the detailed data on sleep duration extracted from the accelerometer, we used restricted cubic spline regressions with Harrell knots [38], Stata command partpred [39], with 7-hour sleep as the reference to assess the shape of the association between sleep duration and multimorbidity risk.

Several post hoc analyses were conducted. First, the main analyses were repeated using inverse probability weighting to account for missing data. Second, we explored whether findings were driven by one specific chronic disease by repeating the analysis on sleep duration at age 50, 60, and 70 and incident multimorbidity, excluding one chronic disease at a time from the definition of multimorbidity. Third, we examined the association of trajectories of sleep duration between age 50 and 70 with incident multimorbidity with age of entry and covariates drawn from the wave sleep measure at age 70 was extracted. Fourth, the association of sleep disturbance at age 60 and age 70 with incident multimorbidity was investigated.

Association of sleep duration with transitions to multimorbidity and death

Among participants at age 50, free from the 13 chronic diseases considered here, multistate models were used (Fig 1) to determine the association of sleep duration at age 50 with transitions from: (1) a healthy state to a first chronic disease (any from the list of 13 diseases considered); (2) a healthy state to death (in those who remained free from any of the 13 diseases during follow-up); (3) a first chronic disease to multimorbidity; (4) a first chronic disease to death; and (5) multimorbidity to death, with follow-up starting at age 50. The advantage of multistate models is that they take into account the time spent within each health state to estimate probabilities of transitions between each state. For comparison, we also examined (post hoc analysis) the association between sleep duration at age 50 and risk of mortality in the same study sample (participants free from chronic disease at age 50) irrespective of incidence of chronic disease over the follow-up.

Fig 1

Schematic representation of the transitions from start of follow-up (free of chronic disease at age 50) to a first chronic disease, multimorbidity, and mortality.

Transition A represents the transition from a healthy state at age 50 (free of the 13 chronic diseases considered) to a first chronic disease (any from the list of 13 diseases considered); Transition B represents the transition from a healthy state to death among those who remained free from any of the 13 diseases during follow-up; Transition C represents the transition from a first chronic disease to multimorbidity (occurrence of a second disease among those with 1 chronic disease); Transition D represents the transition from a first chronic disease to death among those who remained free from multimorbidity during the follow-up; and Transition E represents the transition from multimorbidity to death.

In additional analyses, we examined the association of sleep duration after onset of a first chronic disease with transitions to multimorbidity and death, again using a multistate model. The follow-up here started at the measure of first sleep duration following the onset of a first chronic disease. In sensitivity analysis, we used inverse probability weighting to account for missing data [22].

Results

Sleep duration at ages 50, 60, and 70 and subsequent risk of multimorbidity

Among the 10,308 participants of the Whitehall cohort, 7,864 (32.5% women) participants free of multimorbidity had data on sleep duration and covariates at age 50 (mean (standard deviation (SD)) = 50.6 (2.6) years). Among them, 2,659 (33.8%) developed multimorbidity at mean age 70.9 (SD = 7.7) years over a mean follow-up of 22.6 (SD = 7.5) years. Among the 6,848 participants with data on sleep duration and covariates at age 60 (mean (SD) = 60.3 (2.2) years) and free of multimorbidity, 2,029 (29.6%) developed multimorbidity at mean age of 72.0 (SD = 6.3) years over a mean follow-up of 13.4 (SD = 6.0) years. Among 5,546 participants free of multimorbidity and with data on sleep duration and covariates at age 70 (mean (SD) = 69.2 (1.9) years), 1,402 (25.3%) subsequently developed multimorbidity at a mean age of 76.0 (SD = 4.8) years over a mean follow-up of 6.8 (SD = 4.5) years. Flowchart of sample selection is shown in Fig 2. Characteristics of participants at age 50 are presented in Table 1 and at age 60 and 70 in S1 and S2 Tables, respectively. At age 50, 544 (6.9% of the study population, N = 7,864) reported sleeping ≤5 hours, 2,562 (32.6%) 6 hours, 3,589 (45.6%) 7 hours, 1,092 (13.9%) 8 hours, and 77 (1.0%) ≥9 hours.

Fig 2

Flowchart for analyses on the association between sleep duration at age 50, 60, and 70 and risk of multimorbidity.

Table 1

Characteristics of the study population at age 50.

		Sleep duration at age 50
	Total	≤5 hours	6 hours	7 hours	8 hours	≥9 hours	P
N	7,864	544	2,562	3,589	1,092	77
Sex							<0.001
Men	5,305 (67.5)	319 (58.6)	1,773 (69.2)	2,491 (69.4)	683 (62.5)	39 (50.6)
Women	2,559 (32.5)	225 (41.4)	789 (30.8)	1,098 (30.6)	409 (37.5)	38 (49.4)
Ethnicity							<0.001
White	7,088 (90.1)	464 (85.3)	2,326 (90.8)	3,286 (91.6)	956 (87.5)	56 (72.7)
Non-white	776 (9.9)	80 (14.7)	236 (9.2)	303 (8.4)	136 (12.5)	21 (27.3)
Education							0.017
Primary school or less	887 (11.3)	81 (14.9)	288 (11.2)	391 (10.9)	121 (11.1)	6 (7.8)
Lower secondary school	2,875 (36.6)	217 (39.9)	909 (35.5)	1,312 (36.6)	412 (37.7)	25 (32.5)
Higher secondary school	1,982 (25.2)	135 (24.8)	681 (26.6)	878 (24.5)	262 (24.0)	26 (33.8)
University	1,598 (20.3)	85 (15.6)	521 (20.3)	764 (21.3)	212 (19.4)	16 (20.8)
Higher degree	522 (6.6)	26 (4.8)	163 (6.4)	244 (6.8)	85 (7.8)	4 (5.2)
Occupational position							<0.001
Low	1,572 (20.0)	153 (28.1)	480 (18.7)	658 (18.3)	258 (23.6)	23 (29.9)
Intermediate	3,373 (42.9)	253 (46.5)	1,135 (44.3)	1,517 (42.3)	432 (39.6)	36 (46.8)
High	2,919 (37.1)	138 (25.4)	947 (37.0)	1,414 (39.4)	402 (36.8)	18 (23.4)
Marital status							<0.001
Married/cohabiting	5,951 (75.7)	344 (63.2)	1,925 (75.1)	2,797 (77.9)	824 (75.5)	61 (79.2)
Single/divorced/widowed	1,913 (24.3)	200 (36.8)	637 (24.9)	792 (22.1)	268 (24.5)	16 (20.8)
Smoking status							0.012
Never smoker	3,885 (49.4)	247 (45.4)	1,272 (49.6)	1,745 (48.6)	574 (52.6)	47 (61.0)
Ex-smoker	2,763 (35.1)	192 (35.3)	883 (34.5)	1,303 (36.3)	361 (33.1)	24 (31.2)
Current smoker	1,216 (15.5)	105 (19.3)	407 (15.9)	541 (15.1)	157 (14.4)	6 (7.8)
Alcohol consumption							<0.001
0 unit/week	1,399 (17.8)	128 (23.5)	419 (16.4)	590 (16.4)	238 (21.8)	24 (31.2)
1–14 units/week	4,310 (54.8)	271 (49.8)	1,407 (54.9)	2,024 (56.4)	572 (52.4)	36 (46.8)
>14 units/week	2,155 (27.4)	145 (26.7)	736 (28.7)	975 (27.2)	282 (25.8)	17 (22.1)
Fruit and vegetable consumption							0.010
Less than once a day	2,857 (36.3)	233 (42.8)	950 (37.1)	1,271 (35.4)	377 (34.5)	26 (33.8)
Once a day	3,025 (38.5)	199 (36.6)	940 (36.7)	1,412 (39.3)	446 (40.8)	28 (36.4)
Twice or more a day	1,982 (25.2)	112 (20.6)	672 (26.2)	906 (25.2)	269 (24.6)	23 (29.9)
Moderate-to-vigorous physical activity (hours), M(SD)	3.3 (3.7)	2.9 (4.0)	3.3 (3.9)	3.4 (3.5)	3.3 (3.6)	2.4 (3.1)	0.009
BMI (kg/m2), M(SD)	25.5 (3.8)	26.3 (4.8)	25.8 (3.9)	25.3 (3.7)	25.1 (3.6)	25.8 (3.5)	<0.001
<18.5 kg/m2	78 (1.0)	8 (1.5)	18 (0.7)	35 (1.0)	16 (1.5)	1 (1.3)	<0.001
18.5–24.9 kg/m2	3,906 (49.7)	232 (42.6)	1,185 (46.3)	1,882 (52.4)	575 (52.7)	32 (41.6)
25–29.9 kg/m2	3,006 (38.2)	201 (36.9)	1,054 (41.1)	1,323 (36.9)	394 (36.1)	34 (44.2)
≥30 kg/m2	874 (11.1)	103 (18.9)	305 (11.9)	349 (9.7)	107 (9.8)	10 (13.0)
Hypertension	1,816 (23.1)	150 (27.6)	613 (23.9)	774 (21.6)	259 (23.7)	20 (26.0)	0.014
Use of sleep medication	107 (1.4)	20 (3.7)	47 (1.8)	33 (0.9)	5 (0.5)	2 (2.6)	<0.001
Prevalence of one chronic diseasea at age 50	647 (8.2)	70 (12.9)	212 (8.3)	266 (7.4)	84 (7.7)	15 (19.5)	<0.001

a Chronic disease among diabetes, cancer, coronary heart disease, stroke, heart failure, chronic obstructive pulmonary disease, chronic kidney disease, liver disease, depression, dementia, other mental disorder, Parkinson’s disease, and arthritis/rheumatoid arthritis.

Values are No. (%) unless stated otherwise.

BMI, body mass index; M, mean; SD, standard deviation.

By design, mean age at multimorbidity onset was higher in analyses of sleep duration at older ages, but the distribution of chronic disease dyads was similar across analyses of sleep duration at 50, 60, or 70 years of age (S3 Table). The 8 most common dyads were the same in these analyses, and they represented over 50% of cases of incident multimorbidity cases in all 3 analyses. Coronary heart disease was present in 5 dyads; diabetes, cancer, and arthritis/rheumatoid arthritis in 3; and depression and heart failure in 1 of these 8 dyads.

Table 2 shows the association of sleep duration at age 50, 60, and 70 with subsequent risk of multimorbidity. In the absence of sex differences (p for interaction between sex and sleep duration >0.05), men and women were combined in the analyses. In analyses adjusted for sociodemographic variables, the risk of multimorbidity was higher in participants with a sleep duration ≤5 hours, 6 hours, and ≥9 hours compared to sleep duration of 7 hours, irrespective of the age at measurement of sleep duration. Further adjustment for health behaviors, BMI, hypertension, use of sleep medication, and prevalence of 1 of the 13 chronic diseases showed sleep duration ≤5 hours at age 50 (hazard ratio (HR) = 1.30, 95% confidence interval, 1.12 to 1.50; p < 0.001), age 60 (HR = 1.32, 1.13 to 1.55; p < 0.001), and at age 70 (HR = 1.40, 1.16 to 1.68; p < 0.001) to be associated with higher risk of multimorbidity. In these analyses, the association of sleep duration ≥9 hours at age 50 with incident multimorbidity did not reach statistical significance (HR = 1.39, 0.98 to 1.96; p = 0.067), while sleep duration ≥9 hours at age 60 (HR = 1.54, 1.15 to 2.06; p = 0.003) and 70 (HR = 1.51, 1.10 to 2.08; p = 0.010) was associated with higher risk of multimorbidity.

Table 2

Association of sleep duration at age 50, 60, and 70 with risk of multimorbidity.

	N cases/N total	Model 1: Unadjusted model (age as timescale)		Model 2: Adjusted for sociodemographic variablesb		Model 3: Model 2 + behavioral and health-related factorsc
		HR (95% CI)	p-value	HR (95% CI)	p-value	HR (95% CI)	p-value
Sleep duration at age 50	N cases/N total = 2,659/7,864; follow-up mean (SD) = 22.6 (7.5) years; mean age at event (SD) = 70.9 (7.7) years
≤5 hours	225/544	1.57 (1.36, 1.81)	<0.001	1.46 (1.27, 1.69)	<0.001	1.30 (1.12, 1.50)	<0.001
6 hours	852/2,562	1.13 (1.03, 1.23)	0.007	1.11 (1.02, 1.21)	0.020	1.08 (0.98, 1.17)	0.109
7 hours	1,184/3,589	1.00 (ref)		1.00 (ref)		1.00 (ref)
8 hours	365/1,092	1.01 (0.90, 1.14)	0.811	0.99 (0.88, 1.12)	0.918	1.00 (0.89, 1.12)	0.953
≥9 hours	33/77	1.60 (1.13, 2.27)	0.008	1.42 (1.00, 2.02)	0.047	1.39 (0.98, 1.96)	0.067
Sleep duration at age 60	N cases/N total = 2,029/6,848; follow-up mean (SD) = 13.4 (6.0) years; mean age at event (SD) = 72.0 (6.3) years
≤5 hours	202/519	1.56 (1.34, 1.83)	<0.001	1.44 (1.23, 1.68)	<0.001	1.32 (1.13, 1.55)	0.001
6 hours	645/2,095	1.16 (1.04, 1.28)	0.006	1.13 (1.02, 1.26)	0.018	1.14 (1.02, 1.26)	0.016
7 hours	793/2,882	1.00 (ref)		1.00 (ref)		1.00 (ref)
8 hours	340/1,230	1.03 (0.91, 1.17)	0.652	1.02 (0.90, 1.16)	0.715	1.06 (0.93, 1.20)	0.375
≥9 hours	49/122	1.64 (1.23, 2.19)	0.001	1.60 (1.20, 2.14)	0.001	1.54 (1.15, 2.06)	0.003
Sleep duration at age 70	N cases/N total = 1,402/5,546; follow-up mean (SD) = 6.8 (4.5) years; mean age at event (SD) = 76.0 (4.8) years
≤5 hours	152/451	1.63 (1.36, 1.95)	<0.001	1.58 (1.31, 1.90)	<0.001	1.40 (1.16, 1.68)	<0.001
6 hours	425/1,574	1.20 (1.05, 1.36)	0.007	1.18 (1.04, 1.34)	0.013	1.12 (0.98, 1.27)	0.092
7 hours	514/2,249	1.00 (ref)		1.00 (ref)		1.00 (ref)
8 hours	269/1,151	1.04 (0.90, 1.21)	0.595	1.03 (0.89, 1.20)	0.649	0.99 (0.85, 1.14)	0.843
≥9 hours	42/121	1.52 (1.11, 2.08)	0.010	1.48 (1.08, 2.03)	0.015	1.51 (1.10, 2.08)	0.010

a Multimorbidity defined as 2 or more of the following chronic diseases: diabetes, cancer, coronary heart disease, stroke, heart failure, chronic obstructive pulmonary disease, chronic kidney disease, liver disease, depression, dementia, other mental disorder, Parkinson’s disease, and arthritis/rheumatoid arthritis.

b Adjusted for age (timescale), sex, ethnicity, education, occupational position, and marital status.

c Additionally adjusted for alcohol consumption, physical activity, smoking status, fruit and vegetable consumption, BMI, hypertension, use of sleep medication, and prevalence of 1 of the 13 chronic diseases.

CI, confidence intervals; HR, hazard ratio; ref, reference; SD, standard deviation.

In sensitivity analyses excluding participants with any of the 13 chronic diseases at the measure of sleep duration (S4 Table), the pattern of associations was similar to the main analyses for sleep duration at age 50, 60, and 70 years apart from the association with sleep duration ≥9 hours at age 70 that did not reach significance although the effect size was comparable (HR = 1.52, 0.98 to 2.35; p = 0.063). Analyses excluding participants on sleep medication were similar, irrespective of age at sleep measure (S5 Table).

Accelerometer data and covariates were available on 3,920 participants who took part to the accelerometer sub-study at the 2012 wave. Of them, 3,368 participants (mean age (SD) = 68.9 (5.6), range = 60 to 83 years) were free of multimorbidity and were included in the analysis. During a mean follow-up of 6 years, 601 developed multimorbidity. The correlation between accelerometer- and questionnaire-assessed sleep duration was moderate (Pearson correlation = 0.41, p < 0.001). The shape of the association between accelerometer-assessed sleep duration and incident multimorbidity was similar to that observed with self-reported sleep duration, with the lowest risk of incident multimorbidity seen at 7 hours of sleep (Fig 3). Given the small number of incident multimorbidity cases in participants in less than 6 hours and more than 8 hours categories (Fig 3, panel D), the focus here was on the shape of the association rather than estimates of the associations.

Fig 3

Association of accelerometer assessed sleep duration in 2012–2013 (age range, 60 to 83 years) with risk of incident multimorbidity (N cases/N total = 601/3,368) over a mean follow-up of 6.0 (SD = 1.6) years.

Multimorbidity defined as 2 or more of the following chronic diseases: diabetes, cancer, coronary heart disease, stroke, heart failure, chronic obstructive pulmonary disease, chronic kidney disease, liver disease, depression, dementia, other mental disorder, Parkinson’s disease, and arthritis/rheumatoid arthritis. (A) Model unadjusted (age as timescale). (B) Model adjusted for age (timescale), sex, ethnicity, education, occupational position, and marital status. (C) Model additionally adjusted for alcohol consumption, physical activity, smoking status, fruit and vegetable consumption, BMI, hypertension, use of sleep medication, and prevalence of 1 of the 13 chronic diseases. (D) Sleep duration distribution among participants with no incident multimorbidity (blue) and those with incident multimorbidity (brown).

In post hoc analysis, first we showed results from analyses using inverse probability weighting to account for missing data to be consistent with those in the main analysis (S6 Table). Second, analysis excluding 1 disease at a time from the definition of multimorbidity showed results to be robust to the list of chronic diseases used to define multimorbidity (S7 Table). Third, we examined trajectories of sleep duration. The Pearson correlation of sleep duration at age 50 with sleep duration at age 60 was 0.49 and with sleep duration at age 70 was 0.42 (p < 0.001). A moderate correlation was also found between sleep duration at age 60 and age 70 (correlation 0.57, p < 0.001). Six trajectories of sleep duration using data on 5,510 participants were used and labeled as persistent short sleep, persistent normal sleep, persistent long sleep, change from short to normal sleep, change from normal to long sleep, and change from normal to short sleep (S8 Table). Compared to persistent normal sleep, persistent short sleep duration between age 50 and 70 was associated with increased risk of incident multimorbidity (HR = 1.17, 1.01 to 1.35; p = 0.040) in analyses adjusted for sociodemographic, behavioral, and health-related factors (S9 Table).

In a fourth post hoc analysis, sleep disturbances at age 60 and 70, measured using the Jenkins sleep problems scale, were examined and found to be greater in those with sleep duration ≤5 hours at age 60 as compared to with 7 hours of sleep (mean (SD) Jenkins sleep problems score = 10.1 (SD = 5.7) versus 4.4 (SD = 3.5); p < 0.001; S10 Table) but not in those with sleep duration ≥9 hours (4.4 (SD = 4.1) versus 4.4 (3.5); p = 0.925). A similar pattern was observed at age 70 (S10 Table). Greater sleep disturbance at age 60 and at age 70 (HR = 1.03, 1.02 to 1.04 per 1-point increase in the score at both ages; p < 0.001) was associated with increased risk of multimorbidity in analyses adjusted for sociodemographic, behavioral, and health-related factors (S11 Table).

Sleep duration at age 50 and subsequent transitions to a first chronic disease, multimorbidity, and death

A total of 7,217 participants with data on sleep duration at age 50 were free from the 13 chronic diseases considered in this study. Among them, over a mean follow-up of 25.2 years, 213 participants died without having developed any of the 13 diseases, 4,446 participants developed 1 chronic disease and of them 2,297 subsequently developed a second disease (multimorbidity), and of them 787 died. In analysis adjusted for sociodemographic, behavioral, and health-related factors, compared to sleep duration of 7 hours, sleep duration ≤5 hours was associated with increased risk of transition to a first chronic disease (HR = 1.20, 1.06 to 1.35; p = 0.003) and subsequent transition to multimorbidity (HR = 1.21, 1.03 to 1.42; p = 0.021) but not mortality (Table 3). Sleep durations longer than 7 hours were not associated with these transitions.

Table 3

Association of sleep duration at age 50 with transitions from a healthy state to first chronic disease, multimorbidity, and mortality (N = 7,217).

Sleep duration at 50y	N cases/N total	Model 1: Unadjusted model (age as timescale)		Model 2: Adjusted for sociodemographic variablesa		Model 3: Model 2 + behavioral and health-related factorsb
		HR (95% CI)	p-value	HR (95% CI)	p-value	HR (95% CI)	p-value
Healthy to first chronic disease c , transition A (N cases/N total = 4,446/7,217; mean age at event (SD) = 66.2 (8.5) years)
≤5 hours	319/474	1.31 (1.16, 1.47)	<0.001	1.28 (1.14, 1.44)	<0.001	1.20 (1.06, 1.35)	0.003
6 hours	1,428/2,350	1.07 (1.00, 1.14)	0.059	1.06 (0.99, 1.13)	0.097	1.03 (0.96, 1.10)	0.418
7 hours	2,038/3,323	1.00 (ref)		1.00 (ref)		1.00 (ref)
8 hours	621/1,008	1.00 (0.91, 1.09)	0.944	0.99 (0.90, 1.08)	0.745	0.98 (0.90, 1.08)	0.730
≥9 hours	40/62	1.16 (0.84, 1.58)	0.367	1.10 (0.80, 1.50)	0.558	1.10 (0.80, 1.50)	0.571
Healthy to death, transition B (N cases/N total = 213/7,217; mean age at event (SD) = 65.2 (9.0) years)
≤5 hours	14/474	1.24 (0.71, 2.18)	0.447	1.12 (0.64, 1.97)	0.689	0.97 (0.55, 1.72)	0.926
6 hours	71/2,350	1.17 (0.86, 1.59)	0.314	1.12 (0.82, 1.52)	0.465	1.06 (0.78, 1.44)	0.727
7 hours	98/3,323	1.00 (ref)		1.00 (ref)		1.00 (ref)
8 hours	29/1,008	0.97 (0.64, 1.47)	0.902	0.96 (0.63, 1.45)	0.829	0.96 (0.63, 1.45)	0.844
≥9 hours	1/62	na		na		na
First chronic diseasec to multimorbidityc, transition C (N cases/N total = 2,297/4,446; mean age at event (SD) = 71.9 (7.2) years)
≤5 hours	183/319	1.34 (1.15, 1.57)	0.000	1.27 (1.09, 1.50)	0.003	1.21 (1.03, 1.42)	0.021
6 hours	739/1,428	1.13 (1.02, 1.24)	0.014	1.12 (1.02, 1.23)	0.019	1.10 (1.00, 1.21)	0.041
7 hours	1,042/2,038	1.00 (ref)		1.00 (ref)		1.00 (ref)
8 hours	310/621	0.98 (0.86, 1.11)	0.768	0.98 (0.86, 1.11)	0.767	0.98 (0.86, 1.11)	0.771
≥9 hours	23/40	1.29 (0.85, 1.95)	0.226	1.18 (0.77, 1.79)	0.451	1.10 (0.72, 1.68)	0.661
First chronic diseasec to death, transition D (N cases/N total = 474/4,446; mean age at event (SD) = 68.9 (7.9) years)
≤5 hours	32/319	1.17 (0.80, 1.69)	0.422	1.16 (0.79, 1.69)	0.454	1.19 (0.81, 1.74)	0.374
6 hours	163/1,428	1.27 (1.03, 1.56)	0.023	1.27 (1.04, 1.57)	0.022	1.30 (1.06, 1.60)	0.013
7 hours	205/2,038	1.00 (ref)		1.00 (ref)		1.00 (ref)
8 hours	69/621	1.09 (0.83, 1.43)	0.543	1.09 (0.83, 1.44)	0.516	1.10 (0.83, 1.44)	0.509
≥9 hours	5/40	na		na		na
Multimorbidityc to death, transition E (N cases/N total = 787/2,297; mean age at event (SD) = 76.0 (6.8) years)
≤5 hours	65/183	1.04 (0.79, 1.35)	0.788	1.07 (0.82, 1.40)	0.636	1.07 (0.81, 1.40)	0.631
6 hours	254/739	1.09 (0.93, 1.28)	0.290	1.10 (0.94, 1.30)	0.236	1.10 (0.94, 1.30)	0.236
7 hours	348/1,042	1.00 (ref)		1.00 (ref)		1.00 (ref)
8 hours	112/310	1.06 (0.85, 1.31)	0.613	1.08 (0.87, 1.34)	0.462	1.09 (0.88, 1.35)	0.420
≥9 hours	8/23	0.93 (0.46, 1.88)	0.843	1.10 (0.54, 2.24)	0.785	1.09 (0.54, 2.22)	0.808

a Adjusted for age (timescale), sex, ethnicity, education, occupational position, and marital status.

b Additionally adjusted for alcohol consumption, physical activity, smoking status, fruit and vegetable consumption, BMI, hypertension, and use of sleep medication.

c Chronic disease among diabetes, cancer, coronary heart disease, stroke, heart failure, chronic obstructive pulmonary disease, chronic kidney disease, liver disease, depression, dementia, other mental disorder, Parkinson’s disease, and arthritis/rheumatoid arthritis; multimorbidity defined as 2 or more of these diseases.

For transitions see Fig 1.

CI, confidence intervals; HR, hazard ratio; na, not applicable (N cases ≤5); ref, reference; SD, standard deviation.

Similar findings were observed in inverse probability weighting analyses to account for missing data (S12 Table). In post hoc analysis, we also examined the association between sleep duration at age 50 and risk of mortality without considering chronic diseases over the follow-up; in analysis adjusted for sociodemographic, behavioral, and health-related factors, sleep duration ≤5 hours (HR = 1.25, 1.02 to 1.53; p = 0.034) and of 6 hours (HR = 1.17, 1.04 to 1.32; p = 0.008) was associated with higher risk of mortality over a mean follow-up of 25.2 (SD = 6.9) years (S13 Table).

Sleep duration after a first chronic disease and subsequent risk of multimorbidity and death

A total of 6,546 participants were diagnosed with 1 of the 13 chronic diseases considered during the follow-up period, constituting the target population of this analysis. Among them, 2,442 did not have data on sleep duration after the onset of this first chronic disease, 464 had data but only after the onset of multimorbidity, and 42 had missing covariates, leading to an analytical sample of 3,702 participants (S1 Fig). In analysis adjusted for sociodemographic, behavioral, and health-related factors, sleep duration ≤5 hours in this sample was associated with higher risk of incident multimorbidity (HR = 1.20, 1.03 to 1.40; p = 0.018) (Table 4). Sleep duration ≥9 hours was associated with a higher risk of multimorbidity (HR = 1.46, 1.07 to 1.99; p = 0.017) in analysis adjusted for sociodemographic factors, although the association was attenuated after adjustment for behavioral and health-related factors (HR = 1.36, 1.00 to 1.86; p = 0.051). No consistent association was found with transition to mortality. Once missing data were taken into account using inverse-probability weighting (S14 Table), findings remain substantially the same.

Table 4

Association of sleep duration after a first chronic disease with transitions from first chronic disease to multimorbidity and mortality (N = 3,702).

Sleep duration after a first chronic disease	N cases/N total	Model 1: Unadjusted model (age as timescale)		Model 2: Adjusted for sociodemographic variablesa		Model 3: Model 2 + behavioral and health-related factorsb
		HR (95% CI)	p-value	HR (95% CI)	p-value	HR (95% CI)	p-value
First chronic diseasec to multimorbidityc (N cases/N total = 1,923/3,702; mean age at event (SD) = 70.1 (7.9) years)
≤5 hours	226/385	1.33 (1.15, 1.55)	<0.001	1.25 (1.08, 1.46)	0.004	1.20 (1.03, 1.40)	0.018
6 hours	588/1,125	1.11 (0.99, 1.24)	0.066	1.10 (0.98, 1.22)	0.105	1.08 (0.97, 1.20)	0.179
7 hours	713/1,440	1.00 (ref)		1.00 (ref)		1.00 (ref)
8 hours	353/685	1.09 (0.96, 1.23)	0.206	1.08 (0.95, 1.23)	0.244	1.08 (0.95, 1.23)	0.242
≥9 hours	43/67	1.50 (1.10, 2.04)	0.010	1.46 (1.07, 1.99)	0.017	1.36 (1.00, 1.86)	0.051
First chronic diseasec to death (N cases/N total = 190/3,702; mean age at event (SD) = 69.7 (8.5) years)
≤5 hours	20/385	1.28 (0.77, 2.10)	0.338	1.23 (0.74, 2.05)	0.420	1.23 (0.74, 2.05)	0.432
6 hours	58/1,125	1.16 (0.81, 1.64)	0.414	1.13 (0.79, 1.61)	0.497	1.11 (0.78, 1.58)	0.564
7 hours	68/1,440	1.00 (ref)		1.00 (ref)		1.00 (ref)
8 hours	40/685	1.30 (0.88, 1.92)	0.192	1.32 (0.89, 1.96)	0.164	1.34 (0.90, 1.99)	0.146
≥9 hours	4/67	na		na		na
Multimorbidityc to death (N cases/N total = 545/1,923; mean age at event (SD) = 74.9 (7.5) years)
≤5 hours	63/226	1.08 (0.81, 1.43)	0.608	1.14 (0.85, 1.52)	0.382	1.17 (0.87, 1.57)	0.292
6 hours	158/588	0.98 (0.80, 1.21)	0.867	1.01 (0.82, 1.25)	0.929	0.99 (0.80, 1.23)	0.922
7 hours	194/713	1.00 (ref)		1.00 (ref)		1.00 (ref)
8 hours	121/353	1.33 (1.06, 1.67)	0.015	1.39 (1.10, 1.75)	0.005	1.41 (1.12, 1.78)	0.003
≥9 hours	9/43	0.70 (0.36, 1.37)	0.295	0.67 (0.34, 1.31)	0.243	0.70 (0.36, 1.37)	0.294

a Adjusted for age (timescale), sex, ethnicity, education, occupational position, and marital status.

b Additionally adjusted for alcohol consumption, physical activity, smoking status, fruit and vegetable consumption, BMI, hypertension, and use of sleep medication.

c Chronic disease among diabetes, cancer, coronary heart disease, stroke, heart failure, chronic obstructive pulmonary disease, chronic kidney disease, liver disease, depression, dementia, other mental disorder, Parkinson’s disease, and arthritis/rheumatoid arthritis; multimorbidity defined as 2 or more of these diseases.

CI, confidence intervals; HR, hazard ratio; na, not applicable (N cases ≤5); ref, reference; SD, standard deviation.

Discussion

This prospective study spanning over 20 years presents 3 key findings. One, short sleep duration was consistently associated with increased risk of multimorbidity, irrespective of sleep being measured in mid or late life. The analysis of transitions in health states showed short sleep to be associated with the onset of a first disease and subsequent multimorbidity but not disease prognosis, measured using mortality. Two, the results for long sleep duration were less robust as associations with multimorbidity were observed when sleep was measured at age 60 and 70 but not at 50 years. In the analyses of transitions in health states, we also found long sleep at age 50 not to be associated with disease progression although some of the transitions could not be examined due to a small number of cases. Three, the accelerometer-based measure of sleep duration, undertaken when the mean age of participants was 69 years (range 60 to 83), confirmed the shape of the association between sleep duration and multimorbidity in the main analysis with results matching those observed for self-reported sleep duration at ages 60 or 70. Taken together, these findings suggest an association between short sleep duration and development of multimorbidity.

Much of the evidence on the role of sleep duration for health comes from studies on individual chronic diseases, rather than multimorbidity. One meta-analysis suggested that the association between short sleep duration and health outcomes is stronger when sleep duration is assessed before age 65 [16] and another found long sleep duration, particularly at older ages, to be more strongly associated with chronic diseases [19]. While these studies are important, their generalizability to real-life settings is limited as most adults live with multiple rather than single chronic conditions [8]. Apart from 1 notable study that examined the association between sleep disturbance (rather than sleep duration) and number of chronic conditions over a 9-year period [31], the few studies that exist on sleep duration and multimorbidity are cross-sectional [9-13]. These studies report a U-shape association between sleep duration and multimorbidity, implying higher prevalence of multimorbidity in both individuals with short and long sleep duration [9-13]. In some studies, stronger association was found for short sleep [10,12,13] while at least 1 study suggested a stronger link with long sleep [11]. Although informative, these cross-sectional findings might reflect both the impact of sleep duration on disease incidence and the converse, that is, the effect of disease on sleep.

Using a prospective design, our results show robust evidence of an association of short sleep duration with incident multimorbidity; this was the case for sleep duration measured either in mid or late life or trajectories of sleep duration between age 50 and 70. These findings were supported by results of the multistate models where short sleep duration at age 50 was associated with higher risk of onset of a first chronic disease and subsequent multimorbidity. Short sleep duration was also associated with greater sleep disturbances—itself associated with increased risk of multimorbidity in the present study as well as in a previous study [31]—suggesting that short sleep duration might be a marker of poor sleep quality. Sleep duration and quality might impact health via their role in regulation of endocrine and metabolic processes, inflammation, and circadian rhythm [15,16]. There was no evidence in the present data that short sleep duration was associated with progression to death among those with existing chronic disease(s). This suggests that the previously reported association between short sleep duration and mortality [2-4] is likely to be driven by the association of short sleep with onset of chronic diseases that are themselves associated with risk of mortality.

There is some evidence of poorer health outcomes in long sleepers in previous studies [2,19], but the mechanisms underlying this association remain unclear [19]. Long sleep duration has been hypothesized to reflect poor overall sleep quality that could have a detrimental impact on health [15,19,31], although this hypothesis was not supported in our study where the sleep disturbances were similar in those sleeping ≥9 hours and those sleeping 7 hours. It is also possible that long sleep duration is a marker of underlying conditions that are themselves associated with an increased risk of chronic disease and mortality. This hypothesis is supported by studies showing the association of long sleep duration with health to be based primarily on older adults, who are more likely to have preexisting medical conditions [19]. Our analyses provide further support for this hypothesis as the association of long sleep duration with multimorbidity was attenuated when sleep duration was measured in disease-free participants at age 50. We also found long sleep duration after a first disease to be associated with subsequent risk of multimorbidity although this association was partly attributable to behavioral and health-related factors. These findings support the notion that previously reported associations between long sleep duration and health might reflect increased sleep duration among those with existing health conditions, rather than long sleep duration being an important risk factor for disease onset.

A major strength of this study is the long follow-up, repeated measures that allowed analyses on sleep duration at ages 50, 60, and 70 along with sleep duration trajectories over this age range. Compared to conventional analyses that examined associations between sleep duration and health outcomes, the use of multistate models provides additional insight into the association of sleep duration with the course of disease, including the finding that sleep duration is associated with onset of a chronic disease and subsequent multimorbidity but not with mortality among persons with these conditions.

Our findings should also be considered in light of the limitations of the study. First, like most large-scale studies on sleep, we used self-reported sleep which is likely to be subject to reporting bias. Although the correlation between accelerometer-assessed and self-reported sleep duration was moderate in our study, the shape of the association with multimorbidity risk was similar with both measures. Second, data on sleep quality were available only at age 60 and 70. Third, participants from the Whitehall II cohort study were all employed at recruitment to the study and likely to be healthier than the general population. However, the association between risk factors and health outcomes has been shown to be similar to that observed in the general population [40]. Fourth, participants were mostly of white ethnicity, reflecting the population of the country in 1991 [41], and whether results are generalizable to other populations is unknown. Fifth, despite the use of several covariates residual confounding can be an issue in observational studies. For example, short sleep duration and sleep disturbances may reflect the symptoms of undiagnosed diseases at sleep measures such as depression or arthritis. Considering the wide range of covariates, a confounder would need to have a risk ratio of 2 (E-value) with both the exposure and the outcome to explain away the association between sleep duration at age 50 and multimorbidity. Finally, the mortality numbers in the ≥9 hours sleep duration group was small in some of the analyses, not allowing firm conclusions to be drawn on the association between long sleep and mortality.

With population ageing and increases in life expectancy, living with multiple chronic conditions is common among older adults in high-income countries [6,42]. Multimorbidity presents a challenge as it is associated with high health care service use, hospitalizations, and disability; a further concern is that contemporary health care systems are organized around treatment and care of individual diseases rather than multimorbidity [43]. Primary prevention of a first chronic disease and secondary prevention to reduce risk of multimorbidity among those with a first chronic disease are thus important in addressing the burden of multimorbidity [42]. The present findings along with evidence from previous studies show the importance of sleep duration across the lifecourse for health outcomes at older ages [1-4]. Further research using objective measures of sleep duration would allow better understanding of the importance of sleep duration for chronic disease and multimorbidity.

In conclusions, findings from the present study suggest short sleep duration in midlife and old age is associated with higher risk of onset of chronic disease and multimorbidity. These findings support the promotion of good sleep hygiene in both primary and secondary prevention by targeting behavioral and environmental conditions that affect sleep duration and quality [44].

STROBE Checklist.

(DOCX)

Click here for additional data file.

Analysis plan drafted in August 2021 before data analysis.

(DOCX)

Click here for additional data file.

Flowchart for the analyses on the association of sleep duration after a first chronic disease with transitions from first chronic disease to multimorbidity and mortality.

(TIF)

Click here for additional data file.

Characteristics of the study population at age 60.

(DOCX)

Click here for additional data file.

Characteristics of the study population at age 70.

(DOCX)

Click here for additional data file.

Description of chronic disease dyads in the analysis of sleep duration at age 50, 60, and 70 and incident multimorbidity.

(DOCX)

Click here for additional data file.

Association of sleep duration at age 50, 60, and 70 with risk of multimorbidity among participants free from prevalent chronic disease.

(DOCX)

Click here for additional data file.

Association of sleep duration at age 50, 60, and 70 with risk of multimorbidity in analyses excluding participants on sleep medication.

(DOCX)

Click here for additional data file.

Association of sleep duration at age 50, 60, and 70 with risk of multimorbidity using inverse probability weighting analyses to take missing data into account.

(DOCX)

Click here for additional data file.

Association of sleep duration at age 50, 60, and 70 with risk of multimorbidity: Impact of removing 1 chronic disease at a time from the list of chronic diseases included in the definition of multimorbidity.

(DOCX)

Click here for additional data file.

Description of sleep duration at age 50, 60, and 70 by groups of trajectories of sleep duration between age 50 and 70.

(DOCX)

Click here for additional data file.

Association of trajectories of sleep duration between age 50 and 70 with risk of multimorbidity.

(DOCX)

Click here for additional data file.

Sleep disturbances assessed using the Jenkins sleep problems scale as a function of sleep duration.

(DOCX)

Click here for additional data file.

Association of the Jenkins sleep problems scale with risk of multimorbidity.

(DOCX)

Click here for additional data file.

Association of sleep duration at age 50 with transitions from a healthy state to first chronic disease, multimorbidity, and mortality (N = 7,217) using inverse probability weighting analyses to take missing data into account.

(DOCX)

Click here for additional data file.

Association of sleep duration at age 50 with risk of mortality.

(DOCX)

Click here for additional data file.

Association of sleep duration after first chronic disease with transitions to multimorbidity and mortality in analyses using inverse probability weighting analyses to take missing data into account.

(DOCX)

Click here for additional data file.

28 Mar 2022

Dear Dr Sabia,

Thank you for submitting your manuscript entitled "Sleep duration at age 50, 60, and 70 and risk of multimorbidity: a 25-year follow-up study" for consideration by PLOS Medicine.

Your manuscript has now been evaluated by the PLOS Medicine editorial staff and I am writing to let you know that we would like to send your submission out for external peer review.

However, before we can send your manuscript to reviewers, we need you to complete your submission by providing the metadata that is required for full assessment. To this end, please login to Editorial Manager where you will find the paper in the 'Submissions Needing Revisions' folder on your homepage. Please click 'Revise Submission' from the Action Links and complete all additional questions in the submission questionnaire.

Please re-submit your manuscript within two working days, i.e. by Mar 30 2022 11:59PM.

Login to Editorial Manager here: https://www.editorialmanager.com/pmedicine

Once your full submission is complete, your paper will undergo a series of checks in preparation for peer review. Once your manuscript has passed all checks it will be sent out for review.

Feel free to email us at plosmedicine@plos.org if you have any queries relating to your submission.

Kind regards,

Caitlin Moyer, Ph.D.

Associate Editor

PLOS Medicine

9 Jun 2022

Dear Dr. Sabia,

Thank you very much for submitting your manuscript "Sleep duration at age 50, 60, and 70 and risk of multimorbidity: a 25-year follow-up study" (PMEDICINE-D-22-00972R1) for consideration at PLOS Medicine.

Your paper was evaluated by a senior editor and discussed among all the editors here. It was also discussed with an academic editor with relevant expertise, and sent to four independent reviewers, including a statistical reviewer. The reviews are appended at the bottom of this email and any accompanying reviewer attachments can be seen via the link below:

[LINK]

In light of these reviews, I am afraid that we will not be able to accept the manuscript for publication in the journal in its current form, but we would like to consider a revised version that addresses the reviewers' and editors' comments. Obviously we cannot make any decision about publication until we have seen the revised manuscript and your response, and we plan to seek re-review by one or more of the reviewers.

In revising the manuscript for further consideration, your revisions should address the specific points made by each reviewer and the editors. Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments, the changes you have made in the manuscript, and include either an excerpt of the revised text or the location (eg: page and line number) where each change can be found. Please submit a clean version of the paper as the main article file; a version with changes marked should be uploaded as a marked up manuscript.

In addition, we request that you upload any figures associated with your paper as individual TIF or EPS files with 300dpi resolution at resubmission; please read our figure guidelines for more information on our requirements: http://journals.plos.org/plosmedicine/s/figures. While revising your submission, please upload your figure files to the PACE digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at PLOSMedicine@plos.org.

We expect to receive your revised manuscript by Jun 30 2022 11:59PM. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

We ask every co-author listed on the manuscript to fill in a contributing author statement, making sure to declare all competing interests. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. If new competing interests are declared later in the revision process, this may also hold up the submission. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT. You can see our competing interests policy here: http://journals.plos.org/plosmedicine/s/competing-interests.

Please use the following link to submit the revised manuscript:

https://www.editorialmanager.com/pmedicine/

Your article can be found in the "Submissions Needing Revision" folder.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

We look forward to receiving your revised manuscript.

Sincerely,

Caitlin Moyer, Ph.D.

Associate Editor

PLOS Medicine

plosmedicine.org

-----------------------------------------------------------

Requests from the editors:

1. Title: Please revise your title according to PLOS Medicine's style. Your title must be nondeclarative and not a question. It should begin with main concept if possible. "Effect of" should be used only if causality can be inferred, i.e., for an RCT. Please place the study design ("A randomized controlled trial," "A retrospective study," "A modelling study," etc.) in the subtitle (ie, after a colon).

2. Data availability statement: The Data Availability Statement (DAS) requires revision. For each data source used in your study:

a) If the data are freely or publicly available, note this and state the location of the data: within the paper, in Supporting Information files, or in a public repository (include the DOI or accession number).

b) If the data are owned by a third party but freely available upon request, please note this and state the owner of the data set and contact information for data requests (web or email address). Note that a study author cannot be the contact person for the data.

c) If the data are not freely available, please describe briefly the ethical, legal, or contractual restriction that prevents you from sharing it. Please also include an appropriate contact (web or email address) for inquiries (again, this cannot be a study author).

3. Line numbers: Please provide line numbers, running continuously throughout the document, with the revised version.

4. Abstract: Methods and Findings: Please provide some summary information on the Whitehall II population. Please mention the study dates, and the setting of your study. Please mention the study design.

5. Abstract: Methods and Findings: Please quantify the main results with both 95% CIs and p values. Please ensure that all numbers presented in the abstract are present and identical to numbers presented in the main manuscript text.

6. Abstract: Methods and Findings: In the last sentence of the Abstract Methods and Findings section, please describe the main limitation(s) of the study's methodology.

7. Abstract: Conclusions: Please address the study implications without overreaching what can be concluded from the data; the phrase "In this study, we observed ..." may be useful. Please revise the Conclusions section such that it provides an interpretation of the study based on the results presented in the abstract, emphasizing what is new without overstating your conclusions. For example, the study does not seem to report evidence suggesting that long sleep is a marker of underlying health conditions.

8. Methods: How was ethnicity defined and by whom?

9. Methods: Repeated measures design: Please describe how clustering of repeat measurements at the individual-level was accounted for.

10. Methods: “We repeated this analysis using inverse probability weighting to account for missing data…” Please provide more information/a table describing the nature and amount of missing data.

11. Methods: “...(3) by the association between accelerometer-assessed sleep duration at mean age 69 years and incident multimorbidity. The accelerometer measure and covariates were drawn from the 2012-2013 wave of data collection. Given the detailed data on sleep duration extracted from the accelerometer, we used restricted cubic spline regressions with Harrell knots [32], Stata command xblc [33], with 7-hour sleep as the reference to examine the shape of the association between sleep duration and dementia risk.” Please clarify if the accelerometer data were available for participants at other ages, as well. Please clarify the reason why the association between sleep duration and dementia is highlighted in this analysis.

12. Methods: Please ensure that the study is reported according to the STROBE guideline, and include the completed STROBE checklist as Supporting Information. Please add the following statement, or similar, to the Methods: "This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S1 Checklist)."

The STROBE guideline can be found here: http://www.equator-network.org/reporting-guidelines/strobe/

When completing the checklist, please use section and paragraph numbers, rather than page numbers.

13. Methods: Did your study have a prospective protocol or analysis plan? Please state this (either way) early in the Methods section.

a) If a prospective analysis plan (from your funding proposal, IRB or other ethics committee submission, study protocol, or other planning document written before analyzing the data) was used in designing the study, please include the relevant prospectively written document with your revised manuscript as a Supporting Information file to be published alongside your study, and cite it in the Methods section. A legend for this file should be included at the end of your manuscript.

b) If no such document exists, please make sure that the Methods section transparently describes when analyses were planned, and when/why any data-driven changes to analyses took place.

c) In either case, changes in the analysis-- including those made in response to peer review comments-- should be identified as such in the Methods section of the paper, with rationale.

14. Results: Please provide both 95% CIs and p values for all results presented in the text.

15. Results: “Sleep duration ≥9 hours at age 60 (1.57, 1.17-2.09) and 70 (1.45, 1.05-1.98) years was associated with higher risk of multimorbidity” Please also present the result for sleep duration greater than 9 hours at age 50.

16. Results: “the pattern of associations was similar for analyses using sleep duration at age 50, 60, and 70 years although some of the associations were under-powered.” We suggest providing more detail on how the analyses were determined to be under powered, or changing the wording to reflect that while the results of these analyses were directionally similar, they were no longer statistically significant.

17. Results: “Accelerometer data and covariates were available on 3920 participants.” Please clarify the age group for this analysis. Please provide the HR, 95% CI and p values for this analysis.

18. Results: For example at: “Sleep duration ≥9 hours was also associated with a higher risk of multimorbidity (1.46, 1.07-1.99), although the association was attenuated after adjustment for behavioral and health-related factors (1.36, 1.00-1.86).” Throughout the results section, it is not always clear where the unadjusted results, or results from Model 1 or Model 2 are being presented. Please indicate which factors are adjusted for each analysis, unless this is consistent throughout.

19. Discussion: Please present and organize the Discussion as follows: a short, clear summary of the article's findings; what the study adds to existing research and where and why the results may differ from previous research; strengths and limitations of the study; implications and next steps for research, clinical practice, and/or public policy; one-paragraph conclusion.

20. Discussion: Here and throughout, please avoid any language that implies causality: “The analyses of health state transitions also indicated that long sleep at age 50 does not affect disease progression…”

21. Discussion: There was no evidence in the present data that short sleep duration plays an important role in disease prognosis…” Please clarify that you are referring to mortality.

22. Figures and Tables: Please be sure that there are titles and legends for each individual table and figure, including those in the Supporting Information. Please fully define any abbreviations used in figures and tables within the text of the legend.

23. Table 1 (and eTable 1 and eTable 2): Please report on the full set of characteristics of the study population (e.g. where there are more than two categories for each variable). Please provide a summary for the total group, in addition to the breakdown by sleep duration.

24. Table 2, Table 3, Table 4 and eTable 4, eTable 5, and eTable 6: Please report exact p values, instead of *p<0.05. Please also present the results from unadjusted analyses.

Comments from the reviewers:

Reviewer #1: This study examines sleep duration as a potential risk factor for first chronic disease, subsequent multimorbidity and mortality using data spanning 25 years.

Comments:

"Self-reported sleep duration was measured at age 50 and incidence of multimorbidity was defined as presence of two or more of 13 chronic diseases."

Can the authors please comment on the potential for self reporting bias?

"In 2012-2013 an accelerometer sub-study was undertaken on participants who attended the central London research clinic or were assessed at home if they resided in the South-Eastern regions of England. Wrist-worn accelerometers, the GENEActiv (Activinsights Ltd, Kimbolton, UK), were worn 24h over 9 consecutive days [22]. Sleep duration was estimated using a validated algorithm guided by a sleep log [23]; data from the first and last nights were removed leading to data over 7 nights. Usual daily sleep duration was estimated as the mean of sleep duration over the 7 nights and for those with less than 7 7 nights of measurement, weighted average of sleep duration was calculated as: 5 x week night sleep duration + 2 x weekend night sleep duration)/7. "

Are the authors able to explore and comment on how this accelerometer data compares to self-reported estimates of sleep duration?

"The association of sleep duration at age 50, 60, and 70 with risk of incident multimorbidity was examined using Cox proportional-hazards regression with age as the time scale in participants free from multimorbidity at the measurement of sleep duration. Data were censored at date of multimorbidity diagnosis, death to account for competing risk [30, 31], or 31st of March 2019, whichever came first. Age at the beginning of the follow-up was the age at clinical assessment closest to target age in the analyses (50, 60, and 70 years) from which the sleep duration measure and covariates were drawn. The proportional hazards assumption was verified using Schoenfeld residuals. Analyses were adjusted first for sociodemographic measures (Model 1), and then additionally for behavioral and health-related factors (Model 2)."

The authors have applied rigorous modelling methods, appropriately adjusting for potential confounding and reporting the results accurately.

"To examine the robustness of our findings, we repeated the main analysis (1) in participants free from any chronic disease at the sleep measurement, (2) excluding users of sleep medication and (3) by investigating the association between accelerometer-assessed sleep duration at mean age 69 years and incident multimorbidity."

and

"In additional analyses, we examined the association of sleep duration after onset of a first chronic disease with transitions to multimorbidity and death, again using a multi-state model. The follow-up here started at the measure of first sleep duration following the onset of a first chronic disease. We repeated this analysis using inverse probability weighting to account for missing data that arise from the fact that 10 sleep duration was not measured after a first chronic disease and before onset of multimorbidity in some participants [34]"

The authors have suitably conducted additional analyses which help to demonstrate the robustness of the study findings.

"No association was found for mortality among those with existing chronic diseases. "

Did the authors consider exploring an overall transition analysis with death modelled as the outcome for the whole cohort (i.e. not conditional on the non-existence or existence of chronic disease(s))?

"In analyses excluding participants with any of the 13 chronic diseases at measurement of sleep duration (eTable 4), the pattern of associations was similar for analyses using sleep duration at age 50, 60, and 70 years although some of the associations were under-powered."

Can the authors please further expand on the power analysis they refer to here?

"The shape of the association between accelerometer-assessed sleep duration and incident multimorbidity confirmed that 7 hours of sleep was associated with the lowest risk of incident multimorbidity (Figure 3). "

Can the authors please comment on Figure 3b, which shows adjusted CIs crossing zero, suggesting no evidence of an increased HR for < or > 7 hours sleep?

Table 1 (and eTables 1 and 2): Can the authors consider additionally presenting BMI as a continuous variable here please?

Reviewer #2:

This is a prospective cohort study addressing the role of sleep duration in the progression from a disease-free state to chronic disease, multimorbidity, and death shows short sleep duration to be a risk factor for the onset of chronic disease and multimorbidity but not for subsequent mortality in those with chronic disease(s). The paper is well written and with an extensive and detailed statistical analysis. I would like to point only one minor suggestion, please see the details bellow.

Page 9 - Statistical analysis, section Association between sleep duration at different ages and incident multimorbidity, last line "…with 7-hour sleep as the reference to examine the shape of the association between sleep duration and dementia risk."

This sentence does not seem to make sense here at this paper since assessing dementia risk was not an objective of this paper. Dementia was only assessed among the remain chronic diseases.

Reviewer #3: This 25-year population-based cohort study leverages data on 13 chronic diseases to assess the relation between (self-reported) sleep duration and multimorbidity at different ages. Using sophisticated multistate models, they are able to show that short sleep duration (shorter or equal to 5 hours) increases the relative risk of any chronic disease and multimorbidity with 20% compared to longer sleep duration, but not to death.

The White Hall II study has several key strengths, including long-term, meticulously acquired health data on thousands of individuals. It also repeatedly assessed sleep behavior at various ages, and what I particularly like: a simultaneous assessment of objective sleep duration with that of self-report, allowing verification of self-reported data, as well as the authors' endeavor to repeat analyses with inverse probability weighting. Finally, I would like to congratulate the authors with their careful selection of diseases to define multimorbidity, while rightfully excluding risk factors.

I have three main concerns, and several minor suggestions to further improve the paper.

1) I like the analyses and idea of this study, but I struggle with interpreting the relevance of findings for public health or clinicians. Assessing relations with general health markers such as physical activity, sleep or diet with multimorbidity seems troublesome for etiological purposes since it is hard to say what specific disease (combinations) particularly drive associations. As this study appears to be of etiological origin (presentation of relative risks, and "Our findings support an etiological role of short sleep duration for onset of chronic diseases, including a first chronic disease and subsequent multimorbidity"), it seems that we would like to know how short sleep duration pathophysiologically relates to what specific diseases. This provides clinicians a direct target for intervention.

2) Only short sleep duration (equal or less than 5 hours) related to disease and multimorbidity. It raises the question why particularly this cutoff was chosen, as 5 hours is exceptionally short for an average weekday, and indeed prevalence is low (7.5%). It would rather be interesting to study the relationship continuously, by entering sleep duration per 1 hour into the models.

3) The authors use sophisticated statistical models to disentangle relations over the life course. It is perhaps valuable to explain to the (general) reader why these models are relevant for this particular study compared to for instance traditional cox models. This can be done briefly in the introduction, and more elaborately in the methods part.

Abstract

-Rather the increased risk of multimorbidity associated with long sleep duration in those with existing disease might reflect the need for longer sleep in those with underlying chronic conditions

-This is an interesting interpretation, but it makes me wonder why this should not also hold for the relation with death?

-briefly add some demographics of the study population, including prevalence of short sleep duration as well as limitations of the study

Introduction

-although a number of outstanding questions remain regarding the nature of this relation. I would then rephrase "one, multiple.." to "First, multiple.." to increase readability.

Methods:

-Through linkage with medical records, data on a host of diseases are available. Were diagnoses also verified by study physicians? Is there any data available on the accuracy of these records?

-Apart from inverse probability weighting, how did the authors handle missing data in the main analyses?

-Perhaps add STROBE checklist

-add analysis with E-value, to address potential residual confounding that could not be adjusted for

Discussion

-Is there any data available on sleep quality? This seems relevant to include as authors specifically relate to "Long sleep duration might reflect poor overall sleep quality that could have a detrimental impact on health".

-Limitations of study are mentioned but could be expanded:

-if data on sleep quality is not present, this should be added

-perhaps the authors can reflect on the diversity of their study in terms of ethnicity?

-Since duration does not directly tell us something about quality, I am not sure whether I agree with the following statement: "Our findings suggest short sleep duration in midlife and old age is associated with higher risk of onset of a first disease, and subsequent multimorbidity among those with an existing condition. This supports promotion of good sleep hygiene in both primary and secondary prevention." Can the authors rephrase? What should clinicians do based on these findings, sole advice to sleep longer or stay longer in bed seems not so relevant if patients are not able to sleep.

Reviewer #4: The present study analyses the association between sleep duration and risk of first chronic disease, multimorbidity and death. While the analytical approach seems robust, there are several aspects that may threaten the internal validity of the findings.

- The main problem I see with this study is the low number of chronic diseases taken into consideration. This increases the risk of reverse causality given the potential heterogeneity in subclinical health states leading to and departing from a state of multimorbidity. If the authors cannot account for more diseases due to data availability issues, they should consider adjusting their models for some other comprehensive measure of health, e.g., walking speed, number of medications, muscle strength, etc.

- One still wonders whether the potential impact of sleep duration on incident chronic disease and multimorbidity would still exist above and beyond the link with specific chronic conditions. One possible way to verify this is that the authors repeat the analyses after removing each of the 13 chronic conditions from their count, one at a time.

- Simply measuring sleep duration in absolute terms seems suboptimal. Data on sleep quality (e.g., sleep efficiency, sleep latency) have been shown to be important aspects of sleep. Moreover, assessing sleep duration in absolute terms could result in bias for those people with usual shorter/longer sleep duration. Also, what is the stability of self-reported sleep duration across time? Given that the authors have access to several of these measures over time, it may be pertinent to check this; even more so considering the long follow-up time for the 50-year and 60-year groups (22.6 and 13.4 years respectively). Last, what is the correlation between the subjective and objectively measured sleep duration? This should also be feasible to answer with their data and could shed further light into the validity of the findings.

- Chronic pain and depressive symptoms are well-known causes of sleeping problems and are, at the same time, closely linked to the development of multimorbidity. In order to untangle these potential confounding effects, could the authors repeat their analyses adjusting for these conditions?

Other minor comments:

- In the abstract, the authors mentioned that "Self-reported sleep duration was measured at age 50", but they later clarify this was measured at ages 50, 60 and 70 in the main text. Please update this information in the abstract.

- Why do authors call it age 50, 60 and 70 if 10-year approximations have been used? I would have rather worked with (and labelled them as) quinquagenarian, sexagenarian, septuagenarians subjects if the authors wish to make such an approximation.

- Revise the following sentence in the methods section concerning the dementia outcome: "we used restricted cubic spline regressions with Harrell knots [32], Stata command xblc [33], with 7-hour sleep as the reference to examine the shape of the association between sleep duration and dementia risk".

Any attachments provided with reviews can be seen via the following link:

[LINK]

30 Jul 2022

Submitted filename: response-final.docx

Click here for additional data file.

9 Sep 2022

Dear Dr. Sabia,

Thank you very much for re-submitting your manuscript "Association of sleep duration at age 50, 60, and 70 with risk of multimorbidity: 25-year follow-up of the Whitehall II cohort study" (PMEDICINE-D-22-00972R2) for review by PLOS Medicine.

I have discussed the paper with my colleagues and the academic editor and it was also seen again by 3 reviewers. I am pleased to say that provided the remaining editorial and production issues are dealt with we are planning to accept the paper for publication in the journal.

The remaining issues that need to be addressed are listed at the end of this email. Any accompanying reviewer attachments can be seen via the link below. Please take these into account before resubmitting your manuscript:

[LINK]

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

In revising the manuscript for further consideration here, please ensure you address the specific points made by each reviewer and the editors. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments and the changes you have made in the manuscript. Please submit a clean version of the paper as the main article file. A version with changes marked must also be uploaded as a marked up manuscript file.

Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. If you haven't already, we ask that you provide a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract.

We expect to receive your revised manuscript within 1 week. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

We ask every co-author listed on the manuscript to fill in a contributing author statement. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT.

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript.

Please note, when your manuscript is accepted, an uncorrected proof of your manuscript will be published online ahead of the final version, unless you've already opted out via the online submission form. If, for any reason, you do not want an earlier version of your manuscript published online or are unsure if you have already indicated as such, please let the journal staff know immediately at plosmedicine@plos.org.

If you have any questions in the meantime, please contact me or the journal staff on plosmedicine@plos.org.

We look forward to receiving the revised manuscript by Sep 16 2022 11:59PM.

Sincerely,

Caitlin Moyer, Ph.D.

Associate Editor

PLOS Medicine

plosmedicine.org

------------------------------------------------------------

Requests from Editors:

1. Response to reviewers: Please address the remaining comment of reviewer 4.

2. Title: Thank you for revising the title. It may be helpful to also include the study setting in the title: Association of sleep duration at age 50, 60, and 70 with risk of multimorbidity among adults in England: 25-year follow-up of the Whitehall II cohort study”. Or similar.

3. Financial disclosure: Please include the URL of each funder website.

4. Data availability statement: Thank you for revising the data availability statement. If possible, please also include a contact email address where data access questions may be directed.

5. Abstract: Methods and Findings: Please ensure that all numbers presented in the abstract are present and identical to numbers presented in the main manuscript text.

6. Abstract: Methods and Findings: Please mention the important covariates adjusted for in the analyses (e.g. socio-demographic measures, behavioral and health-related factors).

7. Abstract: Methods and Findings: Please report p values to 2 decimal places for values of 0.01 or greater, and to 3 decimal places for values between 0.001 and 0.01.

8. Abstract: Line 49-50: Please consider mentioning some limitations of the study design, for example the potential for reverse causality, issues with generalizability, or self-reported sleep data could be mentioned here.

9. Abstract: Line 54-55: We suggest tempering this sentence slightly, given possibility of reverse causality: “Our findings suggest an association between short sleep duration and multimorbidity.”

10. Author summary: Line 80: We suggest: “There was no clear evidence for an association between long sleep duration at age 50 and risk of chronic disease.”

11. Lines 514-529: Please remove the Competing interest statement, Funding disclosure statement, and Data availability statement from the main text. Please make sure all information is completely and accurately entered into the manuscript submission system.

12. Figure 1: In the legend, please include some explanation for the various transitions (A-E).

Comments from Reviewers:

Reviewer #1: Many thanks to the authors for considering and responding to each comment in turn, undertaking additional analyses and amending the manuscript as required.

Reviewer #3: Dear authors, thank you for your extensive and clear replies. I have no further comments.

Best

Reviewer #4: The authors have adequately responded to my comments.

Regarding my first comment related to the operationalization of multimorbidity based on 13 chronic conditions, while I understand the justification provided by the authors, I still think they should add a limitation in relation to the risk of reverse causality given subclinical health states leading to and departing from a state of multimorbidity. In other words, short sleep duration and sleep disturbances may in fact represent the symptoms of already established disorders (e.g., depression, arthritis) that were still undiagnosed at baseline.

Any attachments provided with reviews can be seen via the following link:

[LINK]

12 Sep 2022

Submitted filename: response-final.docx

Click here for additional data file.

13 Sep 2022

Dear Dr Sabia,

On behalf of my colleagues and the Academic Editor, Sanjay Basu, I am pleased to inform you that we have agreed to publish your manuscript "Association of sleep duration at age 50, 60, and 70 with risk of multimorbidity in the UK: 25-year follow-up of the Whitehall II cohort study" (PMEDICINE-D-22-00972R3) in PLOS Medicine.

Before your manuscript can be formally accepted you will need to complete some formatting changes, which you will receive in a follow up email. Please be aware that it may take several days for you to receive this email; during this time no action is required by you. Once you have received these formatting requests, please note that your manuscript will not be scheduled for publication until you have made the required changes.

In the meantime, please log into Editorial Manager at http://www.editorialmanager.com/pmedicine/, click the "Update My Information" link at the top of the page, and update your user information to ensure an efficient production process.

Please also address the following editorial points:

-Title: Thank you for revising, please add “years” for the ages, and please update the title in both the main manuscript file and the manuscript submission system: “Association of sleep duration at age 50, 60,

and 70 years with risk of multimorbidity in the UK: 25-year follow-up of the Whitehall II cohort study”

-Line 89, and throughout: Please remove spaces within reference brackets [1,2]. Please check and revise this throughout the main text.

PRESS

We frequently collaborate with press offices. If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximise its impact. If the press office is planning to promote your findings, we would be grateful if they could coordinate with medicinepress@plos.org. If you have not yet opted out of the early version process, we ask that you notify us immediately of any press plans so that we may do so on your behalf.

We also ask that you take this opportunity to read our Embargo Policy regarding the discussion, promotion and media coverage of work that is yet to be published by PLOS. As your manuscript is not yet published, it is bound by the conditions of our Embargo Policy. Please be aware that this policy is in place both to ensure that any press coverage of your article is fully substantiated and to provide a direct link between such coverage and the published work. For full details of our Embargo Policy, please visit http://www.plos.org/about/media-inquiries/embargo-policy/.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Thank you again for submitting to PLOS Medicine. We look forward to publishing your paper.

Sincerely,

Caitlin Moyer, Ph.D.

Associate Editor

PLOS Medicine