Work stress and risk of death in men and women with and without cardiometabolic disease: a multicohort study

Abstract

Background:
Although some cardiovascular disease prevention guidelines suggest a need to manage work stress in patients with established cardiometabolic disease, the evidence base for this recommendation is weak. We sought to clarify the status of stress as a risk factor in cardiometabolic disease by investigating the associations between work stress and mortality in men and women with and without pre-existing cardiometabolic disease.

Methods:
In this multicohort study, we used data from seven cohort studies in the IPD-Work consortium, initiated between 1985 and 2002 in Finland, France, Sweden, and the UK, to examine the association between work stress and mortality. Work stress was denoted as job strain or effort-reward imbalance at work. We extracted individual-level data on prevalent cardiometabolic diseases (coronary heart disease, stroke, or diabetes [without differentiation by diabetes type]) at baseline. Work stressors, socioeconomic status, and conventional and lifestyle risk factors (systolic and diastolic blood pressure, total cholesterol, smoking status, BMI, physical activity, and alcohol consumption) were also assessed at baseline. Mortality data, including date and cause of death, were obtained from national death registries. We used Cox proportional hazards regression to study the associations of work stressors with mortality in men and women with and without cardiometabolic disease.

Results:
We identified 102 633 individuals with 1 423 753 person-years at risk (mean follow-up 13·9 years [SD 3·9]), of whom 3441 had prevalent cardiometabolic disease at baseline and 3841 died during follow-up. In men with cardiometabolic disease, age-standardised mortality rates were substantially higher in people with job strain (149·8 per 10 000 person-years) than in those without (97·7 per 10 000 person-years; mortality difference 52·1 per 10 000 person-years; multivariable-adjusted hazard ratio [HR] 1·68, 95% CI 1·19-2·35). This mortality difference for job strain was almost as great as that for current smoking versus former smoking (78·1 per 10 000 person-years) and greater than those due to hypertension, high total cholesterol concentration, obesity, physical inactivity, and high alcohol consumption relative to the corresponding lower risk groups (mortality difference 5·9-44·0 per 10 000 person-years). Excess mortality associated with job strain was also noted in men with cardiometabolic disease who had achieved treatment targets, including groups with a healthy lifestyle (HR 2·01, 95% CI 1·18-3·43) and those with normal blood pressure and no dyslipidaemia (6·17, 1·74-21·9). In all women and in men without cardiometabolic disease, relative risk estimates for the work stress-mortality association were not significant, apart from effort-reward imbalance in men without cardiometabolic disease (mortality difference 6·6 per 10 000 person-years; multivariable-adjusted HR 1·22, 1·06-1·41).

Interpretation:
In men with cardiometabolic disease, the contribution of job strain to risk of death was clinically significant and independent of conventional risk factors and their treatment, and measured lifestyle factors. Standard care targeting conventional risk factors is therefore unlikely to mitigate the mortality risk associated with job strain in this population.

Funding:
NordForsk, UK Medical Research Council, and Academy of Finland.

Full text

www.thelancet.com/diabetes-endocrinology Published online June 5, 2018 http://dx.doi.org/10.1016/S2213-8587(18)30140-2
1
Articles
Lancet Diabetes Endocrinol 2018
Published Online
June 5, 2018
http://dx.doi.org/10.1016/
S2213-8587(18)30140-2
See Online/Comment
http://dx.doi.org/10.1016/
S2213-8587(18)30172-4
Clinicum, Faculty of Medicine,
and Helsinki Institute of Life
Science
(Prof M Kivimäki FMedSci,
J Pentti MSc, S T Nyberg PhD,
R Luukkonen PhD,
Prof T Strandberg MD), Institute
of Behavioural Sciences
(M Jokela PhD), Department of
Public Health,
(Prof M Koskenvuo MD), and
Faculty of Social Sciences
(Prof A Kouvonen PhD),
University of Helsinki, Helsinki,
Finland; Department of
Epidemiology and Public Health
(Prof M Kivimäki, J E Ferrie PhD,
Prof G D Batty DSc,
A Singh-Manoux PhD,
Prof A Steptoe DSc) and
National Centre for
Cardiovascular Prevention and
Outcomes
(Prof J Deanfield FRCP),
University College London,
London, UK; Department of
Public Health, University of
Turku, Turku, Finland (J Pentti,
S Suominen MD,
Prof J Vahtera MD); School of
Social and Community
Medicine, University of Bristol,
Bristol, UK (J E Ferrie); Institute
of Public Health and Caring
Sciences, University of Uppsala,
Uppsala, Sweden
(Prof M Virtanen PhD); Centre
for Occupational and
Environmental Medicine,
Stockholm County Council,
Stockholm, Sweden
(Prof L Alfredsson PhD,
E I Fransson PhD); Institute of
Environmental Medicine,
Karolinska Institutet,
Stockholm, Sweden
(Prof L Alfredsson); Institute for
Medical Sociology, Medical
Faculty, University of
Düsseldorf, Düsseldorf,
Work stress and risk of death in men and women with and
without cardiometabolic disease: a multicohort study
Mika Kivimäki, Jaana Pentti, Jane E Ferrie, G David Batty, Solja T Nyberg, Markus Jokela, Marianna Virtanen, Lars Alfredsson, Nico Dragano,
Eleonor I Fransson, Marcel Goldberg, Anders Knutsson, Markku Koskenvuo, Aki Koskinen, Anne Kouvonen, Ritva Luukkonen, Tuula Oksanen,
Reiner Rugulies, Johannes Siegrist, Archana Singh-Manoux, Sakari Suominen, Töres Theorell, Ari Väänänen, Jussi Vahtera, Peter J M Westerholm,
Hugo Westerlund, Marie Zins, Timo Strandberg, Andrew Steptoe, John Deanfield, for the IPD-Work consortium
Summary
Background Although some cardiovascular disease prevention guidelines suggest a need to manage work stress in
patients with established cardiometabolic disease, the evidence base for this recommendation is weak. We sought to
clarify the status of stress as a risk factor in cardiometabolic disease by investigating the associations between work
stress and mortality in men and women with and without pre-existing cardiometabolic disease.
Methods In this multicohort study, we used data from seven cohort studies in the IPD-Work consortium, initiated
between 1985 and 2002 in Finland, France, Sweden, and the UK, to examine the association between work stress and
mortality. Work stress was denoted as job strain or eort–reward imbalance at work. We extracted individual-level
data on prevalent cardiometabolic diseases (coronary heart disease, stroke, or diabetes [without dierentiation by
diabetes type]) at baseline. Work stressors, socioeconomic status, and conventional and lifestyle risk factors (systolic
and diastolic blood pressure, total cholesterol, smoking status, BMI, physical activity, and alcohol consumption) were
also assessed at baseline. Mortality data, including date and cause of death, were obtained from national death
registries. We used Cox proportional hazards regression to study the associations of work stressors with mortality in
men and women with and without cardiometabolic disease.
Results We identified 102 633 individuals with 1 423 753 person-years at risk (mean follow-up 13·9 years [SD 3·9]), of
whom 3441 had prevalent cardiometabolic disease at baseline and 3841 died during follow-up. In men with
cardiometabolic disease, age-standardised mortality rates were substantially higher in people with job strain
(149·8 per 10 000 person-years) than in those without (97·7 per 10 000 person-years; mortality dierence 52·1 per
10 000 person-years; multivariable-adjusted hazard ratio [HR] 1·68, 95% CI 1·19–2·35). This mortality dierence for
job strain was almost as great as that for current smoking versus former smoking (78·1 per 10 000 person-years) and
greater than those due to hypertension, high total cholesterol concentration, obesity, physical inactivity, and high
alcohol consumption relative to the corresponding lower risk groups (mortality dierence 5·9–44·0 per
10 000 person-years). Excess mortality associated with job strain was also noted in men with cardiometabolic disease
who had achieved treatment targets, including groups with a healthy lifestyle (HR 2·01, 95% CI 1·18–3·43) and
those with normal blood pressure and no dyslipidaemia (6·17, 1·74–21·9). In all women and in men without
cardiometabolic disease, relative risk estimates for the work stress–mortality association were not significant,
apart from eort–reward imbalance in men without cardiometabolic disease (mortality dierence 6·6 per
10 000 person-years; multivariable-adjusted HR 1·22, 1·06–1·41).
Interpretation In men with cardiometabolic disease, the contribution of job strain to risk of death was clinically significant
and independent of conventional risk factors and their treatment, and measured lifestyle factors. Standard care targeting
conventional risk factors is therefore unlikely to mitigate the mortality risk associated with job strain in this population.
Funding NordForsk, UK Medical Research Council, and Academy of Finland.
Copyright © 2018 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license.
Introduction
Meta-analyses of prospective cohort studies have shown
that psychosocial stress might increase the risk of
cardiovascular disease and diabetes.1–4 The underlying
pathophysiological mechanisms include disturbed sym-
pathetic-parasympathetic balance and dysregulation of
the hypothalamic–pituitary–adrenal axis, which can
accelerate the development of metabolic syndrome and
lead to left-ventricular dysfunction, dysrhythmia, and
proinflammatory and procoagulant responses.5,6 Stress
has also been linked to worsening health-related lifestyle
factors, such as physical inactivity and increased alcohol
consumption, and, in people with existing illness,
suboptimal treatment adherence.6
Although prevention guidelines for cardiovascular
disease do not prioritise the management of stress in
the general population,7–9 some guidelines recommend
stress management for individuals with established

Articles
2
www.thelancet.com/diabetes-endocrinology Published online June 5, 2018 http://dx.doi.org/10.1016/S2213-8587(18)30140-2
cardio vascular disease or major cardiovascular risk
factors, such as diabetes.7 The rationale for these
recommendations is that people with cardio metabolic
disease have many more adverse health events than do
the general population—therefore, assuming that the
relative risk associated with stress is the same for all
people exposed, a greater number of adverse events will
be prevented by targeting those already at high risk.
However, the evidence base for this recommendation is
weak, relying on studies of disease incidence1–4,10,11 with
very few large-scale studies of mortality11–15 and stress in
patients with cardiometabolic disease.13–17 Importantly, it
is unknown whether the excess risk associated with
stress at work and private life can be mitigated by
controlling conventional risk factors (eg, blood pressure
and cholesterol concentration) and improving lifestyle
(eg, physical activity and weight control).
The Individual-Participant-Data Meta-analysis in
Working Populations (IPD-Work) consortium is the
largest multicohort research collaboration on work stress
and clinically verified cardiovascular disease and
diabetes.1,3,10 In this study, we sought to clarify the status
of stress as a risk factor in cardiometabolic disease by
investigating the associations of two common work
stressors, job strain and eort–reward imbalance, with
mortality in individuals with pre-existing diabetes or
coronary heart disease or a history of stroke. For
comparison, we examined the stress–mortality
association in individuals without these diseases. To
investigate whether management of conventional and
lifestyle risk factors is likely to eliminate any excess risk
associated with work stress, we also assessed the stress–
mortality relation among patients with cardiometabolic
disease who otherwise had low risk factor levels (ie, were
normotensive, non-obese, physically active, had normal
blood cholesterol concentrations, and were not smokers
or heavy drinkers). If stress was associated with excess
mortality, even in subgroups of low-risk patients, then
better stress management might be an improvement on
standard care.
Methods
Study population
Established in 2008, the objective of the IPD-Work
Consortium1,10 is to provide a large-scale harmonised
database for the longitudinal estimation of associations
between predefined psychosocial working conditions
and chronic disease outcomes. The participating studies
comply with the Declaration of Helsinki and were
approved by local ethics review boards. Informed consent
was obtained from all participants.
Of the 12 original studies in the IPD-Work Consortium,1
seven independent cohort studies, initiated between
1985 and 2002 in Finland (FPS, HeSSup, Still Working),
France (GAZEL), Sweden (WOLF S, WOLF N) and the
UK (Whitehall II), had data relevant to the present
research. From each cohort study, eligible participants
were those who were employed at the time of the
baseline assessment, had data for age, sex, job strain,
eort–reward imbalance at work, and prevalent
cardiovascular disease and diabetes, and were being
followed up for mortality. Data were anonymised and
Germany (Prof N Dragano PhD,
Prof J Siegrist PhD); School of
Health and Welfare, Jönköping
University, Jönköping, Sweden
(E I Fransson); Stress Research
Institute, Stockholm
University, Stockholm, Sweden
(E I Fransson, Prof T Theorell MD,
Prof H Westerlund PhD); Inserm
UMS 011, Population-Based
Epidemiological Cohorts Unit,
Villejuif, France
(Prof M Goldberg MD,
M Zins PhD); Versailles
St-Quentin University, UMS
011, Villejuif, France
(Prof M Goldberg, M Zins);
Department of Health Sciences,
Mid Sweden University,
Sundsvall, Sweden
(Prof A Knutsson PhD); Finnish
Institute of Occupational
Health, Helsinki, Finland
(A Koskinen MSc, T Oksanen MD,
A Väänänen PhD); Division of
Health Psychology, SWPS
University of Social Sciences
and Humanities in Wroclaw,
Wroclaw, Poland
(Prof A Kouvonen);
Administrative Data Research
Centre Northern Ireland, Centre
for Public Health, Queen’s
University Belfast, Belfast, UK
(Prof A Kouvonen); National
Research Centre for the
Working Environment,
Copenhagen, Denmark
(Prof R Rugulies PhD);
Department of Public Health
and Department of Psychology,
University of Copenhagen,
Copenhagen, Denmark
(Prof R Rugulies); Inserm UMR
1018, Centre for Research in
Epidemiology and Population
Health, Villejuif, France
(A Singh-Manoux); Folkhälsan
Research Center, Helsinki,
Finland (S Suominen); School of
Health and Education,
University of Skövde, Skövde,
Sweden (S Suominen); School
of Social Policy, Sociology and
Social Research,
University of Kent, Canterbury,
UK (S Suominen); Turku
University Hospital, Turku,
Finland (Prof J Vahtera);
Department of Medical
Sciences, Uppsala University,
Uppsala, Sweden
(Prof P J M Westerholm MD);
Department of Internal
Medicine, Helsinki University
Hospital, Helsinki, Finland
(Prof T Strandberg); and Center
for Life Course Health Research,
University of Oulu, Oulu,
Finland (Prof T Strandberg)
Research in context
Evidence before this study
Work stressors, such as job strain and effort–reward imbalance
at work, are common sources of stress in adulthood. Work
stressors have been examined as risk factors for cardiometabolic
disease, such as coronary heart disease, stroke, and diabetes,
but few studies are available on their role as prognostic factors
for these diseases. We searched PubMed and Embase databases
from inception up to Feb 1, 2018 using the search terms:
“work stress”, “job stress”, “job strain”, “effort–reward
imbalance”, and “mortality”, without language restrictions.
We identified no large-scale studies comparing the association
between work stressors and mortality in people with and
without cardiometabolic disease.
Added value of this study
We pooled individual-participant data from seven European
cohort studies, including a total of 102 633 men and women.
Job strain was associated with substantial relative and absolute
increases in mortality risk in men with cardiometabolic disease.
The mortality difference between groups with and without job
strain was clinically significant and independent of
socioeconomic status and several conventional and lifestyle risk
factors, including hypertension and dyslipidaemia and their
pharmacological treatments, obesity, smoking, physical
inactivity, and high alcohol consumption. In absolute terms, the
difference in age-standardised mortality was greater for current
smoking versus not smoking than for with versus without job
strain, but, job strain was associated with a greater mortality
difference than were high cholesterol, obesity, high alcohol
consumption, and physical inactivity. In women and
participants without cardiometabolic disease, the work
stress–mortality associations were small or absent, both in
relative and absolute terms.
Implications of all the available evidence
The finding that job strain increases mortality risk, even in
subgroups of men with cardiometabolic disease but a favourable
cardiometabolic risk profile, suggests that standard care
targeting conventional risk factors is unlikely to mitigate the
mortality risk associated with job strain. Subsequent research
should employ intervention designs to establish whether
systematic screening and management of work stressors such as
job strain would contribute to improved health outcomes in
men with coronary heart disease, stroke, or diabetes.

Articles
www.thelancet.com/diabetes-endocrinology Published online June 5, 2018 http://dx.doi.org/10.1016/S2213-8587(18)30140-2
3
available at the individual level. Details of the studies
included in the present multicohort analysis are
summarised in the appendix.
Clinical characteristics
Baseline characteristics recorded were age, sex, and
harmonised measures of smoking (never smoker,
ex-smoker, or current smoker), alcohol consumption
(non-drinkers, moderate drinkers [1–14 drinks per week
for women and 1–21 drinks per week for men], and heavy
drinkers [>14 drinks per week for women and >21 drinks
per week for men]), leisure-time physical activity
(none or very little, moderate, or vigorous physical
activity or exercise), BMI (<18·5 kg/m² [underweight],
18·5–24·9 kg/m² [normal weight], 25–29·9 kg/m²
[overweight], or ≥30 kg/m² [obese]), and socioeconomic
status (high, intermediate, or low, defined on the basis of
an occupational title or, in the HeSSup study, a
participant’s highest educational qualification).1 In three
studies (Whitehall II, WOLF S, and WOLF N), assess-
ments of systolic and diastolic blood pressure and total
cholesterol concentration were also available.1 In
four studies (FPS, HeSSup, WOLF S, and WOLF N), it
was pos sible to assess prescriptions in participants with
diabetes, coronary heart disease, or stroke by linkage to
pre scription registers during the baseline year. Pre scrip-
tions for antidiabetes (Anatomical Therapeutic Chemical
[ATC] code A10), antihypertensive (ATC C02, C03,
C07–C09), lipid-lowering (ATC C10AA), and anticoag-
ulation (ATC B01) medications were considered to
indicate adherence.
Work stress
Analyses were based on two indicators of work stress:
job demand–control (ie, job strain) and eort–reward
imbalance at work. Reports from the IPD-Work
consortium are based on predefined, harmonised, and
validated definitions of work stress. The psychometric
properties of these data were published before the
extraction of outcome data.18,19 Job strain, referring to a
combination of high demands and low control at work,
was measured with sets of questions from the validated
Job Content Questionnaire and Demand-Control
Questionnaire, which were included in the baseline
self-report questionnaire of all of seven studies.18 Using
both questionnaires, we defined high job demands as
having a job-demand score that was greater than the
study-specific median score; similarly, we defined low
job control as having a job control score that was lower
than the study-specific median score. The Pearson
correlations between the harmonised scales used in this
study and complete versions of the Job Content
Questionnaire and Demand Control Questionnaire all
had r greater than 0·9, apart from one study in which
r was 0·8.18 In the present analyses, the exposure was
defined as job strain versus no job strain according to
the job strain model.1
The Eort–Reward Imbalance at Work questionnaire at
baseline included items on work demands and eorts (the
eort items) and monetary and non-monetary rewards at
work (the reward items). Dierent questionnaire versions
were harmonised and validated across the constituent
studies before the mortality analyses.19 Pearson correlation
coecients between the harmonised scales used in this
study and complete versions of the Eort–Reward
Imbalance questionnaire were high: r was greater than
0·9 for the eort scales and greater than 0·8 for the
reward scales.19 For each participant, mean response
scores were calculated separately for the eort and reward
items. We constructed a ratio of the two scores to quantify
the degree of mismatch between eort and rewards. The
eort–reward ratio was dichotomised at a cuto point of 1
with a ratio greater than 1 indicating eort–reward
imbalance and a ratio of 1 or lower indicating no eort–
reward imbalance at work.19
To examine the combined eects of job strain and
eort–reward imbalance, we constructed a three-level
exposure variable, where 0 represented no job strain or
eort–reward imbalance, 1 represented either job strain or
eort–reward imbalance (not both), and 2 represented both
job strain and eort–reward imbalance.2 More details of the
work stress measurements are provided in the appendix.
Baseline cardiometabolic disease
Baseline (existing) cardiometabolic diseases included
common causes of death: coronary heart disease, stroke,
and diabetes (without distinguishing between types of
diabetes). Coronary heart disease was ascertained from
national hospital admission records and discharge
registries (participants were linked to these registers with
individual identification numbers) and denoted with
version 10 of the International Classification of Diseases
(ICD; codes I21–I22 or the corresponding ICD-9 or
ICD-8 codes),1 or clinical examination with the MONICA
definition (Whitehall II).20 Agreement between the
national hospital admission records and clinical exami-
nations for coronary heart disease has been shown to be
high (sensitivity 70% and specificity >95%, with clinical
examination used as the gold standard ascer tainment
method).21 We identified history of stroke using self-
reported doctor-diagnosed events, event tracing, and
linkage to national hospital admission records (ICD-10
codes I60, I61, I63, I64).10 Prevalent diabetes was defined
with information from any of the following data sources:
hospital admission records with ICD-10 diagnoses
(E10, E11; all studies apart from Whitehall II), antidiabetes
drug reimbursements (only FPS, Still Working, and
HeSSup),3 or 2 h oral glucose tolerance test (WHO criteria)
complemented by self-report of diabetes diagnosis and
medication (Whitehall II).22
Mortality follow-up
Mortality data, including date and cause of death, were
obtained from national death registries. In each study,
Correspondence to:
Prof Mika Kivimäki, Department
of Epidemiology and Public
Health, University College
London, London WC1E 6BT, UK
m.kivimaki@ucl.ac.uk
See Online for appendix

Articles
4
www.thelancet.com/diabetes-endocrinology Published online June 5, 2018 http://dx.doi.org/10.1016/S2213-8587(18)30140-2
participants were linked to mortality records using their
unique identification numbers. Because these records do
not include date of death for people who emigrate and
die abroad, such participants were flagged as emigrants
and censored at the date of emigration.
Statistical analysis
Means and SDs were calculated according to cardio-
metabolic disease status at baseline (prevalent coronary
heart disease, stroke, or diabetes vs none of these
diseases). Each participant was followed up from the date
of the assessment of their work stressors and prevalent
cardiometabolic disease to the earliest event out of death,
loss to follow-up, or end of follow-up (maximum
20 years). We computed time to death to obtain
age-adjusted incidence rates per 10 000 person-years in
men and women in the pooled dataset. We used Cox
proportional hazards regression to study the associations
Figure 1: Sample selection
86 696 included in multivariable-adjusted
lifestyle analysis
17 415 included in multivariable-adjusted
biochemistry analysis
102 663 included in minimally adjusted
analysis
15 967 had missing lifestyle data
85 248 had missing biochemistry
data
102 663 with data for work stress,
prevalent cardiometabolic disease,
and mortality
105 284 participants had IPD-Work data
available
75 242 included in analysis of work
stressors
2621 excluded
1121 had missing data for work stress
1490 had missing data for baseline disease
10 had missing data for mortality
27 421 had missing stressor data
Participants without prevalent cardiometabolic disease (n=99 222) Participants with prevalent cardiometabolic disease (total n=3441)
Deaths/
participants
Minimally adjusted* Multivariable adjusted† Deaths/
participants
Minimally adjusted* Multivariable adjusted†
HR (95% CI) p value HR (95% CI) p value pcorrected‡HR (95% CI) p value HR (95% CI) p value pcorrected‡
Men
Job strain
No 2049/37 287 1 (ref) ·· 1 (ref) ·· ·· 256/1734 1 (ref) ·· 1 (ref) ·· ··
Yes 296/5246 1·06 (0·94–1·20) 0·35 1·01 (0·86–1·19) 0·92 1·00 51/241 1·66 (1·23–2·25) 0·001 1·68 (1·19–2·35) 0·003 0·024
Effort–reward imbalance
No 755/19 675 1 (ref) ·· 1 (ref) ·· ·· 133/1027 1 (ref) ·· 1 (ref) ·· ··
Yes 340/7911 1·21 (1·05–1·39) 0·009 1·22 (1·06–1·41) 0·006 0·048 61/498 0·70 (0·51–0·97) 0·03 0·70 (0·50–0·98) 0·04 0·32
Women
Job strain
No 972/46 242 1 (ref) ·· 1 (ref) ·· ·· 77/1147 1 (ref) ·· 1 (ref) ·· ··
Yes 247/10 447 1·05 (0·91–1·20) 0·52 0·96 (0·82–1·12) 0·59 1·00 26/319 1·21 (0·78–1·90) 0·40 1·11 (0·68–1·84) 0·67 1·00
Effort–reward imbalance
No 559/28 280 1 (ref) ·· 1 (ref) ·· ·· 46/703 1 (ref) ·· 1 (ref) ·· ··
Yes 284/16 611 0·93 (0·80–1·07) 0·32 0·91 (0·78–1·06) 0·23 1·00 32/537 1·01 (0·64–1·60) 0·96 0·99 (0·62–1·60) 0·98 1·00
HR=hazard ratio. *Minimal adjustment includes age and study. †Multivariable adjustment includes study, age, smoking status, physical activity, alcohol consumption, BMI, and socioeconomic status. ‡p value
corrected for multiple testing (Bonferroni correction).
Table 1: Association between work stressors and total mortality in men and women, by baseline cardiometabolic disease

Articles
www.thelancet.com/diabetes-endocrinology Published online June 5, 2018 http://dx.doi.org/10.1016/S2213-8587(18)30140-2
5
of work stressors (job strain and eort–reward imbalance
at work) with mortality in men and women with and
without cardiometabolic disease. Bonferroni correction
was used to compensate for multiple testing (a total of
eight tests from two stressors, two cardiometabolic
disease statuses, and two sexes). Minimally adjusted
models included age and study as covariates. Multi-
variable models were also adjusted for socioeconomic
status, BMI category, smoking status, alcohol con-
sumption, physical activity and, in a subgroup analysis
of cohort studies with relevant data, blood pressure and
total cholesterol concentration. Interactions with age
were tested by grouping participants by age (<45 years,
45–54 years, and >55 years). Heterogeneity in study-
specific estimates was examined by repeating the main
analyses with a two-step procedure, with separate
analyses in each cohort study and then pooling of the
study-specific hazard ratios by use of random-eects
meta-analysis.
Robustness of the association between work stressors
and mortality in subgroups was tested in analyses
stratified by number of lifestyle risk factors (zero, one, or
two or more from current smoking, physical inactivity,
obesity, and high alcohol consumption). To examine
whether any eect of job strain is present in individuals
with cardiometabolic disease but an otherwise low-risk
profile, we assessed the association between work
stressors and mortality in subgroups of participants who
had met treatment targets—ie, they had adhered to
pharmacotherapy, had normal blood pressure (systolic
and diastolic blood pressure <140/90 mm Hg), and normal
fasting total cholesterol concentration (<6·2 mmol/L
and, in sensitivity analysis, <5·0 mmol/L). To minimise
residual confounding, systolic blood pressure and total
cholesterol concentration, treated as continuous variables,
were added to the model as covariates.
In further analyses, we examined the association of
exposure to neither, either, or both of the work stressors
with mortality and the associations of obesity, current
smoking, high alcohol consumption, and physical
inactivity with mortality.
All analyses were done with SAS statistical software
version 9.4. Statistical significance was inferred at a
two-sided p value less than 0·05.
Role of the funding source
The funders of the study had no role in study design,
data collection, data analysis, data interpretation, or
writing of the report. MKi and JP had full access to all the
data in the study and MKi and JD had final responsibility
for the decision to submit for publication.
Results
105 284 people were recruited into the seven studies
between 1985 and 2002. Of the eligible population,
102 663 participants had data on prevalent cardiometabol-
ic disease, at least one of the work stressors (job strain
or eort–reward imbalance), and mortality, and were
therefore included in this study (figure 1). Characterist-
ics were similar between the eligible and included
populations in terms of the proportion of men (43·5% in
eligible vs 43·4% in enrolled participants), mean age
(44·0 years vs 43·9 years), and proportion of participants
of low-socioeconomic status (26·2% vs 25·8%).
Mean follow-up for mortality was 13·9 years (SD 3·9).
During 1 423 753 person-years at risk, we identified
3841 deaths, of which 397 were among the 3441 individuals
with cardiometabolic disease at baseline. Of the 1975 men
with cardiometabolic disease at baseline, 396 had a
history of coronary heart disease, 214 had stroke, 1425 had
diabetes, 54 had two of these disorders, and three had all
three. Of the 1466 women with cardiometabolic
disease at baseline, 73 had a history of coronary heart
disease, 153 had stroke, 1266 had diabetes, 18 had
two of these disorders, and four had all three (appendix).
In men without cardiometabolic disease at baseline,
eort–reward imbalance was associated with an in-
creased risk of mortality (mortality dierence 6·6 per
10 000 person-years), and this association remained after
multivariable adjustment and correction for multiple
testing (table 1). There was no significant heterogeneity
Figure 2: Job strain and age-adjusted mortality
Job strain and mortality in participants without (A) and with cardiometabolic disease (B) at baseline, and
cumulative hazard in participants with cardiometabolic disease at baseline (C).
WomenMen
0
Mortality per 10 000 person-years
Sex Sex
160
140
120
100
80
60
40
20
AParticipants without cardiometabolic disease
WomenMen
0
160
140
120
100
80
60
40
20
BParticipants with cardiometabolic disease
No job strain
Job strain
Number at risk
Men without job strain
Men with job strain
Women without job strain
Women with job strain
02015105
1734
241
1147
319
1705
233
1138
314
1555
211
1009
281
1076
142
370
113
289
51
181
42
Years
0
35
30
25
20
15
5
10
Cumulative hazard (%)
C
Men without job strain
Men with job strain
Women without job strain
Women with job strain

Articles
6
www.thelancet.com/diabetes-endocrinology Published online June 5, 2018 http://dx.doi.org/10.1016/S2213-8587(18)30140-2
in the study-specific estimates (I²=0%, p=0·44; appendix)
or interaction with age (p=0·10). In absolute terms, the
dierence in mortality between those with and without
eort–reward imbalance was smaller than those related
to conventional lifestyle factors, such as smoking,
physical inactivity, and obesity (appendix). In women
without cardiometabolic disease at baseline, eort–
reward imbalance was not associated with mortality
(table 1). Job strain alone (table 1 and figure 2A), or in
combination with eort–reward imbalance (appendix),
was not associated with mortality in men or women
without cardiometabolic disease.
In men with cardiometabolic disease at baseline, job
strain was associated with an increased risk of mortality
that remained after multivariable adjustment and cor-
rection for multiple testing (table 1). The age-adjusted
mortality rate per 10 000 person-years was 149·8 in men
with job strain and 97·7 in those without (risk dierence
52·1 per 10 000; figure 2B). The excess mortality risk in
men with cardiometabolic disease who reported job
strain was apparent across the entire follow-up period,
rather than becoming apparent only in the early or late
phases of the follow-up (figure 2C), and was robust to
adjustment for lifestyle risk factors (smoking status,
alcohol consumption, physical activity, BMI, and
socioeconomic status; table 1). There was no significant
heterogeneity in study-specific estimates (I²=0%, p=0·96;
appendix) or dierence in the association between age
groups (pinteraction=0·62). After further adjustment for
blood pressure and total cholesterol in the subgroup of
participants with these data available, the HR for job
strain compared with no job strain was 1·84
(95% CI 1·06–3·18; p=0·029; 94 deaths among
569 participants). In analyses of cause-specific mortality,
job strain had a minimally adjusted (study and age) HR
of 1·71 (95% CI 1·08–2·71; p=0·02) for risk of mortality
from cardiovascular disease (appendix), but no robust
associations were observed with cancer mortality or
non-cardiovascular, non-cancer mortality. Eort–reward
imbalance seemed to be associated with lower risk of
death in men with previous cardiometabolic disease, but
this association was lost after correction for multiple
testing (table 1).
In women with cardiometabolic disease at baseline, the
age-adjusted death rates per 10 000 were 64·0 for job
strain and 53·2 for no job strain (dierence 10·8 per
10 000; figure 2) and mortality was not significantly
associated with job strain or eort–reward imbalance
(table 1). Job strain in combination with eort–reward
imbalance was not associated with mortality in men or
women with cardiometabolic disease (appendix).
To examine the relative importance of job strain as a risk
factor for mortality in men with cardiometabolic disease,
we compared death rates associated with job strain with
those associated with established risk factors. The mortality
dierence between men with and without job strain
(52·1 per 10 000) was almost the same as that for current
smokers versus never or former smokers (78·1 per 10 000),
and higher than those for the presence of hypertension,
high total cholesterol, obesity, physical inactivity, and high
alcohol consumption (5·9–44·0 per 10 000; figure 3).
Furthermore, job strain was associated with a two to six
times higher risk of mortality in subgroups of men with
cardio metabolic disease but favourable risk factor profiles,
including participants who were not obese, physically
inactive, smokers, or heavy drinkers (table 2) and
normotensive participants, those with no dyslipidaemia,
and those who adhered to antihyper tensive, lipid-lowering,
or anticoagulation treatments, according to prescription
records (figure 4). Additional adjustments did not alter
these findings (appendix).
Discussion
Evidence from our pooling of individual-participant data
from seven European cohort studies suggests that job
strain is a risk factor for mortality in men with
cardiometabolic disease, as defined by the presence of
coronary heart disease, stroke, or diabetes. The mortality
dierence between groups with and without job strain
Figure 3: Mortality in men with cardiometabolic disease by job strain and lifestyle factors
*Data available only from Whitehall II, WOLF-N, and WOLF-S.
020015010050
Baseline risk factor
Age-adjusted mortality per 10 000 person-years
Smoking
p value
<0·0001
0·04
0·51
0·13
0·65
0·35
0·005
High cholesterol
High blood pressure*
Obesity
Physical inactivity
High alcohol consumption
Job strain
No Yes
Job strain Participants Deaths HR (95% CI) p value
0 lifestyle risk factors No 718 88 1 (ref) ··
0 lifestyle risk factors Yes 80 16 2·01 (1·18–3·43) 0·010
1 lifestyle risk factor No 626 94 1·39 (1·03–1·86) 0·029
1 lifestyle risk factor Yes 95 19 2·09 (1·27–3·45) 0·004
>2 lifestyle risk factors No 389 73 2·21 (1·60–3·06) <0·0001
>2 lifestyle risk factors Ye s 66 16 3·03 (1·76–5·20) <0·0001
Lifestyle risk factors are current smoking, obesity, physical inactivity, and high alcohol consumption. HRs are adjusted
for age and study. HR=hazard ratio.
Table 2: Association between job strain and mortality by number of lifestyle risk factors in men with
cardiometabolic disease at baseline

Articles
www.thelancet.com/diabetes-endocrinology Published online June 5, 2018 http://dx.doi.org/10.1016/S2213-8587(18)30140-2
7
was clinically significant and independent of socio-
economic status, the conventional and lifestyle risk factors
measured (current smoking, obesity, physical inactivity,
high alcohol consumption, hypertension, dyslipidaemia),
and pharmacotherapy. This finding is unlikely to be
attributable to type I error due to multiple testing because
it was robust to correction for multiple comparisons.
In absolute terms, the dierence in age-standardised
mortality among men with cardiometabolic diseases was
greater between current and non-smokers than between
men with and without job strain. However, high
cholesterol, obesity, high alcohol consumption, and
physical inactivity were associated with smaller mortality
dierences than job strain.
Our findings agree with those of a study of patients with
acute myocardial infarction from the USA, in which
individuals who reported life stress had higher mortality
than those free of life stress (HR 1·4, 95% CI 1·2–1·8)14
and the few previous, small-scale prognostic studies on
patients with cardiovascular disease,23–26 which in com-
bination suggest a 1·6 times (95% 1·2–2·2) increased
risk of recurrent events associated with job strain.17 The
observed associations are also biologically plausible. The
stress hormone cortisol stimulates glucose production
in the liver and antagonises the action of insulin in
peripheral tissues—both processes have the potential to
contribute to worse prognoses in people with diabetes.5,6
Stress can also have adverse eects on cardiometabolic
systems by inducing transient endothelial dysfunction,
myocardial ischaemia, and cardiac arrhythmia and thus
increasing the risk of fatal and non-fatal cardiac events.6
To our knowledge, this is the first large-scale study to
examine the work stress–mortality association stratified
by cardio metabolic risk profile. Our data showed that job
strain substantially increased mortality risk even in
subgroups of men with prevalent cardiometabolic
disease but a favourable cardiometabolic risk profile,
suggesting that standard care targeting conventional and
lifestyle risk factors (eg, blood pressure, lipids, smoking,
obesity, physical inactivity) does not necessarily mitigate
the excess mortality risk associated with job strain. The
European prevention guidelines7 and the American
Heart Association policy statements8 highlight psycho-
social stress as a potential barrier to healthy lifestyles and
optimal medication adherence, and recommend manage-
ment of stress in individuals with high cardiovascular
risk or established cardiovascular disease. Our findings
are consistent with these recommendations, but also
suggest that harmful eects of stress in men were not
attributable to the lifestyle risk factors measured or poor
adherence to pharmacotherapy; excess mortality risk
was observed even among patients successfully treated
for cardiometabolic disease who were normotensive,
non-obese, physically active, had normal blood cholesterol,
and were not smokers or heavy drinkers.
There are various ways of expanding standard care to
address work stress in patients, including systematic
screening for stress and, if needed, interventions such as
consultation, rehabilitation, job redesign, reductions in
working hours, and retirement on health grounds.6,7 In a
Cochrane review of 35 randomised controlled trials
including a total of 10 703 patients with coronary heart
disease who had at least 6 months’ follow-up, psycho-
logical interventions that alleviated stress and other
psychological symptoms were successful in reducing
cardiac mortality for people with coronary heart disease.27
However, it is unclear whether those inter ventions would
benefit men with job strain and cardiometabolic disease.
For other groups stress-related dierences in mortality
were small or absent, both in relative and absolute terms.
In working-aged women with cardiometabolic disease,
for example, job strain was not associated with a
Deaths
Healthy subgroup
No lifestyle risk factors*
No job strain
Job strain
Normotensive (systolic/diastolic blood pressure <140/90 mm Hg)†
No job strain
Job strain
No dyslipidaemia (total cholesterol <6·2 mmol/L)†‡
No job strain
Job strain
Normotensive and no dyslipidaemia†
No job strain
Job strain
High adherence to pharmacotherapy§¶
No job strain
Job strain
88
16
54
20
26
8
10
8
58
15
Participants
718
80
372
66
241
37
135
30
498
64
p value
0·010
0·0001
0·086
0·0049
0·0029
HR (95% CI)
1·00
2·01 (1·18–3·43)
1·00
3·77 (1·92–7·39)
1·00
2·26 (0·89–5·72)
1·00
6·17 (1·74–21·9)
1·00
2·38 (1·34–4·20)
0·5 8·02·0
Figure 4: Job strain and mortality in men with cardiometabolic disease and a favourable risk profile
HRs are adjusted for age and study. In analyses of normotensive and non-dyslipidaemic participants, HRs are also adjusted for systolic and diastolic blood pressure
and total cholesterol. HR=hazard ratio. *Analysis included all studies. †Analysis included Whitehall II, WOLF-N, and WOLF-S. ‡For a subgroup of participants with total
cholesterol <5·0 mmol/L, the corresponding hazard ratio is 4·82 (95% CI 1·15–20·2) for those with job strain (three deaths among nine participants) compared with
those without job strain (five deaths among 71 participants). §Antidiabetes (ATC A10), antihypertensive (ATC C02, C03, C07-C09), lipid-lowering (ATC C10AA),
and anticoagulation (ATC B01) medication. ¶Analysis includes FPS, HeSSup, WOLF-S, WOLF-N.

Articles
8
www.thelancet.com/diabetes-endocrinology Published online June 5, 2018 http://dx.doi.org/10.1016/S2213-8587(18)30140-2
significant increase in mortality risk and the absolute
mortality dierence between those with and without
job strain was only 10·8 per 10 000 person-years (for
comparison, the corresponding mortality dierence
in men was 52·1 per 10 000 person-years). Similarly,
eort–reward imbalance was not associated with
increased mortality in men or women with cardio-
metabolic disease, suggesting that job strain and eort–
reward imbalance are of dierent prognostic value. Job
strain encompasses only external sources of stress,
whereas eort–reward imbalance also involves the
individual’s own behaviours. People with more severe
cardiometabolic disease tend to shorten their working
hours as a consequence of their condition, thus potentially
reducing any eort–reward imbalance through reduced
eort.10,28 This change could mitigate the link between
eort–reward imbalance and mortality. By contrast,
external characteristics of work that relate to job strain
remain unchanged after the onset of disease. Finally, as
expected, in healthy people work stress did not
substantially increase mortality risk, although in men
free of cardiometabolic disease, we observed a moderate
association between eort–reward imbalance and risk of
death.
Our study benefits from a large sample size, predefined
exposure assessment, coverage of several European
countries, and a mortality outcome assessed via record
linkage with very little loss to follow-up. The limitations of
our study include the use of a single measurement of
work stressors and risk factors, which does not include
any measure of chronicity or change over time. There is
also the possibility that prevalent cardiometabolic disease
was underestimated in those studies with no measures
of undiagnosed diabetes and cardiovascular disease
(eg, silent myocardial infarctions). These drawbacks could
contribute to an underestimation or overestimation of
associations with mortality. We adjusted the associations
for several conventional and lifestyle risk factors, but data
for blood pressure and blood cholesterol concentration
were not available in all the studies. This limitation could
lead to overestimation of the status of job strain as an
independent predictor of mortality, although there was no
evidence to support this possibility in supplementary
analyses of the three cohort studies with relevant data. We
did not have detailed data on the duration or severity of
the cardiometabolic diseases. Several factors that are more
common in individuals with stress that can precipitate a
fatal cerebrovascular or cardiovascular event, or otherwise
increase risk of premature death, were not covered by
our baseline measurement. These include, for example,
stress-induced ischaemia, cardiac arrhyth mia, low-grade
systemic inflammation, increased blood viscosity, platelet
activation and increases in the levels of coagulation
and fibrinolytic factors, short and long sleep durations
and sleep disorders, and reduced self-care.6,29,30 Further
research is needed to establish the role of such factors in
the excess mortality risk seen in men with job strain and
cardiometabolic disease and to examine mechanisms
underlying the observed sex dierences in the eects of
job strain.
In conclusion, the results of this large pan-European
study suggest that in men with cardiometabolic disease,
the contribution of job strain to risk of death is clinically
significant and independent of conventional risk factors
and their treatment, as well as the lifestyle factors
measured. Subsequent research should employ inter-
vention designs to establish whether systematic screening
and management of work stressors, such as job strain,
would contribute to improved health outcomes in men
with prevalent coronary heart disease, stroke, or diabetes.
Contributors
All authors participated in designing the study, generating hypotheses,
interpreting the data, and critically reviewing the report. MKi wrote the
first draft of the report. JEF and JD were also members of the writing
group. JP analysed the data with support from STN. JP and MKi had
full access to anonymised individual participant data from all
constituent studies.
Declaration of interests
We declare no competing interests.
Acknowledgments
The IPD-Work consortium was supported by NordForsk (the Nordic
Research Programme on Health and Welfare), the UK Medical Research
Council (K013351, R024227), and the Academy of Finland (311492).
References
1 Kivimäki M, Nyberg ST, Batty GD, et al. Job strain as a risk factor
for coronary heart disease: a collaborative meta-analysis of
individual participant data. Lancet 2012; 380: 1491–97.
2 Dragano N, Siegrist J, Nyberg ST, et al. Eort–reward imbalance at
work and incident coronary heart disease: a multi-cohort study of
90,164 individuals. Epidemiology 2017; 28: 619–26.
3 Nyberg ST, Fransson EI, Heikkilä K, et al. Job strain as a risk factor
for type 2 diabetes: a pooled analysis of 124 808 men and women.
Diabetes Care 2014; 37: 2268–75.
4 Huang Y, Xu S, Hua J, et al. Association between job strain and risk
of incident stroke: a meta-analysis. Neurology 2015; 85: 1648–54.
5 Brotman DJ, Golden SH, Wittstein IS. The cardiovascular toll of
stress. Lancet 2007; 370: 1089–100.
6 Kivimäki M, Steptoe A. Eects of stress on the development and
progression of cardiovascular disease. Nat Rev Cardiol 2018;
15: 215–29.
7 Piepoli MF, Hoes AW, Agewall S, et al. 2016 European Guidelines
on cardiovascular disease prevention in clinical practice:
the Sixth Joint Task Force of the European Society of Cardiology and
Other Societies on Cardiovascular Disease Prevention in Clinical
Practice (constituted by representatives of 10 societies and by
invited experts). Developed with the special contribution of the
European Association for Cardiovascular Prevention &
Rehabilitation (EACPR). Eur Heart J 2016; 37: 2315–81.
8 Pearson TA, Blair SN, Daniels SR, et al. AHA guidelines for
primary prevention of cardiovascular disease and stroke:
2002 update: consensus panel guide to comprehensive risk
reduction for adult patients without coronary or other
atherosclerotic vascular diseases. Circulation 2002; 106: 388–91.
9 Anderson TJ, Gregoire J, Pearson GJ, et al. 2016 Canadian
Cardiovascular Society guidelines for the management of
dyslipidemia for the prevention of cardiovascular disease in the
adult. Can J Cardiol 2016; 32: 1263–82.
10 Kivimäki M, Jokela M, Nyberg ST, et al. Long working hours and
risk of coronary heart disease and stroke: a systematic review and
meta-analysis of published and unpublished data for
603 838 individuals. Lancet 2015; 386: 1739–46.
11 Eaker ED, Sullivan LM, Kelly-Hayes M, D’Agostino RB Sr,
Benjamin EJ. Marital status, marital strain, and risk of coronary
heart disease or total mortality: the Framingham Ospring Study.
Psychosom Med 2007; 69: 509–13.

Articles
www.thelancet.com/diabetes-endocrinology Published online June 5, 2018 http://dx.doi.org/10.1016/S2213-8587(18)30140-2
9
12 Sheps DS, McMahon RP, Becker L, et al. Mental stress-induced
ischemia and all-cause mortality in patients with coronary artery
disease: results from the Psychophysiological Investigations of
Myocardial Ischemia study. Circulation 2002; 105: 1780–84.
13 Sabbath EL, Mejía-Guevara I, Noelke C, Berkman LF. The long-term
mortality impact of combined job strain and family circumstances:
a life course analysis of working American mothers. Soc Sci Med 2015;
146: 111–19.
14 Stewart RAH, Colquhoun DM, Marschner SL, et al.
Persistent psychological distress and mortality in patients with stable
coronary artery disease. Heart 2017; 103: 1860–66.
15 Arnold SV, Smolderen KG, Buchanan DM, Li Y, Spertus JA.
Perceived stress in myocardial infarction: long-term mortality and
health status outcomes. J Am Coll Cardiol 2012; 60: 1756–63.
16 Orth-Gomer K, Wamala SP, Horsten M, Schenck-Gustafsson K,
Schneiderman N, Mittleman MA. Marital stress worsens prognosis
in women with coronary heart disease: the Stockholm Female
Coronary Risk Study. JAMA 2000; 284: 3008–14.
17 Li J, Zjang M, Loerbroks A, Angerer P, Siegrist J. Work stress and the
risk of recurrent coronary heart disease events: a systematic review
and meta-analysis. Int J Occup Med Environ Health 2015; 28: 8–19.
18 Fransson EI, Nyberg ST, Heikkila K, et al. Comparison of alternative
versions of the job demand-control scales in 17 European cohort
studies: the IPD-Work consortium. BMC Public Health 2012; 12: 62.
19 Siegrist J, Dragano N, Nyberg ST, et al. Validating abbreviated
measures of eort–reward imbalance at work in European cohort
studies: the IPD-Work consortium. Int Arch Occup Environ Health
2014; 87: 249–56.
20 Tunstall-Pedoe H, Kuulasmaa K, Amouyel P, Arveiler D,
Rajakangas AM, Pajak A. Myocardial infarction and coronary
deaths in the World Health Organization MONICA Project.
Registration procedures, event rates, and case-fatality rates in
38 populations from 21 countries in four continents.
Circulation 1994; 90: 583–612.
21 Kivimäki M, Batty GD, Singh-Manoux A, Britton A, Brunner EJ,
Shipley MJ. Validity of cardiovascular disease event ascertainment
using linkage to UK hospital records. Epidemiology 2017; 28: 735–39.
22 Tabák AG, Jokela M, Akbaraly TN, Brunner EJ, Kivimäki M,
Witte DR. Trajectories of glycaemia, insulin sensitivity, and insulin
secretion before diagnosis of type 2 diabetes: an analysis from the
Whitehall II study. Lancet 2009; 373: 2215–21.
23 Aboa-Eboule C, Brisson C, Maunsell E, et al. Job strain and risk of
acute recurrent coronary heart disease events. JAMA 2007;
298: 1652–60.
24 Biering K, Andersen JH, Lund T, Hjollund NH. Psychosocial
working environment and risk of adverse cardiac events in patients
treated for coronary heart disease. J Occup Rehabil 2015; 25: 770–75.
25 Aboa-Eboule C, Brisson C, Maunsell E, et al. Eort–reward
imbalance at work and recurrent coronary heart disease events:
a 4-year prospective study of post-myocardial infarction patients.
Psychosom Med 2011; 73: 436–47.
26 Laszlo KD, Ahnve S, Hallqvist J, Ahlbom A, Janszky I. Job strain
predicts recurrent events after a first acute myocardial infarction:
the Stockholm Heart Epidemiology Program. J Intern Med 2010;
267: 599–611.
27 Richards SH, Anderson L, Jenkinson CE, et al. Psychological
interventions for coronary heart disease: Cochrane systematic
review and meta-analysis. Eur J Prev Cardiol 2018; 25: 247–59.
28 Sokejima S, Kagamimori S. Working hours as a risk factor for acute
myocardial infarction in Japan: case-control study. BMJ 1998;
317: 775–80.
29 St-Onge M-P, Grandner MA, Brown D, et al. Sleep duration and
quality: impact on lifestyle behaviors and cardiometabolic health:
a scientific statement from the American Heart Association.
Circulation 2016; 134: e367–86.
30 Liu T-Z, Xu C, Rota M, et al. Sleep duration and risk of all-cause
mortality: a flexible, non-linear, meta-regression of 40 prospective
cohort studies. Sleep Med Rev 2017; 32: 28–36.