1Vol.:(0123456789)
Scientific Reports |        (2023) 13:16526  | https://doi.org/10.1038/s41598-023-43666-x
www.nature.com/scientificreports
Midlife cardiovascular health 
factors as predictors of retirement 
age, work‑loss years, and years 
spent in retirement among older 
businessmen
Markus J. Haapanen 1,2,3*, Timo Törmäkangas 4, Monika E. von Bonsdorff 5, 
Arto Y. Strandberg 6, Timo E. Strandberg 6,7 & Mikaela B. von Bonsdorff 2,4
Cardiovascular disease (CVD) is one of the leading causes of premature retirement. However, the 
relationship between CVD risk factors and workforce participation is not well known. We studied the 
relationship between midlife CVD risk, age at retirement, work‑loss years, and survival in retirement. 
Middle‑aged Finnish men (initial n = 3490, mean age = 47.8 years) were assessed for CVD risk factors 
and general health in the 1970s. They worked as business executives and provided information on 
their retirement status in the year 2000. Survival was followed up to the 9th decade of life with a 
follow‑up of up to 44 years. Work‑loss years were calculated as death or retirement occurring at 
age ≤ 65 years. Smoking, body mass index, and alcohol use were used as covariates, excluding models 
of CVD risk, which were adjusted for alcohol use only. Higher risk of 10‑year fatal CVD was associated 
with 0.32 more years (relative risk < 1 vs. 1, covariate‑adjusted β = 0.32, 95% CI = 0.13, 0.53) of work‑
loss. Higher risk of 5‑year incident (covariate‑adjusted time‑constant HR = 1.32, 95% CI = 1.19, 1.47) 
and 10‑year fatal (covariate‑adjusted time‑dependent HR = 1.55, 95% CI = 1.30, 1.85) CVD in midlife 
were associated with fewer years spent in retirement. Poorer self‑rated health and physical fitness 
and higher levels of triglycerides were associated with increased hazard of earlier retirement, more 
work‑loss years, and fewer years spent in retirement. Poorer health and greater midlife CVD risk may 
be associated with earlier exit from the workforce and fewer years spent in retirement. Management 
of CVD risk in midlife may support people to work longer.
Cardiovascular disease (CVD), a leading cause of premature mortality  worldwide1, is associated with poorer 
workforce  participation2 among older employees by increasing the risk of transitioning into disability  pension3,4 
and exiting the labour force  early5. It bears high costs to societies, amounting to a quarter of health-related 
productivity losses in the European Union, with estimates of 54 billion in the year  20196. Whether CVD risk 
factors, including elevated blood pressure, serum lipids, and blood glucose among others, would also associate 
with workforce participation among individuals of working age, is poorly  understood7,8. Understanding this 
association is becoming increasingly important as more individuals are expected to work longer in the future, 
regardless of coexisting chronic illnesses.
A recent meta-analysis highlighted the need to consider prior health when studying the relationship between 
retirement timing and  mortality9. This can only be examined with follow-up studies started at earlier stages of 
the life course, with subjectively and objectively measured data on health at working age and in retirement, as 
well as data on mortality. There are few previous studies that have investigated the association between prior 
OPEN
1Department of General Practice and Primary Health Care, University of Helsinki and Helsinki University Hospital, 
Helsinki, Finland. 2Folkhälsan Research Centre, Helsinki, Finland. 3Department of Medical Epidemiology and 
Biostatistics, Karolinska Institutet, Stockholm, Sweden. 4Gerontology Research Center and Faculty of Sport and 
Health Sciences, University of Jyväskylä, Jyväskylä, Finland. 5Management and Leadership, Jyväskylä University 
School of Business and Economics, University of Jyväskylä, Jyväskylä, Finland. 6Department of Medicine, 
University of Helsinki and Helsinki University Hospital, Helsinki, Finland. 7Centre for Life Course Health Research, 
University of Oulu, Oulu, Finland. *email: markus.haapanen@helsinki.fi
2Vol:.(1234567890)
Scientific Reports |        (2023) 13:16526  | https://doi.org/10.1038/s41598-023-43666-x
www.nature.com/scientificreports/
health and survival in retirement over longer periods of time. Years spent in retirement is a way of describing 
this phenomenon and it can be calculated for cohorts with mortality follow-up until very old age.
Our aim was to increase the understanding of midlife CVD risk factors and general health as predictors 
for pre-retirement workforce participation and post-retirement survival; the former measured through age at 
retirement and calculation of work-loss years and the latter represented as survival in retirement. The rationale 
for including work-loss years was that it is uncertain whether previous findings are influenced by mortality 
before retirement. To study this, we assessed CVD risk factors and aspects of general health among Finnish male 
executives in midlife (mean age 47.8 years) and obtained information on their retirement status in the year 2000 
and by that time the youngest participants had reached the age of 66 years. Thereafter, assessment of register-
based all-cause mortality continued until the 9th decade in the year 2018. We hypothesized that greater CVD 
risk and poorer general health in midlife would associate with premature retirement and with poorer survival 
in retirement.
Methods
Study population
The Helsinki Businessmen Study is a longitudinal study of 3,490 initially healthy men who were born between 
1919 and  193410. The present sample (n = 3,309) consists of participants who could be traced using personal 
identification numbers assigned to all Finnish residents in the 1970s. Figure 1 presents a flowchart of the study 
population. The men (mean age 47.8 years in the year 1974) participated in voluntary health check-ups covering 
a range of cardiovascular health factors at the Finnish Institute of Occupational Health. This information was 
then used to calculate estimates of the participants’ 5-year risk of incident  CVD11 and 10-year risk of incident 
fatal  CVD12. Information on retirement was obtained when the youngest participants had reached the age of 
66 years in the year 2000; out of 2,286 traceable participants, 1,864 returned the questionnaire. Full retirement 
data were available for 1,766 participants, of whom 1,612 also had information on midlife cardiovascular health 
factors. The participants’ deaths were tracked from The Digital and Population data services Agency’s Population 
Information System between the years 1974 and 2018.
Cardiovascular health factors in 1974–1975
Midlife cardiovascular health factors, measured in the 1970s, included clinical and laboratory  data10, such as 
blood pressure (BP), heart rate, fasting total cholesterol (TC)13, triglyceride levels (TG), and 1 h blood glucose 
levels after an oral glucose tolerance test (1 h-OGTT load)14. Participants’ global self-rated health (SRH) and 
self-rated physical fitness (SRF) were assessed with the question, "What do you think about your present state of 
health/physical fitness; is it ‘very good’, ‘fairly good’, ‘average’, ‘fairly poor’ or ‘very poor’?" SRH categories ’fairly 
poor’ and ’very poor’ were combined into ’poor’ due to the small number of men who perceived their SRH as 
’very poor’.
We estimated the participants’ 5-year risk of incident  CVD11 using a tool, that was developed among Euro-
pean and U.S. men aged 40–59 years free from coronary heart disease. The risk  score11 uses information on age, 
systolic BP, serum TC, smoking status and body mass index (BMI). We also calculated the participants’ 10-year 
composite risk score of fatal CVD in midlife and old age using the European ’low risk’ SCORE-chart (Finnish)12. 
The chart gives estimates of the relative risk of fatal CVD events in a 10-year period. SCORE has been externally 
 validated15 and uses sex, smoking status, systolic BP, and serum TC to estimate CVD risk.
Covariates in the year 1974
In 1974, height and weight measurements were used to calculate BMI (kg/m2). Smoking status, based on self-
reported data, was categorized as current smokers, ex-smokers, and never smokers. Similarly, self-reported 
weekly alcohol consumption in grams of pure alcohol was recorded.
Retirement age and time spent in retirement evaluated in 2018
In 2000, participants reported their retirement status and age at  retirement16. We excluded participants report-
ing no retirement (n = 98) and analysed data from 1766 men. Time spent in retirement was calculated in years 
from the onset of retirement until death or the end of follow-up in the year 2018. Approximately two thirds 
(n = 1,999, 67.9%) of the sample had died by 2018, and we observed 43,086 person-years of mortality follow-up 
from the onset of retirement.
Working years lost due to death or early retirement
To study whether the inclusion of participants who died before retiring would affect our results, we calculated 
work-loss  years17 as death or retirement occurring before the general age limit for old-age pensions. We used 
65 years as the limit for old age  pensions17, which in this cohort was also the most common age to retire (n = 250, 
14.2%). The respective observations in person-years were 4,804 and 24,627, for deceased and those retiring 
younger than at age 65 years.
Statistical methods
We present continuous variables as means and standard deviations and used one-way analyses of variance and 
Kruskall-Wallis tests for group central tendency statistics. Categorical data are presented as proportions. Covar-
iate-adjusted Cox proportional hazard regression models and models extended for time-dependent effects were 
used to estimate hazard ratios (HRs) and 95% confidence intervals. For complex associations figures of HRs are 
shown. When the proportionality of hazards assumption is not met for covariate Z, the Cox- regression model 
3Vol.:(0123456789)
Scientific Reports |        (2023) 13:16526  | https://doi.org/10.1038/s41598-023-43666-x
www.nature.com/scientificreports/
can be extended to handle time-dependent effects by adding an effect of covariate interaction with a function 
of follow-up time. Thus, in our models the hazard function, hi(t), for participant i (i = 1, …, N) was specified as:
 where h0(t) is the baseline hazard function, zi is the covariate value with a time-invariant and time-dependent 
effects γ0 and γt, respectively, g(t) is a suitable function of time, t, and β is the vector of regression coefficients for 
other time-invariant fixed-effects for other covariates in vector xi. Notably, the effect of covariate Z is partitioned 
into two components: the time-invariant effect of the predictor and a time-varying quantifying dynamically 
the difference to the time-invariant effect over the follow-up. The function of time can take various functional 
forms, and we have indicated the selected function for each predictor in the tables for time-dependent effects. 
To get an idea of the total impact of covariate Z, the time-varying and time-invariant effects must be interpreted 
together, the HR estimates themselves are difficult to digest, and it is more informative to present the hazard 
ratio as a figure. In absence of time-dependent effects, the model Eq. (1) becomes the conventional proportional 
hazards Cox regression model.
(1)hi(t) = h0(t)eγ 0zi+γ tzig(t)+β
Txi ,
Figure 1.  Flowchart over the 3,309 men and their inclusion in the present study. Solid lines represent groups 
relevant to the present study.
4Vol:.(1234567890)
Scientific Reports |        (2023) 13:16526  | https://doi.org/10.1038/s41598-023-43666-x
www.nature.com/scientificreports/
As model validation, we assessed time-dependency in model effects using plots and tests based on scaled 
Schoenfeld-residuals, functional form of model effects based on martingale residuals, and impact from outliers 
based on dfbeta-statistics. Extended model effects (non-linear and time-dependent effects) are shown using 
predicted hazard plots. We adjusted our models with smoking, alcohol consumption, and continuous BMI in 
midlife. As some of these variables were used in estimating CVD risk, analyses of CVD risk were adjusted for 
alcohol consumption only. Supplementary Tables 1–2 show information about support for model assumptions 
and count of participants at risk in our analysis regarding age at retirement. Supplementary Tables 3–4 show the 
respective information in our analysis on survival in retirement. We normalized work-loss data by performing a 
log (base 10) transformation and back-transformed the point estimates. For the models we report hazard ratios 
or regression coefficients together with their 95% confidence intervals. The analyses were carried out using R 
 software18 (version 4.2.1) with packages  survival19 and  extrafont20, and linear regression models in Stata (Release 
17. College Station, TX: StataCorp LLC).
Ethical approval
The research protocol of the follow-up study of HBS has been approved by the Ethics Committee of the Depart-
ment of Medicine, University of Helsinki. The study adheres with the principles stated in the Declaration of 
Helsinki.
Consent to participate
Informed consent was obtained from all individual participants included in the study.
Results
Pre‑retirement workforce participation: age at retirement and work‑loss years
The men had retired at the mean age of 61.2 years (SD = 4.4 years). Table 1 shows that men in the lowest quintile 
of retirement age (< 58 years) had had poorer SRH and SRF in midlife and lower mean risk of 5- and 10-year 
estimated risk of non-fatal/fatal CVD than those in the highest quintile retiring at age ≥ 65 years. In addition, 
441 men had died before the statutory retirement age of 65 years. The combined work-loss through retiring or 
dying before age 65 years was 4.8 years (SD = 4.3 years; range = 0–38 years).
Post‑retirement survival
After the men had retired, their mean survival time was 24.4 years (SD = 6.6 years; range 3–54 years). The lexis 
diagram indicated that survival in retirement was relatively evenly distributed (Supplementary Fig. 1). Mortality 
increased during the study period according to the degree of estimated risk of CVD in midlife; in the year 2018 
fewer than 1 out of 10 had survived in the groups with the highest risk, whereas slightly less than half in the 
low-risk groups were alive in the year 2018 (Supplementary Table 5).
Health in midlife and sample selection
We report findings on men who had been measured for CVD health factors in midlife and who provided informa-
tion on retirement age in a questionnaire (Fig. 1). Participants who had been assessed for CVD health factors in 
midlife but who were not included in the present study because of non-respondence to the questionnaire (n = 557) 
or death (n = 889) had been less healthy in most considered health factors in midlife (Supplementary Table 6).
Self‑rated health and physical fitness in midlife as determinants of pre‑retirement workforce 
participation
Poorer SRH (average or poor vs. very good) and SRF (very good vs. fair) were associated with higher hazards of 
retiring at a younger age, of which poorer SRH (poor vs. very good, covariate-adjusted HR = 1.63, 95% CI = 1.14, 
2.33) and SRF (very good vs. fair, covariate-adjusted HR = 0.71, 95% CI = 0.52, 0.94) remained statistically signifi-
cant after adjusting for smoking, alcohol consumption, and BMI in midlife (Table 2). Similarly, men with poor 
SRH (poor vs. very good) had 0.32 years more work-loss (95% CI = 0.13, 0.52) in analyses adjusted with smoking, 
BMI, and alcohol consumption in midlife (Table 3). Poor SRF (very poor vs. fair) was initially associated with 
more work-loss (crude β = 0.45, 95% CI = 0.14, 0.66), but this association attenuated following adjustment for 
covariates (covariate-adjusted β = 0.38, 95% CI = − 0.29, 0.69).
Laboratory and clinical CVD risk factors in midlife as determinants of pre‑retirement work‑
force participation
Some but not all individual CVD risk factors showed associations with earlier retirement or more work-loss 
years, with many associations becoming statistically non-significant after covariate adjustment (Tables 2 and 3). 
Systolic BP had a curvilinear association to retirement hazard, suggesting that those in either low or high end of 
the range of observed readings had a lower hazard of retiring than those with readings in the mid-range (Fig. 2A). 
The association was weak, and it did not change substantially after covariate adjustment, which also reduced the 
sample size. Similarly, one-unit increases in systolic BP and diastolic BP showed initial positive associations of 
more work-loss years, but these associations attenuated after adjustment for smoking, BMI, and alcohol con-
sumption (Table 3). Time-dependent hazards of unit-higher triglyceride levels showed an elevated hazard rate 
(HR(t) ~ 2.2) of earlier retirement, which then levelled off by age 60 years (Table 2 and Fig. 2B). For each 1 mmol/l 
increase in 1 h-OGTT (covariate-adjusted β = 0.01, 95% CI = 0.00, 0.03) and triglycerides (covariate-adjusted 
β = 0.05, 95% CI = 0.01, 0.09), the men were more likely to have more work-loss years (Table 3).
5Vol.:(0123456789)
Scientific Reports |        (2023) 13:16526  | https://doi.org/10.1038/s41598-023-43666-x
www.nature.com/scientificreports/
Table 1.  Characteristics of the study population according to quintiles of age at retirement. SD = standard 
deviation, BP = blood pressure, CVD = cardiovascular disease; OGTT = oral glucose tolerance test.
Quintiles of age at retirement
N 0–20th percentile, ≤ 58y
20–40th percentile, 
59-60y
40–60th percentile, 
61-62y
60–80th percentile, 
63-64y
80–100th 
percentile, ≥ 65y p
Midlife characteristics in the year 1974
 Age, average years (SD) 1766 45.8 (3.7) 46.5 (3.9) 47.4 (3.9) 47.9 (3.8) 48.4 (4.2)  < 0.001
 Self-rated health, % 1554  < 0.001
  Very good 101 3.7 6.7 6.2 6.9 8.7
  Fairly good 711 38.7 46.0 44.5 50.2 48.9
  Average 667 49.7 39.6 46.6 40.6 39.0
  Poor 75 8.0 7.6 2.8 2.3 3.3
 Self-rated physical fit-
ness, % 1556 0.002
  Very good 56 1.0 4.9 3.1 3.3 5.4
  Good 490 32.6 29.3 25.3 34.1 35.7
  Fair 795 47.8 51.5 59.2 51.5 46.2
  Poor 200 16.3 13.7 12.1 10.8 11.4
  Very poor 15 2.3 0.6 0.3 0.3 1.2
 1 h-OGTT, mmol/l (SD) 1285 7.1 (2.1) 6.7 (1.8) 7.0 (1.9) 6.8 (1.9) 7.0 (2.1) 0.262
 Systolic BP, mmHg (SD) 1037 142.7 (18.5) 140.7 (17.3) 142.2 (16.8) 141.1 (18.6) 142.7 (18.6) 0.775
 Diastolic BP, mmHg 
(SD) 1037 93.1 (12.0) 90.3 (10.9) 91.6 (10.5) 90.2 (11.4) 91.3 (11.7) 0.179
 Pulse, bpm (SD) 976 65.5 (10.7) 64.3 (10.0) 64.1 (11.1) 63.9 (10.7) 63.8 (10.3) 0.559
 Total cholesterol, mmol/l 
(SD) 1306 6.3 (1.1) 6.1 (1.0) 6.2 (1.0) 6.1 (1.0) 6.2 (1.0) 0.464
 Triglycerides, mmol/l 
(SD) 1290 1.7 (1.0) 1.6 (0.8) 1.6 (0.8) 1.5 (0.7) 1.5 (0.9) 0.028
 Body mass index, kg/
m2 (SD) 1357 26.1 (2.7) 26.1 (2.8) 25.8 (2.5) 25.3 (2.7) 25.6 (2.7) 0.002
 Alcohol consumption, g/
week (SD) 1371 168 (153) 163 (136) 140 (122) 136 (129) 145 (127) 0.026
 Smoking status, % 1752 0.905
  Current smoker 23.1 23.2 22.5 20.8 21.2
 Midlife 5-year risk of 
incident CVD, per SD 1006 0.38 (0.98) 0.37 (0.96) 0.48 (0.93) 0.42 (0.94) 0.63 (0.95) 0.018
 Midlife SCORE 10-year 
risk of fatal CVD, % 
(SD)
1034 0.80 (0.93) 0.85 (0.98) 1.08 (1.21) 1.11 (1.40) 1.35 (1.56) 0.002
 Categorical midlife 
SCORE relative risk of 
10-year fatal CVD, n (%)
1034 0.003
  < One fold 42 71 (42.0) 97 (42.9) 74 (37.2) 83 (38.8) 69 (30.5)
  1 543 77 (45.6) 86 (38.1) 75 (37.7) 75 (35.0) 94 (41.6)
  Two fold 201 11 (6.5) 28 (12.4) 27 (13.6) 32 (15.0) 26 (11.5)
  Three Four fold 145 8 (4.7) 14 (6.2) 18 (9.0) 17 (7.9) 23 (10.2)
   ≥ Five fold 103 2 (1.2) 1 (0.4) 5 (2.5) 7 (3.3) 14 (6.2)
Old-age characteristics in the year 2000
 Age, average years (SD) 1766 71.8 (3.7) 72.5 (3.9) 73.4 (3.9) 73.9 (3.8) 74.4 (4.2)  < 0.001
 CVD, n (%) 1767 70.8 63.0 60.2 60.0 59.4 0.010
 Old age SCORE 10-year 
risk of fatal CVD, % 
(SD)
1212 5.8 (2.6) 5.8 (2.4) 5.6 (2.0) 5.9 (2.2) 5.7 (2.5) 0.259
Survival characteristics
 Survival in retirement, 
years (SD) 1767 29.2 (6.9) 25.0 (6.1) 24.4 (5.5) 23.3 (5.6) 20.4 (5.7)  < 0.001
 Mortality between 2000 
and 2010, n (%) 1766 122 (35.7) 144 (38.6) 107 (33.0) 115 (33.3) 137 (35.9) 0.530
 Mortality between 2011 
and 2018, n (%) 1766 112 (32.7) 105 (28.2) 109 (33.6) 109 (31.6) 141 (36.9) 0.141
6Vol:.(1234567890)
Scientific Reports |        (2023) 13:16526  | https://doi.org/10.1038/s41598-023-43666-x
www.nature.com/scientificreports/
Composite CVD risk estimates in midlife as determinants of pre‑retirement workforce 
participation
The associations between midlife CVD risk and retirement and work-loss years were conflicting; while higher 
risk of CVD was associated with retiring at an older age, there was evidence suggesting that those with high risk 
were more likely to have more years of work-loss (Tables 2 and 3). Higher 5- and 10-year CVD risk scores in 
midlife were associated with a lower covariate-adjusted hazard of retirement (Table 2). Respectively for work-
loss years, each increase of one standard deviation in the 5-year risk of incident CVD was not associated with 
work-loss years. The association between 10-year fatal CVD risk estimates and work loss showed that individuals 
with a 10-year fatal CVD risk estimate of 1 or higher vs. less than 1 relative risk had 0.32 more work-loss years 
(covariate-adjusted β = 0.32, 95% CI = 0.13, 0.53).
Health in midlife as a determinant of post‑retirement survival
Poorer SRH (average or poor vs. very good) and SRF (poor or very poor vs. fair) were associated with fewer years 
spent in retirement, of which poor SRH (poor vs. very good, covariate-adjusted HR = 2.05, 95% CI = 1.37, 3.07) 
and SRF (poor vs. fair, covariate-adjusted HR = 1.26, 95% CI = 1.20, 1.55) remained after adjustment for smoking, 
alcohol consumption, and BMI in midlife (Table 4). Per each 1 mmol/l higher total cholesterol (covariate-adjusted 
HR = 1.08, 95% CI = 1.01, 1.15) and triglycerides (covariate-adjusted HR = 1.08, 95% CI = 1.01, 1.15) the men 
were more likely to spend fewer years on retirement (Table 4). Higher systolic BP, diastolic BP, and midlife car-
diovascular risk scores showed age-adjusted time-dependent hazard of spending years in retirement (Table 4 and 
Fig. 3). In panel A, unit-higher midlife systolic BP was associated with an almost linearly increasing hazard from 
age 75 years to HR(t) ~ 1.015 at age 97. In panel B, unit-higher midlife diastolic BP showed an exponential rate 
of growth in the time-dependent hazard rate of fewer years spent in retirement from age 67 years with a peak of 
HR(t) ~ 1.05 at age 97. In panel C, one standard deviation higher estimated 5-year incident risk of cardiovascular 
Table 2.  Cox regression models of midlife health factors as predictors of retirement age in years among older 
retired businessmen. HR = hazard ratio, CI = confidence interval, BP = blood pressure, CVD = cardiovascular 
disease; OGTT = oral glucose tolerance test. a Analyzed individually. b Model adjusted with alcohol 
consumption, smoking, and body mass index in 1974. Models of midlife 5-year and 10-year risk of CVD were 
adjusted with alcohol consumption only. c Mean-centered. Significant values are in bold.
Predictora
Crude Adjustedb
Events Crude HR (95% CI) Events Adjusted HR (95% CI)
Self-rated health 1533 1270
 Very good 101 Ref 90 Ref
 Fairly good 702 1.20 (0.97, 1.48) 620 1.15 (0.92, 1.44)
 Average 660 1.32 (1.07, 1.62) 513 1.19 (0.95, 1.49)
 Poor 70 1.80 (1.32, 2.44) 47 1.63 (1.14, 2.33)
Self-rated physical fitness 1535 1272
 Very good 56 0.69 (0.52, 0.90) 53 0.71 (0.52, 0.94)
 Good 479 0.93 (0.83, 1.05) 407 0.93 (0.82, 1.05)
 Fair 789 Ref 659 Ref
 Poor 197 1.05 (0.90, 1.23) 146 1.03 (0.86, 1.24)
 Very poor 14 1.14 (0.67, 1.94) 7 0.79 (0.38, 1.67)
1 h-OGTT, mmol/l 1270 1.00 (0.98, 1.03) 1210 1.00 (0.97, 1.03)
Systolic BP, mmHg 1024 991
 - linear 1.04 (1.00, 1.08) 1.04 (1.00, 1.08)
 - quadratic 1.00 (1.00, 1.00) 1.00 (1.00, 1.00)
Diastolic BP, mmHg 1024 1.00 (1.00, 1.00) 991 1.00 (1.00, 1.01)
Pulse, bpm 963 1.00 (1.00, 1.01) 899 1.00 (1.00, 1.01)
Total cholesterol, mmol/l 1291 1.02 (0.96, 1.07) 1231 1.02 (0.96, 1.07)
Triglycerides, mmol/lc 1275 1217
 - time-constant 0.99 (0.92, 1.07) 0.97 (0.89, 1.05)
 - time-varying 1.00 (1.00, 1.00) 1.00 (1.00, 1.00)
Midlife 5-year risk of incident CVD, per SD 993 0.91 (0.85, 0.97) 987 0.90 (0.84, 0.96)
Midlife SCORE 10-year relative risk of fatal CVD 1021 988
 < 1 389 Ref 370 Ref
 1 404 0.91 (0.79, 1.05) 393 0.90 (0.78, 1.03)
 2 121 0.79 (0.64, 0.97) 120 0.79 (0.64, 0.97)
 3–4 60 0.85 (0.65, 1.12) 59 0.85 (0.65, 1.11)
 ≥ 5 47 0.54 (0.40, 0.73) 46 0.51 (0.30, 0.70)
7Vol.:(0123456789)
Scientific Reports |        (2023) 13:16526  | https://doi.org/10.1038/s41598-023-43666-x
www.nature.com/scientificreports/
disease in midlife showed curvilinear time-dependent elevated hazard rates peaking at approximately age 85 years 
(HR(t) ~ 1.3)., which levelled off so that the hazard rate remained indicative of elevated hazard for higher risk 
scores. In panel D, unit-higher 10-year risk of cardiovascular disease showed a linear increase from age 82 years 
to HR(t) ~ 1.4 at age 97.
Discussion
Cardiovascular disease (CVD) is a leading cause of disability pension and premature exit from the workforce, sig-
nificantly affecting work ability and labour force  participation2,5. Despite its impact, limited research has focused 
on studying CVD risk factors or composite CVD risk estimates in populations without pre-existing CVD. Our 
study, based on business executives followed for up to 44 years, found a time-dependent relationship between 
higher triglyceride levels in midlife and earlier retirement, independent of lifestyle-related factors including 
smoking, alcohol use, and BMI in midlife. Furthermore, although men with higher CVD risk in midlife were 
more likely to retire at an older age, we found evidence to support that they also experienced more work-loss 
years, i.e., to retire or die before age 65 years. After the participants had retired, those who had had higher CVD 
risk in their midlife also had higher mortality risk in retirement. Poorer self-rated health and physical fitness 
predicted earlier retirement, more work-loss years, and poorer survival in retirement.
Interpretation in light of known literature
We were able to expand the previous scarce literature on CVD risk and retirement, where both early and on-
time retirees displayed similar CVD risk  factors7, by providing evidence of higher triglyceride levels associating 
with increased hazard of earlier retirement. While this association was independent of lifestyle factors includ-
ing alcohol consumption, higher triglyceride levels could still reflect overall poor health, stress, or other factors 
that predispose to early retirement. Poorer self-rated health, a known predictor of transitioning into disability 
 pensions21–23, was also associated with increased hazard of earlier retirement in our study, thus corroborating 
previous findings. Individuals who self-rate their health as favourable may exhibit improved work ability and 
the absence of conditions which may compromise their ability to  work21–23. However, our results suggest that 
CVD risk may not be a major determinant of earlier retirement overall, as most other CVD risk factors were 
Table 3.  Linear regression model estimates of midlife health factors as predictors of work-loss years among 
older business executives (back-transformed from log10). Note. CI = confidence interval; BP = blood pressure; 
CVD = cardiovascular disease; OGTT = oral glucose tolerance test β = back-transformed point estimate. 
a Analyzed individually. b Adjusted for alcohol consumption, smoking, and body mass index in 1974. Models of 
midlife 5-year and 10-year risk of CVD were adjusted with alcohol consumption only. Significant values are in 
bold.
Midlife health  factorsa N
Crude
N
Adjustedb
β (95% CI) β (95% CI)
Self-rated health 1520 1256
 Very good Ref Ref
 Fairly good 0.05 (-0.12, 0.19) 0.05 (-0.13, 0.21)
 Average 0.13 (-0.03, 0.26) 0.11 (-0.06, 0.27)
 Poor 0.32 (0.16, 0.46) 0.32 (0.13, 0.52)
Self-rated physical fitness 1522 1258
 Very good -0.01 (-0.23, 0.19) 0.04 (-0.19, 0.25)
 Good 0.02 (-0.06, 0.11) 0.06 (-0.04, 0.14)
 Fair
 Poor 0.10 (-0.00, 0.21) 0.08 (-0.04, 0.20)
 Very poor 0.45 (0.14, 0.66) 0.38 (-0.29, 0.96)
1 h-OGTT, mmol/l 1273 0.02 (0.01, 0.04) 1200 0.01 (0.00, 0.03)
Systolic BP, mmHg 965 0.00 (0.00, 0.00) 935 0.00 (-0.00, 0.00)
Diastolic BP, mmHg 965 0.00 (0.00, 0.01) 935 0.00 (-0.00, 0.01)
Pulse, bpm 890 0.00 (-0.00, 0.01) 830 0.00 (-0.00, 0.01)
Total cholesterol, mmol/l 1301 0.03 (-0.00, 0.07) 1222 0.03 (-0.00, 0.07)
Triglycerides, mmol/l 1284 0.07 (0.03, 0.12) 1208 0.05 (0.01, 0.09)
5-year risk of incident CVD, per SD 940 -0.00 (-0.05, 0.04) 933 -0.01 (-0.06, 0.04)
Midlife SCORE 10-year relative risk of fatal CVD 963
 < 1 Ref Ref
 1 0.33 (0.15, 0.55) 0.32 (0.13, 0.53)
 2 0.33 (0.15, 0.57) 0.31 (0.11, 0.54)
 3–4 0.19 (-0.06, 0.40) 0.16 (-0.09, 0.38)
 ≥ 5 0.17 (-0.09, 0.40) 0.16 (-0.11, 0.38)
8Vol:.(1234567890)
Scientific Reports |        (2023) 13:16526  | https://doi.org/10.1038/s41598-023-43666-x
www.nature.com/scientificreports/
unrelated to hazard of earlier retirement. CVD risk alone may not be sufficient to generate symptoms with sig-
nificant impact work ability. Moreover, men at higher risk of developing CVD in midlife were not at increased 
hazard of retiring early; in fact, they showed a decreased hazard. Considering that the men underwent clinical 
examinations in the Finnish Institute of Occupational Health and that by old age the difference in CVD risk by 
retirement age appeared to have levelled off, it is possible that men with high CVD risk received interventions 
to mitigate their high risk.
Considering these findings and the fact that one can exit the workforce also through premature mortality, we 
found parallel evidence to support CVD risk factors as determinants of more work-loss years, where participants 
retiring or dying prematurely (younger than age 65 years) were considered. We could expand the previous lit-
erature on estimated productivity losses due to medical conditions, e.g.  hypertension24,25, heterozygous familial 
 hypercholesterolemia26, and  diabetes27,28 to include CVD risk factors and estimated CVD risk by showing evi-
dence that higher CVD risk, levels of 1 h-OGTT, and serum triglycerides in midlife were associated with more 
work-loss years. However, as with retirement age, most other CVD risk factors were unrelated to work-loss, and 
findings on CVD risk and work-loss years were less consistent.
Our follow-up extended to the 9th decade, enabling us to describe survival among the study participants and 
provide valuable information on the impact of CVD risk factors in old age. While CVD risk factors are known 
predictors of overall  survival13,29,30, a meta-analysis on the association between retirement age and  survival9 
highlighted the significance of prior health in their analysis. Our findings contribute to this gap by suggesting 
that midlife CVD health factors not only potentially impact productivity losses but also contribute to fewer 
years in retirement. We found that men with higher CVD risk factors, higher estimated risk of 5-year incident 
or 10-year fatal CVD, and poorer self-rated health and fitness, had a higher hazard for all-cause mortality in 
retirement, regardless of smoking, alcohol consumption, and BMI in midlife. Some of these associations also 
showed time-dependent  hazards9,13,17,24–30.
Importance of the results
These findings highlight the impact of CVD risk on the work-life trajectory of individuals and emphasize the 
importance of early identification and management of modifiable risk factors to reduce CVD-related morbidity 
and mortality. First, although CVD is a leading cause of premature retirement, we found no evidence to support 
CVD risk factors as major determinants of earlier retirement, except for triglyceride levels, self-rated health, 
and physical fitness. However, when premature mortality before age 65 years was also considered in combina-
tion with earlier retirement in work-loss years, greater overall CVD risk also emerged as a potential factor in 
Figure 2.  Predicted time-dependent hazard ratios for systolic blood pressure (panel A, top) and triglyceride 
concentration (panel B, bottom) assessed in midlife. The curves show the time-varying HR (solid dark grey line) 
of retirement age in years, and reference lines for HR estimated under the proportional hazards assumption 
(dashed grey line) and HR for strictly equal hazards (grey, dotted line).
9Vol.:(0123456789)
Scientific Reports |        (2023) 13:16526  | https://doi.org/10.1038/s41598-023-43666-x
www.nature.com/scientificreports/
productivity losses. Second, our results suggest that potential productivity losses related to higher CVD risk 
may also occur in healthy individuals, as only a small proportion of our study participants had CVD or diabetes 
in midlife. Therefore, efficient management of CVD risk may prevent potential productivity losses even among 
healthy individuals. Third, our study supports the notion that both general health and CVD risk factors are criti-
cal factors to consider when examining retirement outcomes and survival in general.
Strengths and limitations
We examined the association between CVD risk factors, retirement age, work-loss years, and mortality in older 
male executives. The study followed the men for up to 44 years, until the 9th decade of life, and used national 
register-linkage to gather data on mortality. The sample was subject to selective survival and loss to follow-up, 
leaving out participants who had poorer general and CVD health in midlife, thus potentially undermining our 
results regarding retirement age and survival in retirement. Given that CVD is one of the leading causes of 
premature exit from the  workforce2,5, we believe that participants at high risk of CVD but who did not survive 
to retirement age may have also been more likely to retire early due to CVD-related health problems. The study 
population consisted of Caucasian men from higher socioeconomic  groups10, which provides internal validity 
but significantly limits generalizability to other socioeconomic groups, women, or individuals with different 
retirement policies. The men worked as executives and entrepreneurs who did not have fixed retirement ages. 
Laboratory CVD risk factors were measured in the 1970s using methods from that time, which may be outdated, 
but have been shown to correlate with clinical outcomes in later studies from this  cohort14. Due to the extended 
time span between the assessment of CVD risk factors and retirement, participants’ cardiovascular risk levels 
may have undergone changes or interventions, such as pharmacological treatments. Our definition of work-loss 
years has potential caveats as the retirement characteristics were self-reported. Nonetheless, we used a cut-off at 
65 years, which has been used by  others17.
Table 4.  Cox regression models of midlife health factors as predictors of survival in retirement among older 
retired businessmen. HR = hazard ratio, CI = confidence interval, BP = blood pressure, CVD = cardiovascular 
disease. a Analyzed individually. bLog-transformed. cLog-transformed, centered time. dCube-transformed, 
rescaled time. eSquared, rescaled time. fModel adjusted with alcohol consumption, smoking, and body mass 
index in 1974. Models of midlife 5-year and 10-year risk of CVD were adjusted with alcohol consumption only. 
Significant values are in bold.
Predictora N
Crude
N
Adjustedf
Events HR (95% CI) Events HR (95% CI)
Self-rated health 1554 1053 1287 873
 Very good 101 58 Ref 90 55 Ref
 Fairly good 711 465 1.31 (0.99, 1.72) 628 418 1.21 (0.91, 1.60)
 Average 667 466 1.42 (1.08, 1.87) 519 357 1.16 (0.87, 1.54)
 Poor 75 64 2.46 (1.72, 3.51) 50 43 2.05 (1.37, 3.07)
Self-rated physical fitness 1556 1053 1289 874
 Very good 56 34 0.79 (0.56, 1.12) 53 33 0.93 (0.65, 1.33)
 Good 490 327 0.92 (0.81, 1.07) 417 284 1.02 (0.87, 1.18)
 Fair 795 527 Ref 664 440 Ref
 Poor 200 153 1.39 (1.16, 1.67) 148 112 1.26 (1.20, 1.55)
 Very poor 15 12 2.04 (1.15, 3.62) 7 5 1.31 (0.54, 3.19)
1-h glucose, mmol/l 1285 875 1.04 (1.01, 1.08) 1335 842 1.02 (0.99, 1.06)
Systolic BP, mmHg 1037 690 1004 673
 - time-constant 0.95 (0.91, 1.00) 0.95 (0.90, 1.00)
 - time-dependent 1.02 (1.00, 1.03) 1.02 (1.00, 1.03)
Diastolic BP, mmHg 1037 690 1004 673
 - time-constant 1.00 (0.99, 1.01) 0.94 (0.87, 1.01)
 - time-dependent 1.02 (1.00, 1.04)d 1.02 (1.00, 1.05)d
Pulse, bpm 976 656 1.00 (1.00, 1.01) 910 614 1.00 (0.99, 1.01)
Total cholesterol, mmol/l 1306 887 1.14 (1.07, 1.21) 1246 855 1.08 (1.01, 1.15)
Triglycerides, mmol/l 1290 875 1.13 (1.06, 1.22) 1246 855 1.08 (1.01, 1.15)
Midlife 5-year risk of incident CHD, per  SDb 1006 673 1000 671
 - time-constant 1.42 (1.28, 1.58) 1.40 (1.26, 1.55)
 - time-dependent 0.76 (0.69, 0.84)e 0.76 (0.68, 0.83)e
Midlife SCORE 10-year risk of fatal CVD 1034 689 1001 672
 - time-constant 0.28 (0.17, 0.47) 0.28 (0.17, 0.47)
 - time-dependent 1.58 (1.33, 1.87)d 1.58 (1.33, 1.87)
Old age SCORE risk of fatal CVD 1212 810 1.05 (1.02, 1.08) 929 624 1.04 (1.00, 1.07)
10
Vol:.(1234567890)
Scientific Reports |        (2023) 13:16526  | https://doi.org/10.1038/s41598-023-43666-x
www.nature.com/scientificreports/
Future recommendations
Efforts aimed at early identification and management of CVD risk factors may not only enhance the productiv-
ity of individuals during their working years but also contribute to a longer and healthier retirement period. To 
achieve this, clinicians should prioritize assessing and addressing CVD risk factors in midlife to reduce morbid-
ity and mortality. Targeted interventions should be implemented for individuals with poorer self-rated health, 
while comprehensive retirement planning should incorporate cardiovascular health considerations to promote a 
longer and healthier retirement period. Further investigation is warranted to explore the underlying mechanisms 
behind the association between higher triglyceride levels and the time-dependent hazard of early retirement, as 
the precise mechanisms remain unclear.
Figure 3.  Predicted time-dependent hazard ratios for systolic blood pressure (panel A, top), diastolic blood 
pressure (panel B), 5-year risk of incident CVD (panel C), and 10-year risk of fatal CVD (panel D, bottom). The 
curves show the time-varying HR (solid dark grey line) of survival in years, and reference lines for HR estimated 
under the proportional hazards assumption (dashed grey line) and HR for strictly equal hazards (grey, dotted 
line).
11
Vol.:(0123456789)
Scientific Reports |        (2023) 13:16526  | https://doi.org/10.1038/s41598-023-43666-x
www.nature.com/scientificreports/
Conclusions
This study found evidence that CVD risk factors can serve as predictors of both pre-retirement workforce 
participation and post-retirement survival. Although high midlife CVD risk was not a consistent predictor of 
earlier retirement, high risk may contribute to premature mortality, potentially leading to poorer pre-retirement 
workforce participation. Of considered CVD risk factors, only high midlife triglyceride levels were associated 
with the hazard to retire earlier. Midlife CVD health factors and composite risk estimates were linked to a higher 
hazard of mortality in retirement. Those with poorer self-rated health or fitness were at higher hazard to retire 
earlier, more likely to have more work-loss years, and at higher hazard for mortality in retirement, rendering them 
potential targets for interventions. The effective management of CVD risk factors may enhance the productivity 
of working-aged individuals and the survival of retirees.
Data availability
The datasets used and/or analysed during the current study available from Timo E. Strandberg on reasonable 
request.
Received: 13 June 2023; Accepted: 27 September 2023
References
 1. Burden, G. et al. Global burden of cardiovascular diseases and risk factors, 1990–2019: Update from the GBD 2019 study. J. Am. 
Coll. Cardiol. 76(25), 2982–3021 (2020).
 2. Bin Sayeed, M. S., Joshy, G., Paige, E., Banks, E. & Korda, R. Cardiovascular disease subtypes, physical disability and workforce 
participation: A cross-sectional study of 163,562 middle-aged Australians. Plos One 16(4), e0249738 (2021).
 3. Oude Hengel, K., Robroek, S. J. W., Eekhout, I., Van Der Beek, A. J. & Burdorf, A. Educational inequalities in the impact of chronic 
diseases on exit from paid employment among older workers: A 7-year prospective study in the Netherlands. Occup. Environ. Med. 
76(10), 718–725 (2019).
 4. Virtanen, M. et al. The joint contribution of cardiovascular disease and socioeconomic status to disability retirement: A register 
linkage study. Int. J. Cardiol. 1(230), 222–227 (2017).
 5. Kouwenhoven-Pasmooij, T. A., Burdorf, A., Roos-Hesselink, J. W., Hunink, M. G. M. & Robroek, S. J. W. Cardiovascular disease, 
diabetes and early exit from paid employment in Europe; The impact of work-related factors. Int. J. Cardiol. 15(215), 332–337 
(2016).
 6. Timmis, A. et al. European society of cardiology: Cardiovascular disease statistics 2019. Eur. Heart J. 41(1), 12–85 (2020).
 7. Salonen, P. Factors associated with premature departure from working life among ageing food industry employees. Occup. Med. 
(Chic Ill). 53(1), 65–68 (2003).
 8. Goetzel, R. Z. et al. The relationship between modifiable health risk factors and medical expenditures, absenteeism, short-term 
disability, and presenteeism among employees at novartis. J. Occup. Environ. Med. 51(4), 487–499 (2009).
 9. Sewdas R, De Wind A, Stenholm S, Coenen P, Louwerse I, Boot C, et al. Association between retirement and mortality: Working 
longer, living longer? A systematic review and meta-analysis. Vol. 74, Journal of Epidemiology and Community Health. BMJ 
Publishing Group; 2020. p. 473–80.
 10. Strandberg, T. E. et al. Cohort profile: The Helsinki businessmen study (HBS). Int. J. Epidemiol. 45(4), 1074–1074h (2016).
 11. Keys, A. et al. Probability of middle-aged men developing coronary heart disease in five years. Circulation. 45(4), 815–828 (1972).
 12. Piepoli, M. F. et al. 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 
representative of 10 societies and by invited experts. Developed with the special contribution of the European association for 
cardiovascular prevention & rehabilitation). Eur. J. Prev. Cardiol. 23(11), 1–96 (2016).
 13. Hyttinen, L. et al. Effect of cholesterol on mortality and quality of life up to a 46-year follow-up. Am J. Cardiol. 108(5), 677–681 
(2011).
 14. Strandberg, T. E. et al. One-hour glucose, mortality, and risk of diabetes: A 44-year prospective study in men. Archiv. Intern. Med. 
171(10), 941–954 (2011).
 15. Aktas, M. K., Ozduran, V., Pothier, C. E., Lang, R. & Lauer, M. S. Global risk scores and exercise testing for predicting all-cause 
mortality in a preventive medicine program. JAMA 292(12), 1462–1468 (2004).
 16. Haapanen, M. J. et al. Retirement age and type as predictors of frailty: A retrospective cohort study of older businessmen. BMJ 
Open 10(12), 37722 (2020).
 17. von Bonsdorff, M. B. et al. Work-loss years among people diagnosed with diabetes: A reappraisal from a life course perspective. 
Acta Diabetol. 55(5), 485–491 (2018).
 18. R Core Team. R: A language and environment for statistical computing. Vienna: R Foundation for Statistical Computing; 2020.
 19. Therneau, T. M. Survival Analysis [R package survival version 3.5–5]. 2023 Mar 12;
 20. CRAN - Package extrafont [Internet]. [cited 2023 Apr 3]. Available from: https:// cran.r- proje ct. org/ web/ packa ges/ extra font/ index. 
html
 21. Halford, C. et al. Effects of self-rated health on sick leave, disability pension, hospital admissions and mortality. A population-based 
longitudinal study of nearly 15,000 observations among Swedish women and men. BMC Public Health 12, 1103 (2012).
 22. Pietiläinen, O., Laaksonen, M., Rahkonen, O. & Lahelma, E. Self-rated health as a predictor of disability retirement–the contribu-
tion of ill-health and working conditions. PloS One 6(9), e25004 (2011).
 23. De Breij, S. et al. Educational differences in the influence of health on early work exit among older workers. Occup. Environ. Med. 
77(8), 568–575 (2020).
 24. Kiiskinen, U., Vartiainen, E., Puska, P. & Aromaa, A. Long-term cost and life-expectancy consequences of hypertension. J. Hyper-
tens. 16(8), 1103–1112 (1998).
 25. Hird, T. R. et al. Productivity burden of hypertension in Australia: A life table modeling study. Hypertension 73(4), 777–784 (2019).
 26. Ademi, Z. et al. The economic impact of familial hypercholesterolemia on productivity. J. Clin. Lipidol. 14(6), 799–806 (2020).
 27. Schofield, D. et al. The costs of diabetes among Australians aged 45–64-years from 2015 to 2030: Projections of lost productive 
life years (PLYs), lost personal income, lost taxation revenue, extra welfare payments and lost gross domestic product from 
Health&WealthMOD2030. BMJ Open 7(1), 13158 (2017).
 28. Dall, T. M. et al. The economic burden of elevated blood glucose levels in 2012: Diagnosed and undiagnosed diabetes, gestational 
diabetes mellitus, and prediabetes. Diabetes Care 37(12), 3172–3179 (2014).
 29. Strandberg, T. E. et al. Low cholesterol, mortality, and quality of life in old age during a 39-year follow-up. J. Am. Coll. Cardiol. 
44(5), 1002–1008 (2004).
12
Vol:.(1234567890)
Scientific Reports |        (2023) 13:16526  | https://doi.org/10.1038/s41598-023-43666-x
www.nature.com/scientificreports/
 30. Strandberg, A. Y. et al. Low midlife blood pressure, survival, comorbidity, and health-related quality of life in old age: The Helsinki 
Businessmen study. J. Hypertens. 32(9), 1797–1804 (2014).
Author contributions
M.J.H. designed the present study, drafted the article, analyzed the data, and drafted the manuscript. T.T. designed 
the present study, analyzed the data, interpreted the data, and revised the article critically for important intellec-
tual content. M.B.B. designed the present study, interpreted the data, and revised the article critically for impor-
tant intellectual content. M.E.B. and A.Y.S. interpreted the data and revised the article critically for important 
intellectual content. T.E.S. acquired funding for the cohort, data collection, interpreted the data and revised the 
article critically for important intellectual content.
Funding
MJH is supported by Finska Läkaresällskapet r.f. and Liv och Hälsa r.f. The Academy of Finland supported MBvB 
with grant 257239 and 349336. The Academy of Finland supported TT with grant 286536. The Academy of 
Finland supported MEvB with grants 294530, 307114 and 303920. TES is supported by the Helsinki University 
Hospital, Department of Medicine.
Competing interests 
The authors declare no competing interests.
Additional information
Supplementary Information The online version contains supplementary material available at https:// doi. org/ 
10. 1038/ s41598- 023- 43666-x.
Correspondence and requests for materials should be addressed to M.J.H.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and 
institutional affiliations.
Open Access  This article is licensed under a Creative Commons Attribution 4.0 International 
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or 
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the 
Creative Commons licence, and indicate if changes were made. The images or other third party material in this 
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the 
material. If material is not included in the article’s Creative Commons licence and your intended use is not 
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from 
the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.
© The Author(s) 2023