As a service to our readers, Harvard Health Publishing provides access to our library of archived content. Please note the date of last review or update on all articles.
No content on this site, regardless of date, should ever be used as a substitute for direct medical advice from your doctor or other qualified clinician.
To evaluate how self-reported leisure-time physical activity [PA] changes during the adult life span, and to study how PA is related to cardiovascular risk factors using longitudinal studies.
Methods
Several Swedish population-based longitudinal studies were used in the present study [PIVUS, ULSAM, SHE, and SHM, ranging from hundreds to 30,000 participants] to represent information across the adult life span in both sexes. Also, two cross-sectional studies were used as comparison [EpiHealth, LifeGene]. PA was assessed by questionnaires on a four or five-level scale.
Results
Taking results from several samples into account, an increase in PA from middle-age up to 70 years was found in males, but not in females. Following age 70, a decline in PA was seen. Young adults reported both a higher proportion of sedentary behavior and a higher proportion high PA than the elderly. Females generally reported a lower PA at all ages.
PA was mainly associated with serum triglycerides and HDL-cholesterol, but also weaker relationships with fasting glucose, blood pressure and BMI were found. These relationships were generally less strong in elderly subjects.
Conclusion
Using data from multiple longitudinal samples the development of PA over the adult life span could be described in detail and the relationships between PA and cardiovascular risk factors were portrayed. In general, a higher or increased physical activity over time was associated with a more beneficial cardiovascular risk factor profile, especially lipid levels.
Citation: Lind L, Zethelius B, Lindberg E, Pedersen NL, Byberg L [2021] Changes in leisure-time physical activity during the adult life span and relations to cardiovascular risk factors—Results from multiple Swedish studies. PLoS ONE 16[8]: e0256476. //doi.org/10.1371/journal.pone.0256476
Editor: David Meyre, McMaster University, CANADA
Received: May 5, 2021; Accepted: August 6, 2021; Published: August 19, 2021
Copyright: © 2021 Lind et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: Due to Swedish laws on personal integrity and health data, as well as the Ethics Committee, we are not allowed to make any data, including health variables, open to the public even if made anonymous. The data could be shared with other researchers after a request to the steering committee [karl.michaelsson@surgsci.uu.se].
Funding: The authors received no specific funding for this work.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Based on observational data, there is a general agreement that a high leisure time physical activity [PA] is beneficial in terms of future CVD [, ]. The same is considered for a high cardiorespiratory fitness, a condition most often accompanying a high PA [, ]. However, high cardiorespiratory fitness is also dependent on biological and genetic factors that may be independent of physical activity []. In a large study including twins in different European countries, the heritability of PA in males and females was similar and ranged from 48% to 71% []. Such high heritability was confirmed in another twin study []. The estimated heritability of CVDs, like coronary heart disease, is somewhat lower [30–60%] [].
Two studies have used the Mendelian randomization [MR] approach to see if genes linked to PA also are linked to CVD, thereby suggesting a causal effect of PA on CVD. One of these studies showed a suggestive causal genetic effect of PA on coronary heart disease [], but not stroke, while no such associations were found in another study using MR []. It should however be remembered that the number and strength of the SNPs used as genetic instrument for PA is quite low and therefore the risk of false negative findings are high.
Long-term randomized trials with increased PA does not exist, but a community-based intervention study performed in the primary care setting over 9 months showed an improvement in blood pressure and lipids after 9 months, but also a significant reduction in incident CVD after 2 years [].
Generally, the amount of high PA declines with ageing, while the amount of sedentary behavior increases with age [–]. This pattern has been reported to be more pronounced in women compared to men []. However, most of those studies are cross-sectional and only a few longitudinal studies exist that are addressing the issue of change in PA over longer time periods [, ]. Thus, there is a need for more longitudinal studies covering the adult lifespan. Since it is hard to cover the whole adult lifespan in a single study, we have in this paper used multiple studies to reflect different parts of the lifespan in both men and women.
The positive impact of a high PA on CVD might well be due to positive effects on traditional risk factors for CVD, such as a better glucose control [], improved lipid levels [], and a lower blood pressure [], as documented in intervention studies. However, since the impact of these risk factors on future risk for CVD vary with age [], we hypothesized that also the relationship between PA and CVD risk factors varied in strength during the lifespan. Ideally, also in this case longitudinal data should be used in which the relationship between PA and CVD risk factors should be studied in the same sample at different ages.
Thus, the primary objective of the study was to evaluate how PA levels change over time and how this is dependent on age and sex. The second objective was to study how PA is related to traditional risk factors for cardiovascular disease, such as blood pressure, lipids, diabetes, obesity, and smoking in different age group and in men and women. For both of these objectives, we used several longitudinal study samples to cover both sexes and most of the adult lifespan. Thus, the major contribution of the present study to the already existing knowledge in this field is that we are able to study these two objectives in the longitudinal setting using several study samples.
As a complement to the longitudinal data, we also report results from a large cross-sectional study covering the age span from 20 to 75 years.
Methods
Samples
ULSAM [Uppsala Longitudinal Study of Adult Men] is a population-based study of all men aged 50 living in Uppsala during the years 1970–74. The participation rate was 82%. Re-examinations have been performed at ages 60, 70, 77, 82, 88, and 93. The two latest investigations are not used in the present analysis. Details on the study sample is given in [] [www.pubcare.uu.se/ULSAM]. Data on PA was present in 2291 of the subjects at baseline. See for n at the other investigations.
Download:
PPT
PowerPoint slide
PNG
larger image
TIFF
original image
Table 1. Description of studied variables in the different samples.
//doi.org/10.1371/journal.pone.0256476.t001
PIVUS [Prospective Investigation of Vasculature in Uppsala Seniors] is a population-based study of men and women aged 70 in Uppsala during the years 2001–2004. The participation rate was 50%. Re-examinations have been performed at ages 75 and 80. Details on the study sample is given in []. Data on PA was present in 1013 of the subjects at baseline. See for n at the other investigations.
EpiHealth [Epidemiology for Health] is a population-based study of men and women aged 45–75 in the cities of Uppsala and Malmö during the years 2011–2018. The participation rate was 20%. No re-examinations have been performed. Details on the study sample is given in [, ]. Data on PA was present in 24,703 of the subjects.
LifeGene is population-based study of men and women from the newborn to mid-age in the cities of Stockholm, Umeå, and Alingsås during the years 2009–2018. No re-examinations have been performed. Details on the study sample is given in []. Data on PA was present in 21,759 of the subjects.
SHE [Sleep and Health in Women] is a population-based study of women aged 20–99 years in Uppsala in 2000. The participation rate was 71.6%. Re-examination was performed in 2010. Details on the study sample is given in []. Data on PA was present in 6955 of the subjects at baseline. See for n at the other investigations.
SHM [Sleep and Health in Men] is a population-based study of men aged 30–69 years and living in Uppsala in 1984. The participation rate was 79.6%. Re-examinations were performed in 1994 and 2007. Details on the study sample is given in []. Data on PA was present in 2626 of the subjects at baseline. See for n at the other investigations.
The participants of each study gave written informed consent and the studies were either approved by the Regional Ethics Committee in Uppsala [ULSAM, PIVUS, EpiHealth, SHE, and SHM] or Stockholm [LifeGene], Sweden.
PA assessment
In all studies, a questionnaire was given to the participants who answered questions regarding leisure time PA. The questions were different across the studies [details given in ], but the results were presented in 4 [5 in EpiHealth and LifeGene] PA categories with 1 denoting a sedentary behavior and 4 [5 in EpiHealth and LifeGene] denoting an athletic lifestyle.
Risk factors
Blood samples were drawn in all studies after an overnight fast, except in EpiHealth in which only 6 hours of fasting was required. Plasma glucose, serum triglycerides, LDL-, and HDL-cholesterol were measured by standard techniques, see the references above for details. Blood pressure was measured in the supine position in ULSAM and PIVUS and in the sitting position in EpiHealth.
Smoking status, education level, and medications were obtained using questionnaires.
The metabolic syndrome [MetS] was defined according to the consensus criteria [], being a slight modification of the NCEP criteria. The number of the five different components were calculated.
Statistical analyses
Changes in PA over time/with age.
In all longitudinal samples the change in PA categories over time was assessed by mixed models for an ordinal outcome [command xtologit]. To evaluate if age or sex were related to the change over time in PA, interaction terms between time and age or time and sex were included in the models. All data from all examinations were used in the longitudinal analyses. In ULSAM and PIVUS, the analyses included the confounders education and smoking status [updated for each examination], and sex in PIVUS [ULSAM consisted of men only].
In the cross-sectional analysis, data from LifeGene and EpiHealth were merged and ordinal logistic regression [command ologit] was used to relate age and sex to the PA categories. An interaction term between age and sex was also included in the model, as well as the confounders alcohol intake, smoking status, and education level. For the corresponding figure, the sample was divided into six age-groups [20–29, 30–39, 40–49, 50–59, 60–69 and 70–75].
For the calculations of trajectories in PIVUS and ULSAM, we calculated group-based trajectories using a finite mixed model [command traj]. Maximum likelihood is used for the estimation of the model parameters. The maximization is performed using a general quasi-Newton procedure.
Cross-sectional relationships between PA and CV risk factors.
Fasting glucose and triglycerides were log-transformed to achieve normal distributions.
A trend test for PA used as a continuous variable vs six risk factors [one by one] were performed by linear regression [command regress] for each examination in PIVUS and ULSAM.
EpiHealth was divided into three age-groups [45–54, 55–64, and 65–75], and the analyses were performed in each age-group.
All analyses included the confounders education, smoking status, antihypertensive medication, antidiabetic medication, lipid-lowering drugs, and sex [in PIVUS and EpiHealth].
In the ULSAM study, we evaluated the role of BMI as a mediator in the PA vs risk factor relationship using structural equation models [SEM] using a maximum likelihood method. We then used data for PA and BMI from age 50, while for the outcomes HDL and triglyceride [TG] data from age 60 was used.
Longitudinal relationships between PA and CV risk factors.
In PIVUS and ULSAM, the relationships between the change in PA and the change in six risk factors [one by one] were performed using mixed models with a random intercept [command xtmixed]. The analyses included the confounders education, smoking status, antihypertensive medication, antidiabetic medication, lipid-lowering drugs [updated for each examination], sex [in PIVUS], and the baseline value of PA.
We also evaluated if the trajectories for PA identified in PIVUS and ULSAM were related to the risk factors during the follow-up period. For this task, we used mixed models [command xtmixed] where the first trajectory [see ] was treated as the reference group and the other trajectories for PA were evaluated vs this reference group.
Cross-sectional relationships between PA and MetS.
A trend test for PA vs MetS [binary] or the number of MetS components were performed by ordinal logistic regression [command ologit] for each examination in PIVUS and ULSAM. EpiHealth was divided in three age-groups [45–54, 55–64, and 65–75], and the analyses were performed in each age-group. The log odds of the beta coefficients are given.
All analyses included the confounders education, smoking status, and sex [in PIVUS and EpiHealth].
Longitudinal relationships between PA and MetS.
In PIVUS and ULSAM the relationships between the change in PA and the change in occurrence of MetS or number of MetS components were performed using mixed models for ordinal logistic regression [command xtologit]. The analyses included the confounders education, smoking status [updated for each examination], and sex [in PIVUS], and the baseline value of PA.
Calculations were performed using STATA16.1 [Stata inc, College Station, TX, USA].
Results
Characteristics of the populations are shown in , and the timing of the data collection in the populations is shown in .
Download:
PPT
PowerPoint slide
PNG
larger image
TIFF
original image
Fig 1. Overview of the different studies in relation to calendar time.
//doi.org/10.1371/journal.pone.0256476.g001
Changes in PA with age
The changes in the proportions of the PA-categories over time in ULSAM are given in . A slight increase in the PA-activity was seen over time [p = 0.00024] when adjusted for smoking and education level. This increase was most pronounced when comparing the 50 to 70-year time-span [p = 5.5e-11]. Thereafter, a decline in PA was seen.
Download:
PPT
PowerPoint slide
PNG
larger image
TIFF
original image
Fig 2. Longitudinal change in physical activity [PA] categories over time in a] ULSAM, b] PIVUS, c] SHE, and d] SHM and cross-sectional relationships between PA categories and age-groups in e] the EpiHealth study.
//doi.org/10.1371/journal.pone.0256476.g002
The changes in the proportions of the PA-categories over time in PIVUS are given in . The PA level declined over time in the PIVUS study [p = 1.3e-06] when adjusted for sex, smoking, and education level. No significant differences between men and women were seen regarding PA [p = 0.32]. The interaction term between time and sex was not significant [p = 0.46].
The changes in the proportions of the PA-categories over time in SHE are given in . The PA level declined over time in the SHE study [p = 1.2e-05] when adjusted for age. The interaction term between time and age was not significant [p = 0.84].
The changes in the proportions of the PA-categories over time in SHM are given in .
A highly significant interaction was seen between time and age [p = 2.6e-08]. In the two youngest age-groups, the PA level increased over time [p = 6.4e-03 and p = 3.4e-04, respectively], while on the contrary a reduction in PA over time was seen in the two oldest groups [p = 3.3e-03 and p = 3.6e-05, respectively].
The relationships between age and PA in the cross-sectional analysis in EpiHealth/LifeGene are given in . The physical activity level declined with age [p = 1.0e-11] and was lower in women then in men [p = 3.7e-09] following adjustment for alcohol intake, smoking, and education level. No interaction between age and sex was seen regarding PA [p = 0.54].
The results in the SHM study were stratified by age-group at the initial investigation.
Individual changes in PA over time
In the ULSAM and PIVUS studies, trajectories for PA based on individual changes were calculated [see ].
Download:
PPT
PowerPoint slide
PNG
larger image
TIFF
original image
Fig 3. Trajectories for physical activity [PA] in a] ULSAM and b] PIVUS. The proportion of subjects in each trajectory is given.
//doi.org/10.1371/journal.pone.0256476.g003
In ULSAM, four significant trajectories were identified [pPA at age 50 ->TG at age 60, PA was not a significant mediator of the effect of BMI on TG [p = 0.13].
Assuming a causal relationship being: PA at age 50 ->BMI at age 50 ->HDL at age 60 showed that BMI was a significant mediator [p = 0.049] accounting for 18% of the total effect of PA on HDL.
If we, on the contrary, postulated a causal relationship being: BMI at age 50 ->PA at age 50 ->HDL at age 60, PA was not a significant mediator of the effect of BMI on HDL [p = 0.18].
Longitudinal relationships between PA and CV risk factors
In ULSAM, the change in PA between 50 and 82 years was related to the change in 6 traditional risk factors [evaluated one by one], with smoking, antihypertensive treatment, statin use, and antidiabetic treatment and education level as confounders. The change in PA was significantly related to the changes in serum triglycerides and HDL-cholesterol. An increase in PA over time was related to a reduction in triglycerides [negative beta coefficient] and an increase in HDL-cholesterol [positive beta coefficient, see ].
Download:
PPT
PowerPoint slide
PNG
larger image
TIFF
original image
Table 2. Relationships between changes in cardiovascular risk factors and change in physical activity from age 50 to age 82 in ULSAM, and between age 70 and 80 in PIVUS.
//doi.org/10.1371/journal.pone.0256476.t002
Similar results were seen when only the 50 to 70-year time-span was used in the analyses, but in this case also the change in systolic blood pressure tended to be related to the change in PA in an inverse fashion [p = 0.034].
When the four PA trajectories described in were related to the risk factors, compared with trajectory 1 [those increasing their PA from a sedentary level], trajectories 3 and 4 [both being at higher levels of PA over time] showed significantly [p