Article Text

Download PDFPDF

Does a population-based multifactorial lifestyle intervention increase social inequality in physical activity? The Inter99 study
  1. M Aadahl1,
  2. L von Huth Smith1,
  3. U Toft1,
  4. C Pisinger1,
  5. T Jørgensen1,2
  1. 1Research Centre for Prevention and Health, Building 84/85, Glostrup University Hospital, DK-2600 Glostrup, Denmark
  2. 2Faculty of Health Sciences, University of Copenhagen, Copenhagen, Denmark
  1. Correspondence to Mette Aadahl, Research Centre for Prevention and Health, Building 84/85, Glostrup University Hospital, DK-2600 Glostrup, Denmark; metaad01{at}glo.regionh.dk

Abstract

Aim To examine the effect of a multifactorial lifestyle intervention on 5-year change in physical activity (PA) and to explore whether length of education had an impact on the effect of the intervention.

Methods Two random samples (high intervention group A, n=11 708; low intervention group B, n=1308) were invited for a health examination, assessment of absolute risk of ischemic heart disease and individual lifestyle counselling. The participation rate was 52.5%. High-risk individuals in group A were also offered group-based counselling on diet and PA and/or smoking cessation. High-risk individuals in group B were referred to usual care. All high-risk individuals were reinvited for examination and counselling after 1 and 3 years, and all participants were reexamined after 5 years. The control group (group C, n=5264, response rate 61.1%) answered a mailed questionnaire. Change in self-reported PA from baseline to 5-year follow-up was the main outcome. Level of education was classified as no vocational training, ≤4 years and >4 years. Data were analysed using longitudinal linear regression models with random intercepts.

Results In men, the high-intensity intervention had a beneficial effect on PA level after 5 years. The age- or time-related decrease in PA was approximately 30 min/week less compared to men in the control group (p<0.0001). Level of education had no significant impact on the effect of the intervention neither in men (p=0.39) nor in women (p=0.32).

Conclusion A population-based multifactorial lifestyle intervention did not influence social inequality in PA.

Keywords Lifestyle, Exercise, Randomised Intervention Study, Ischemic Heart Disease, Socioeconomic Position.

View Full Text

Statistics from Altmetric.com

Physical inactivity is a major risk factor for ischemic heart disease (IHD) and premature death1,,6; other factors are smoking7 and unhealthy dietary habits.8 Several studies have found that a healthy lifestyle, including regular leisure time physical activity (PA), is more prevalent among individuals with high socioeconomic position (SEP) than among individuals and groups of low SEP.9,,12 Individual, social and environmental factors may all contribute to the social inequality in PA level.13 14

Randomised, controlled trials have found modest effects of multifactorial lifestyle interventions on biological IHD risk factors,15,,22 whereas the long-term effect on PA level has been investigated in a limited number of studies only.16 18 20 22

The Inter99 study is a large-scale Danish randomised multifactorial intervention study aimed at reducing the incidence of IHD in the general population through increased PA, improved dietary habits and smoking cessation using a high-risk approach.23,,25 After 5 years with repeated individualised and group-based counselling, a beneficial effect of the intervention on dietary habits26 and smoking cessation rates27 was observed. After 3 years the intervention was found to increase PA level among men but not among women.28

Ideally, a lifestyle intervention based on a high-risk strategy should reduce social inequality in PA. However, given that individuals with high SEP may respond better to health education messages, lifestyle interventions, such as the Inter99, involve a risk of increasing the social inequality in PA. The aim of the present study was to investigate whether a multifactorial lifestyle intervention is effective in increasing long-term PA and whether level of SEP has an impact on the effect of the intervention. Little is known about behavioural changes in different socioeconomic groups, and to our knowledge, this research question has not previously been addressed in a population-based intervention study.

Methods

Study population and design

The Inter99 study is a population-based intervention study initiated in March 1999 and terminated in April 2006. The aim of the study was to prevent cardiovascular disease and type 2 diabetes by non-pharmacological intervention. The study focused on improvements in smoking, diet and PA habits (only to a small extent on avoidance of excessive alcohol consumption), and was performed at the Research Centre for Prevention and Health, Glostrup, Denmark. The study was approved by the local ethics committee (KA 98155) and registered at Clinical Trials.gov (NCT00289237).

The study design is described in detail elsewhere23 (http://www.Inter99.dk). In short, the study population consisted of all 61 301 individuals born in 1939–1940, 1944–1945, 1949–1950, 1954–1955, 1959–1960, 1964–1965 and 1969–1970 and living in the southwestern part of Copenhagen county. An age- and sex-stratified random sample of 13 016 individuals was drawn from the study population, with most of the individuals aged between 40 and 50 years. Before invitation the sample was prerandomised into group A (high-intensity intervention) and group B (low-intensity intervention). The invitation included a questionnaire to be completed before the start of the intervention. Eighty-two persons had died or could not be traced. Of the remaining 12 934, a total of 6906 (53.4%) participated in the study. Of these, 122 were excluded because of alcoholism, drug abuse or linguistic barriers, leaving 6784 (52.5%) for analyses in groups A and B. From the remaining 48 285 individuals, a random sample of 5264 individuals was drawn for a non-intervention control group (group C). Out of these, 3324 (61.1%) answered a postal questionnaire. Contrary to the sampling of groups A and B, where 40-, 50- and 55-year-olds were over-represented, group C had equal representation of age groups. Written informed consent was obtained from all participants.

All participants in group A and group B underwent a health examination, an assessment of their risk of IHD and individual lifestyle counselling. Absolute risk of IHD within the next 10 years was estimated by the Copenhagen Risk Score using the computer program PRECARD.29 Individuals were classified as high-risk persons if they were in the upper quintile of the distribution stratified according to sex and age, or if they had one or more of the following risk factors: being a daily smoker, systolic blood pressure ≥160 mm Hg or in antihypertensive treatment, total cholesterol ≥7.5 mmol/l, body mass index ≥30 kg/m2, diabetes or impaired glucose tolerance. Numbers of participants at baseline, 1-, 3- and 5-year follow-up are presented in the Inter99 flow sheet in fig 1.

Figure 1

Inter99 study population and flow sheet. Single asterisk denotes 714 persons participated at least once in diet/physical activity group counselling. Double asterisks denote random sample of the reference population, group C.

Intervention

The individual lifestyle counselling was based on the personal risk estimate and the participant's motivation for change, and comprised a lifestyle counselling consultation focusing on smoking, PA, diet and/or alcohol. The counselling was based on aspects from the following theories: Health Belief Model,30 the Social Cognitive Theory31 and the Transtheoretical Model32 focusing on individual perceived health risk and perceived benefits and barriers, self-efficacy and individual readiness for behavioural change. Trained nurses, dieticians and doctors performed the motivational interview and the consultation proceeded as a dialogue, respecting the wishes of each participant. The consultation lasted 15–45 min, depending on the risk status and motivation of the participant.

Participants were advised to aim for at least 30 min of moderate-intensity PA per day in accordance with current public health recommendations.

In addition to the individualised lifestyle counselling, participants at high IHD risk in group A were offered group counselling on diet/physical activity (DPA) and/or smoking cessation or reduction. High-risk persons in group B were referred to usual care—for example, to their general practitioner. The DPA groups were lead by a nurse or a dietician and consisted of 15–20 participants who met six times for 2 h during a 6-month period. The theoretical basis was similar to that of the individual counselling. The participants set specific goals for behavioural change together with the staff. The aim was to increase PA in everyday life. Participation in group counselling is described in detail elsewhere.25

Follow-up

Individuals in groups A and B, who were at high risk at baseline, were reinvited after 1 year to a health examination and an individualised lifestyle counselling. Persons still at high risk in group A were again offered group counselling and high-risk persons in group B were referred to usual care. Low-risk persons in groups A and B and all in group C were sent a questionnaire. This was repeated again after 3 years. At 5-year follow-up, all baseline participants were invited for a health examination and a short finishing lifestyle counselling.

Outcome measures

PA was measured with a self-administered questionnaire at baseline, 1-, 3- and 5-year follow-up. It was based on two questions: (1) in your leisure time, “how many hours a week are you physically active? (including walking, cycling, gardening, but excluding transportation to and from work)” (answer categories: 0 min, ∼½ h/week, ∼1 h/week, ∼2–3 h/week, ∼4–6 h/week, and 7 h/week or more) and (2) “how much time do you spend walking, cycling or running on your way to and from work?” (answer categories: less than 15 min/day, 15–30 min/day, 30 min to 1 h/day, 1 h or more/day, and I do not work at the moment). Commuting PA was calculated as minutes per week using a 5-day working week and responses to the two PA questions were summed into one PA variable.28 33 Five-year changes in PA has been found to reflect the corresponding changes in biological cardiovascular disease risk factors among Inter99 participants.34

Covariate assessment

The covariates were measured with a self-administered questionnaire at baseline and are shown in table 1. SEP was defined by length of education and was categorised in three levels: (1) short education, corresponding to no vocational training; (2) medium-length education, corresponding to ≤4 years of vocational training; and (3) long education, representing >4 years of vocational training.

Table 1

Baseline characteristics according to randomisation group in the Inter99 study population (N=10 108)

The three-class validated dietary quality score was developed from a 48-item food frequency questionnaire by Toft et al.35 Participants were classified as eating a healthy, an average or an unhealthy diet, based on the intake of fruit, vegetable, fish and fat. Alcohol intake was based on average weekly consumption of normal beer, strong beer, wine, fortified wine, schnapps or other spirits within the last 12 months. Smoking status was categorised as being a daily smoker (yes or no). Functional limitation was measured as being limited in climbing several flights of stairs because of one's health. Living with a partner was defined as being married or cohabiting. Employment status was dichotomised as being currently employed (yes/no), and self-rated health was categorised into three groups: very good, good or poor.

Statistical analysis

Data analyses were performed using multilevel regression analyses with repeated measurements. We used the SAS Proc Mixed procedure with normally distributed random intercepts (SAS statistical software, V.9.2; SAS Institute, Cary, North Carolina, USA). To determine the effects of the lifestyle intervention on PA level across time, an interaction term between intervention group and time was included. When exploring whether level of vocational training had an impact on the effect of the intervention on PA level across time, a three-way interaction term composed of level of vocational training, intervention group and time was included. The models were adjusted for baseline age, diet, smoking, alcohol intake, self-rated health, being limited in stair-climbing, living with partner and current employment status. These covariates were selected, based on (1) being differently distributed between the intervention and the control group at baseline and/or (2) being differently distributed at baseline between participants and dropouts at 1-, 3- and 5-year follow-up. For comparison of baseline characteristics between intervention groups, age-adjusted Cochran–Mantel–Haenszel tests and age-adjusted linear regression models were used because group C was sampled with a larger proportion of individuals in the lower and upper age groups compared to groups A and B. Multiple logistic regression analysis was used to compare responders with non-responders at 1-, 3- and 5-year follow-up.

The total model sample size in the longitudinal analyses was 40 432 observations from 10 108 subjects at baseline, 1-, 3- and 5-year follow-up. Missing information on covariates reduced the sample size, leaving 27 131 observations from 8405 subjects for analyses in multiadjusted models. All analyses were stratified by sex and p=0.05 was chosen as the significance level.

Results

Baseline characteristics of participants across groups of randomisation are presented in table 1. Women in group B were more physically active than women in groups A and C, whereas we found no statistically significant difference in PA level among men.

At baseline, women with short education had a lower PA level compared to women with ≤4 years and >4 years of education (fig 2). The finding was statistically significant in groups A (p=0.008) and C (p=0.01), but not in group B. Among men, the same pattern was seen, but the differences in PA across groups of education were not statistically significant (fig 2).

Figure 2

Baseline physical activity for groups with different length of education in the three randomisation groups. Mean and 95% CI. p Values are age-adjusted.

PA decreased from baseline to 1-, 3- and 5-year follow-up in both sexes (fig 3). Among men, the decrease was approximately 30 min/week less in group A compared to control group C (p<0.0001) and approximately 20 min/week less for group A, when compared to group B at 5-year follow-up (p=0.035). There was no significant difference between groups B and C. Among women, there were no significant differences in 5-year change in PA between any of the randomisation groups.

Figure 3

The change in physical activity from baseline to 1-, 3- and 5-year follow-up in longitudinal analyses. The results are adjusted for baseline age, diet, smoking status, alcohol intake, self-rated health, being limited in stair-climbing, living with partner, length of education and current employment status.

Change in PA from baseline to 5-year follow-up for the three education groups in randomisation groups A, B and C are presented in fig 4. In the multilevel analyses including the three-way interaction term with time, randomisation group and length of education, no significant interaction was found in men (p=0.39) or in women (p=0.32); that is, the effect of the intervention on change in PA over the 5-year intervention period did not differ significantly among groups with different length of education (fig 4).

Figure 4

The change in physical activity from baseline to 5-year follow-up for groups with different length of education in the three randomisation groups. The results are adjusted for baseline age, diet, smoking status, alcohol intake, self-rated health, being limited in stair-climbing, living with partner, length of education and current employment status. p=0.32 for difference among education groups in effect of the intervention on change in physical activity in women. p=0.39 for difference among education groups in effect of the intervention on the change in physical activity in men.

Discussion

In the present study, we found a significant 5-year beneficial effect on PA level of the multifactorial lifestyle intervention in men. The (age- or time-related) decline in PA observed in the no-intervention control group was attenuated over the 5-year intervention period in the high-intensity intervention group in men, whereas no effect of the intervention was observed among women. There was no significant effect of the low-intensity intervention in neither men nor women.

The intervention did not seem to have a different impact on groups of varying lengths of education, neither in men nor in women. In other words, the high-risk strategy used in the Inter99 study did not seem to increase the social inequality in PA level observed at baseline.

Our findings on the overall effect at 5-year follow-up are in accordance with results from the 3-year follow-up of the Inter99 study, reported in a prior publication.28 At the 3-year follow-up, there was no significant difference in change in PA among men, between the high-intensity intervention group (group A) and the low-intensity intervention group (group B). The conclusion then was that a low-intensity intervention with referral to usual care had the same impact on change in PA among men, as did an offer of an extensive DPA group counselling programme. However, our results after the 5-year follow-up suggest that the high-intensity intervention is more beneficial.

In our study, the lack of intervention effect on change in PA among women seems consistent, also after 5 years. This contrasts findings from the Tromsø Study15 that only found an effect in women, but it corresponds to findings in the OXCHECK Study,17 where a modest effect in men but no effect in women was seen. The two studies used crude self-report measurement of PA, repeated interventions and had follow-up time of 3 and 6 years, respectively. Other multifactorial intervention studies aiming at preventing IHD also show small to modest effects on PA18 20 or physical fitness,22 whereas one study found no effect on PA.19

To our knowledge, the impact of a multifactorial lifestyle intervention has not been evaluated among different SEP groups before. We found that the intervention effect did not vary with length of education. Hence, the high-risk strategy used in the Inter99 study did not seem to affect the social inequality in PA either way. SEP has been identified as a strong and consistent correlate of PA,9,,14 and education-based disparities in PA are apparent in cross-sectional data.9 11 13 14 In a recent observational study by Shaw et al,36 the age-related rates of decline in self-reported PA differed by educational level in a population of “early old age” (54–72 years of age), community-dwelling Americans. The decline in PA was steeper among individuals with low education. In our study we did not observe a similar significant difference in PA decline or a difference in intervention effect among participants with different levels of education. Considering that we found no overall effect of the intervention on PA level in women and a rather modest effect in men, this finding is not unexpected. However, SEP is still an important factor to consider when choosing future strategies for promotion of PA. Reduction of social inequality in lifestyle, including PA, is an important challenge and a necessary target for future public health interventions. In the present study, there is likely to be social inequality in recruitment of study participants at baseline, that is, individuals with no vocational training are likely to be over-represented among non-participants. If so, the Inter99 study may indirectly have caused an increase in social inequality. This emphasises that a high-risk strategy may not be sufficient when attempting to combat social inequality in lifestyle and health. A structural prevention strategy theoretically targets everyone, regardless of SEP.37 Therefore, a combination of high-risk and structural prevention strategies should be considered in the future.

The concept of SEP is complex and the measurement of SEP is equally challenging.38 39 Different measures can be chosen and each measurement has different strengths and weaknesses. In general, there is no single best indicator of SEP. However, we chose to define SEP by length of education. First, length of education is relatively easy to measure by self-report and therefore provide relatively valid measurements compared to—for example, self-reported income. Second, education is recommended as a SEP indicator when exploring behavioural and knowledge-related aspects of health education messages.38 39 This is based on the assumption that the knowledge and skills attained through education may affect a person's cognitive functioning, making them more receptive to health education messages.38 39

The present study has several strengths and limitations that should be addressed. The long follow-up time and the repeated intervention made longitudinal analyses possible, which is an important strength of the study. Loss to follow-up is a major concern in large-scale intervention studies such as ours. However, using multilevel regression models with random effects enabled us to take account of the loss to follow-up under the assumption of missing at random. We realise that the results may be biased if the missing at random assumption is false. This could be the case if the change in PA among those who drop out depends systematically on factors that we have not measured. However, the fact that non-response could be related to measured baseline characteristics was taken into account by inclusion of covariates that were differently distributed at baseline between participants and dropouts at 1-, 3- and 5-year follow-up.

Finally, the self-report nature of the PA measurement involves a risk of bias. It is well-known in the general population that PA is beneficial for health. Therefore, possible overestimation of PA is likely to be equally distributed across randomisation groups and hence would not have caused differential misclassification.

In conclusion, the population-based multifactorial lifestyle intervention applied in the present study did not seem to increase social inequality in PA. However, the social inequality in PA level at baseline was not reduced either. Because social inequality in physical inactivity and other lifestyle factors—for example, smoking, contribute substantially to social inequality in health, future lifestyle interventions should consider structural prevention strategies, either alone or in conjunction with high-risk strategies, to combat social inequality in health. A combination of high-risk strategy and structural or environmental strategies has the potential for reaching the entire population regardless of level of education. This may require comprehensive, multilevel interventions targeting activity-related attitudes and skills as well as social and physical environments.

What this study adds

  • Compared to a no-intervention control group, a population-based multifactorial lifestyle intervention had a beneficial long-term effect on physical activity level in men but not in women.

  • The lifestyle intervention did not seem to increase social inequality in physical activity, but the social inequality found at baseline was not reduced either.

What is already known on this subject

  • A healthy lifestyle, including regular leisure time physical activity, is more prevalent among individuals with high socioeconomic position than among individuals and groups of low socioeconomic position.

  • The long-term effect of multifactorial lifestyle interventions on physical activity level has been investigated in a limited number of studies only.

Acknowledgments

The Inter99 was initiated by TJ (PI), Knut Borch-Johnsen (co-PI), Hans Ibsen and Troels F Thomsen. The steering committee is composed of the former two and Charlotta Pisinger. The study was financially supported by research grants from the Danish Research Council, the Danish Centre for Health Technology Assessment, Novo Nordisk Inc., Research Foundation of Copenhagen County, Ministry of Internal Affaires and Health, The Danish Heart Foundation, The Danish Pharmaceutical Association, the Augustinus Foundation, the Ib Henriksen Foundation, and the Becket Foundation.

References

View Abstract

Footnotes

  • Competing interests None.

  • Ethics approval This study was conducted with the approval of the local ethics committee of Copenhagen County (KA 98155).

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.