Background To describe trends in cardiovascular risk factors and change over time across education levels, and study the influence from medicine use and gender.
Methods Data from participants (30–74 years) of the Tromsø Study in 1994–1995 (n=22 108) and in 2007–2008 (n=11 565). Blood samples, measurements and self-reported educational level and medicine use were collected.
Results Differences in risk factor levels across education groups were persistent for all risk factors over time, with a more unfavourable pattern in the lowest education group. The exception was cholesterol, with the reduction being largest in the lowest educated, resulting in weakened educational trends over time. While a significant educational trend in cholesterol persisted among the non-users of lipid-lowering drugs (LLD), no educational trend in cholesterol was found among the LLD users in 2007–2008.
The strongest educational trends were found for daily smoking and Body Mass Index (BMI). In 2007–2008 the odds for being a smoker were five times higher among the lowest educated compared to the highest educated. In men, the odds for being in the highest quintile of the BMI distribution were, in 2007–2008, almost doubled in the lowest compared to the highest educated. The lowest educated women had 6.2 mm Hg higher mean systolic blood pressure than the highly educated, mean BMI of 26.4 kg/m 2 and smoking prevalence of 37.7%.
Conclusions The difference across education groups for cholesterol levels decreased, while the educational gap persisted over time for the other risk factors. Use of LLD seemed to contribute to the reduction of social differences in cholesterol levels.
- Cardiovascular disease
- Health inequalities
- Cohort studies
Statistics from Altmetric.com
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.
Despite an overall decline in cardiovascular mortality, relative differences in mortality between educational groups have widened over the last five decades. This is due to a stronger decrease in mortality in the highest as compared with the lowest education group. Cardiovascular disease in men, and lung cancer and chronic obstructive pulmonary disease in women were important causes for the observed differences in mortality.1 In the large European Prospective Investigation into Nutrition and Cancer study, total mortality was 43% lower among men and 29% lower in women with the highest attained education as compared to those men and women with the lowest education.2
About half the mortality decrease in coronary heart disease between the 1980s and the turn of the century was attributable to favourable changes in risk factors.3 ,4 The influence of medicines has been a part of the discussion.3–7
Cardiovascular risk factors are inversely correlated with socioeconomic status (SES).8–15 Inequalities in cardiovascular risk factors, particularly smoking and overweight, have a major impact on inequalities in cardiovascular morbidity and mortality.16 ,17 Norway has free schools and universities, low unemployment and high average income per capita compared to other European countries. Every inhabitant has, in principle, the same right and access to health services. However, while Mackenbach et al18 concluded that social inequalities are not smallest in the equality-focused Nordic countries, Popham et al19 concluded that the Nordic countries had the smallest inequalities in mortality for men, but not women. Steingrimsdottir et al20 showed that the probability of surviving to the age of 64 years did not improve from 1961 to 2009 for Norwegian women with primary education only. Norwegian studies have shown that cardiovascular risk factors had divergent patterns when it comes to changes in educational trends over time.8 ,21 ,22 More information is needed about whether or not the socioeconomic inequalities persist or change over time.
The aim of this paper was to analyse differences in cardiovascular risk factors across education groups, whether educational differences in risk factor levels have changed over time, and whether any such changes were related to gender or use of medicines.
The Tromsø Study is a population-based, prospective study of various health issues and chronic diseases, and is a resource for the surveillance of disease risk factors. It consists of six surveys referred to as Tromsø 1–6, which have been conducted in the municipality of Tromsø from 1974 to 2008. Eligible for the present study were those who participated in Tromsø 4 in 1994–1995 (n=27 158) and/or Tromsø 6 in 2007–2008 (n=12 984).23 ,24 The attendance rates were 72% and 66% in Tromsø 4 and Tromsø 6, respectively. The present analyses are limited to subjects aged 30–74 years with known educational status (Tromsø 4, n=22 108, Tromsø 6, n=11 565; 7822 attended both surveys). Due to missing values for some variables, the subjects included in the analyses may differ slightly from these numbers. The data collection is described in detail elsewhere.24 An English translation of the questionnaires is available at the Tromsø Study homepage.25
The highest attained level of education was collected through self-report in a questionnaire and classified as follows: (1) primary/partly secondary education (up to 9 years of schooling); (2) upper secondary education (10–12 years of schooling); (3) tertiary education, short (college/university less than 4 years); (4) tertiary education, long (college/university 4 years or more).
Smoking status was defined as the participants filling in ‘yes/no’, to the question ‘Do you smoke daily?’ Height (cm) and weight (kg) were measured to the first decimal in participants who wore light clothing and no footwear on an automatic electronic scale, the Jenix DS 102 stadiometer (Dong Sahn Jenix, Seoul, Korea). Body Mass Index (BMI) was calculated as weight (in kilogramme)/height2 (in metres). Systolic blood pressure (systolic BP) was measured using an automated device, the Dinamap Vital Signs Monitor 1846, Criticon (Tromsø 4) and Dinamap Pro care 300 Monitor, GE Healthcare, Norway (Tromsø 6). The cuff was chosen after the circumference of the upper arm was measured. Three readings on the upper right arm were taken in a sitting position, separated by 1 min intervals, and after 2 min of rest. The average of the last two measurements was used in the analyses. Non-fasting blood samples were collected. Venipuncture was performed with subjects in a sitting position. A light tourniquet was used and released before sampling. Serum total cholesterol was analysed within 10 h by an enzymatic colorimetric method in both surveys. Total serum cholesterol is referred to simply as cholesterol in the text. Current use of antihypertensives (AHT) and lipid-lowering drugs (LLD) (yes/no) was assessed through self-report (questionnaire). Participants were also asked to write a list of brand names of medicines used on a regular basis. The questionnaire information was checked by health personnel at the study site.
Data analyses were performed with SAS V.9.3 (SAS Institute, North Carolina , USA). Most analyses were stratified by sex. Descriptive characteristics were reported with means (SD) for continuous variables and proportions for binary variables. Age-adjusted means, proportions, and ORs between educational groups, test of linear trend over education, and test of differences in educational trends between Tromsøs 4 and 6 were assessed using multivariable linear mixed models and generalised estimating equation models with the logit link function. The ORs were estimated for being in the highest quintile, or being above specified threshold values for BMI, total cholesterol and systolic BP, or being a smoker. The models included attained age, indicator variable for survey (Tromsø 4 or 6), education (as an ordinal variable or as indicator variables for each level of education), and two-way cross-product terms between survey and education. To account for dependencies between repeated measures, all models included an exchangeable covariance structure. The proportions of AHT and LLD users in 1994–1995 and 2007–2008 were age-adjusted using the direct method with the Norwegian population as of 1 January 2008 as standard population. There were no LLD users among women aged 30–39 years, therefore, we age-adjusted the proportions of LLD users among participants aged 40–74 years. All p values were two-sided using a 5% significance level.
The Tromsø Study was approved by the Data Inspectorate of Norway and the Regional Committee of Medical and Health Research Ethics, North Norway. Participation was voluntary and each subject gave written informed consent prior to participation.
The educational status increased over time in both genders, more in women compared with men (see online supplementary table S1). The proportion of smokers, mean systolic BP and cholesterol decreased, while mean BMI increased in all sex, age and education groups from 1994 to 2008 (data not shown).
Daily smoking was strongly inversely associated with education, and the educational trends were significant in both sexes and surveys. The difference between the surveys regarding the trend in the prevalence in daily smoking across education groups was borderline significant in men (p=0.054) and not significant in women (tables 1 and 2). The odds for being a smoker among the lowest educated compared to the highly educated group was about five times higher in both sexes in 2007–2008 (table 2). The prevalence of smoking decreased over time in all education groups. Relatively speaking, the prevalence decreased more in men than in women, and among the highly educated compared to lowest educated (table 3). The educational trends in daily smoking increased in all age groups in the time-period, but this was significant in men aged 45–59 years only (see online supplementary table S2).
Body Mass Index
BMI was inversely associated with education, and the educational trends were significant in both sexes and surveys (table 1). There was no change between the surveys regarding differences in BMI across education groups. However, in women, the percentage of obesity showed a reduction in educational trend between 1994–1995 and 2007–2008 (tables 1 and 2). BMI and obesity increased in all education groups in the time period, and in 2007–2008 approximately 20% of the lowest educated were obese. The educational trends in BMI increased in the time-period in men aged 45–59 years (see online supplementary table S3).
Systolic BP was inversely associated with education in both sexes and surveys (table 1). The difference in systolic BP between the lowest and highest educated (2007–2008) was 6.2 and 2.6 mm Hg in women and men, respectively. In 2007–2008, the odds for being in the highest quintile of the systolic BP distribution was about 39% (women) and 18% (men) lower in the highest compared to the lowest educated group (table 2). The changes in the educational trends systolic BP in the time period were modest and insignificant in all age groups (see online supplementary table S4).
The prevalence of AHT use increased in all sex and age groups in the period (table 4). In the male AHT users, there was no significant educational trend in systolic BP in any of the surveys, and no change in educational trend over time (table 1). In male AHT non-users, we observed significant and persistent educational trends in systolic BP in both surveys.
In the female AHT users, the significant educational trend in systolic BP observed in 1994–1995 was not found in 2007–2008. In female AHT non-users, there were significant educational trends in both surveys and a significant change between the two surveys regarding the differences in systolic BP across educational groups (p≤0.001). However, in 2007–2008, the differences in trends in systolic BP between the AHT users and non-users were not significant in men (p=0.88) or in women (p=0.18).
Systolic BP was higher in AHT users compared with non-users, and the decrease in systolic BP over time in the users was considerable, especially in women (table 3).
Total serum cholesterol
Cholesterol was inversely associated with education in both sexes and in both surveys. The educational trend in cholesterol weakened significantly between 1994–1995 and 2007–2008 in both sexes (table 1). In women, the educational trends in cholesterol tended to become weaker in the time period in all age groups, however, this was significant only in women aged 60–74 years (see online supplementary table S5).
Use of LLD was practically non-existent in 1994, but the prevalence of LLD use increased considerably between the surveys (table 4). The LLD brands were dominated by statins in 1994–1995 (83%) and 2007–2008 (99%). In 1994–1995, cholesterol was higher in LLD users compared to the non-users in all education groups, but this was reversed in 2007–2008 (table 1).
In the LLD users, the educational trends in cholesterol weakened significantly over time in both sexes (p=0.001). No educational trend in cholesterol was observed in users in 2007–2008 in neither men nor women (table 1). However, the educational trends weakened significantly over time in the LLD non-users, too, but the educational trend persisted in both sexes (p<0.001). In 2007–2008, the differences in trends between the LLD users and non-users were significant in women (p<0.001), but of borderline significance in men (p=0.055).
The educational trends were persistent for all risk factors over time from 1994–1995 to 2007–2008, except for cholesterol, where the educational trends were weakened over time. The strongest educational trends were found for daily smoking. Use of LLD seemed to contribute to the reduction of social differences in cholesterol level.
SES includes occupation, income and education. The highest attained education level has shown to be a good proxy for SES.26 Education level was chosen as an indicator of SES because it is an important determinant for type of work and economic circumstances, available for all participants regardless of employment status, and may be an indicator for participation in health-promoting activities.
Higher rates of daily smoking have been found among the lower-educated women in the Nordic countries, while the opposite has been found in Southern Europe18 ,27 Studies have shown an increasing educational gap.12 ,15 ,21 In the time period, a smoking ban law which has prohibited smoking in public places, public transportation and working places, was introduced in Norway, and tobacco prices have been very high. We observed strong and persistent differences in smoking prevalence across educational groups. Although the percentage reduction in the prevalence of daily smoking was largest among the highly educated, especially in men, the absolute reduction in the prevalence of daily smoking was largest among the lowest educated, suggesting a larger reduction in the absolute risk of smoking-related diseases in this group of subjects.
The gender pattern of daily smoking changed between 1994–1995 and 2007–2008. More women than men were smokers in 2007–2008, and while almost 38% of the lowest educated women were daily smokers, less than 10% of the highly educated men were so. Smoking is a more important risk factor for coronary heart disease in women compared to men (RR=1.25).28 As smokers may lose at least 10 years of lifespan, this difference in smoking pattern is an important finding for understanding differences in mortality.29
Molarius et al9 found that low education was associated with higher BMI in about half the male and almost all the female populations. The differences increased over the 10-year study period. The estimate is an increase in BMI in western Europe of 0.4–0.5 kg/m2 per decade, with substantial differences across regions and sexes.30 We observed a considerable increase in BMI in all educational groups, and an inverse relationship between SES and being overweight as in other studies.11 ,13 ,14 ,31 In women, the inverse relationship between the prevalence of obesity (BMI ≥30 kg/m2) and educational level was attenuated from 1994–1995 to 2007–2008.
In the total sample, we found no change in the educational trends in systolic BP over time, confirming the inverse relationship between systolic BP and education.10 ,12 ,14 ,21 Educational trends were more evident in women compared to men, as reported by others.21 The difference of 6.2 mm Hg in systolic BP across the educational groups in women is important with regard to cardiovascular risk and difficult to explain. The reduction of mean systolic BP in AHT users in the period is substantial, especially in female AHT users, probably due to stricter European guidelines for treatment of cardiovascular disease published in 2003.32 The educational gap in systolic BP increased among female AHT non-users; on the other hand we observed no significant relationship between educational level and systolic BP in AHT users in 2007–2008.
Our results support an inverse relationship between cholesterol level and education.7 ,12 ,33 While Kanjilal et al12 reported stable trends, others showed a more dynamic situation where the educational gap increased in one period and decreased in others.33 Our study showed that the educational gap in cholesterol diminished in users and non-users of LLD in the 15-year period. In 2007–2008, no educational trend in cholesterol was found in the users of LLD, indicating that LLD treatment contributes to the reduction of educational differences in distribution of the cholesterol risk factor. The educational trends persisted significantly among non-users of LLD. The considerable reduction in cholesterol concentration among drug users in the period is probably due to introduction of more extensive treatment guidelines32 and a generous reimbursement system. Selmer et al34 reported a tendency to higher statin treatment prevalence among highly educated subjects with a history of cardiovascular disease or diabetes compared to people of lower educational level, even after adjustment for other cardiovascular disease risk factors, particularly in women. Socioeconomic differences in statin use were also observed by others.35
Several authors6 ,7 ,33report that the reduction in cholesterol levels has been driven by dietary change alone rather than pharmacological treatment. Others argue that the decrease in cholesterol is partly due to better screening and pharmacological treatment in high-income regions compared to low-income regions.5 Dietary change is a factor explaining the observed reduction in cholesterol in the period, for example, reduction in the intake of saturated fat, transfat, and higher consumption of polyunsaturated fats and fish oil.36 Transfatty acids were removed from all margarines in Norway in 1998, probably contributing to a favourable reduction in all education groups. Norway had been one of the countries with the highest intake of transfatty acids in Europe, and this has been reduced to a recommended level.37 Johansson et al31 showed an unexpected rise in cholesterol in the population, probably due to increased fat consumption related to new eating habits. The positive reduction in cholesterol we observed may be reversed, and it is still important to monitor the cholesterol level in the population in the future.
The strengths of our study are the population-based design, high attendance, and repeated data collection over a long time period. However, there are some limitations. For the age groups 30–44 years and 60–74 years, different birth cohorts are compared between Tromsø 4 and Tromsø 6. Thus, it is not possible to separate the effect of period from that of birth cohort. There were few old subjects with a university education, especially among women and in Tromsø 4. This may have led to unstable point estimates. Smoking data was self-reported and, therefore, subject to the respondent's desire to falsely report sociably acceptable behaviour. Use of medication was based on questionnaire data about specific groups of drugs and list of brand names of current medication controlled by trained health personnel. The participants may still have failed to report. Self-report of medication used regularly for chronic conditions, such as LLD and AHT, is good.38 A comparison of data from Tromsø 6 with information from a national prescription database (time-window 180 days) showed a κ value of 0.940 (95% CI 0.932 to 0.949) and 0.952 (95% CI 0.856 to 0.876) for LLD and AHT, respectively, (AE Eggen, unpublished data).
Participants in epidemiologic studies tend to be female, married or cohabiting, and employed. They have higher SES, are more educated, tend to be non-urban residents, and live a healthier lifestyle.39 ,40 Therefore, we may be missing information from young, unmarried, less educated men who also tend to have unhealthier lifestyle, and non-participation may lead to an overestimation of, for example, the reduction in smoking prevalence.
In conclusion, we found a decrease in difference across education groups for cholesterol levels, while the educational gap persisted over time for the other cardiovascular risk factors. Use of preventive medicines seemed to contribute to the reduction of social differences in distribution of the related risk factors.
What is already known on this subject
Studies have shown an increasing socioeconomic gap in cardiovascular disease mortality, and most of the excess cardiovascular disease mortality in lower socioeconomic groups can be explained by known risk factors. Studies of socioeconomic differences in cardiovascular disease risk factors show a complex picture, and we need to know whether the educational gap in cardiovascular risk factor levels is changing over time.
What this study adds
The educational trends were persistent for all risk factors over time, except for cholesterol, where the educational trend was attenuated. The educational gap persisted significantly among the non-users of lipid lowering drugs (LLD), but not among LLD users, indicating that LLD treatment contributed to the reduction of social differences in cholesterol levels. Use of antihypertensives also tended to show the same pattern on the systolic blood pressure level. Daily smoking and body mass index showed the strongest educational trends.
We thank the technical staff who examined the participants, and the National Screening Services (SHUS) for its contribution to the data collection in the fourth Tromsø Study. Above all, we thank the residents of Tromsø; their willingness to participate was the largest resource in our epidemiological research.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
Contributors AEE took part in the data analyses and interpretation of data and wrote the first draft of the manuscript. TW did the statistical analyses and drafted the tables. EBM, IN, BKJ, and TW took part in the interpretation of data and critically revised the manuscript for important intellectual content. All authors read and approved the submitted version of the manuscript.
Funding Tromsø 6 was mainly funded by the University of Tromsø and The Northern Norway Regional Health Authority Medical Programme, and supported by grants from The Norwegian Research Council.
Competing interests None.
Patient consent Obtained.
Ethics approval Data Inspectorate of Norway and the Regional Committee of Medical and Health Research Ethics, North Norway.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement Available variables from the Tromsø Study (Tromsøs 1–6) may be viewed at the website http://www.tromsostudy.com.