Methods

PARTICIPANTS

The design of the Danish WHO MONICA surveys has been described in detail previously.13 Briefly, random equal sized samples of 30, 40, 50, and 60 year old men and women living in the south western part of Copenhagen County were drawn from the National Central Person Registry and invited to participate in surveys conducted in 1982–84 (DAN-MONICA I), 1987 (DAN MONICA II), and 1991–92 (DAN-MONICA III). The response rates were 79%, 75%, and 73%, respectively, and after exclusion of immigrants, mainly persons from Turkey, former Yugoslavia and Pakistan (less than 5%), data for 3317 women (1822 + 737 + 778) and 3378 men (1876 + 725 + 777) were available. Immigrants were excluded because they may tend to cluster on both biological variables and behavioural traits.

DATA COLLECTION

Data on well established biological and behavioural cardiovascular risk factors were collected by clinical examination and questionnaire, respectively. Comparable survey methods, which followed the WHO-MONICA protocol were used in all three surveys, and they have been described in earlier reports.13 Data on the biological variables, smoking, drinking and physical activity were missing from less than 1% of the participants, whereas 9.5% of the participants had not completed the questionnaire on dietary habits.

CARDIOVASCULAR RISK FACTORS

All anthropometric and blood pressure measures were collected by the same trained nurse in all studies. BMI was calculated as weight in kilograms divided by the square of height in metres. Blood pressure was measured with a London School of Hygiene sphygmomanometer, using one of three cuffs of different sizes. Measurements were done twice on the left arm with the participants having rested at least five minutes in supine position. All measurements in the three surveys were done by the same trained nurse who was subject to control of hearing. Preliminary data analyses showed that mean systolic blood pressure apparently decreased from 1982–84 to 1986–87, and then rose again in 1991–92. This peculiarity led to identification of a subtle difference in the examination programmes schedules (duration of rest before blood pressure measurements) as a likely source of variation, and data were re-analysed with exclusion of blood pressure data from the second survey. However, this did not change the results significantly. Blood samples for serum lipid analyses were taken after an overnight fast. Total cholesterol, and HDL-cholesterol were measured using the CHOD-PAP enzymatic method (Monotest R Cholestrol, Boehringer Mannheimer, Germany). The biochemical analyses were subject to internal laboratory quality control following the recommended protocols for the WHO MONICA project, and took full part throughout the three survey periods in the external quality assessment programme offered collaborating WHO MONICA centres by the WHO Regional Lipid Reference Centre in Prague. Results from this programme revealed consistent biases in the measurements of total cholesterol in the laboratory, amounting to an average of +5.8% in 1982–84, to +3.4% in 1986–7, and to +1.1% in 1991–2. The corresponding figures for HDL cholesterol measurements were +3.9%,+3,0% and 2.7%. All data used in this paper were adjusted to comply with these biases.13

Questions on smoking concerned current and previous habits (yes, occasionally, no/never), and the kinds and average daily quantities of tobacco consumed (cigarettes, cheroots, cigars, pipe). In the analyses of the data, occasional smokers were considered as non-smokers. A daily consumption of more than 15 grams of tobacco was defined as heavy smoking. Tobacco consumption was calculated by equating a cigarette to 1 g of tobacco, a cheroot to 3 g tobacco, and a cigar to 5 g tobacco.

Dietary habits were assessed using a short food frequency questionnaire with 26 food items as previously described.14 The alternatives in the frequency scale were: never, once a month or less, twice a month, once a week, two to three times a week, once a day, two to three times a day, and four times or more daily. Based on indices of overall diet quality,15 and the current recommendations for a healthy diet, we calculated a crude index by giving one point for each of the following five characteristics of the diet: (1) not consuming butter, margarine or animal fats daily, (2) consuming either raw or boiled vegetables at least once daily, (3) consuming either wholemeal white bread or wholemeal rye bread at least once daily, and (4) consuming fruit at least once daily. As there were relatively few participants in the two extreme categories (that is, those with 0 or 4 points), they were combined with the neighbouring categories to give three categories: a low value of the index (0 or 1 point), an average value (2 points), and a high value (3 or 4 points). Such indices of overall diet quality have been related to the risk of all cause mortality diseases in several prospective population studies.15

Data on leisure time physical activity were based on answers to a question for which the participants had to mark one of four alternatives: A, mostly sedentary; B, walking, bicycling or otherwise active at corresponding level at least four hours per week; C, going jogging or demanding sports, or doing heavy activity during leisure for at least three hours per week; D, long distance running or competitive sports several times per week. We defined people in group A as being inactive.

We calculated a cardiovascular disease risk index with both biological and lifestyle related variables. It was comparable to an index used in the FINMONICA survey,16 and was based on seven different risk factors each contributing 0,1, or 2 points as shown in table 1. This risk score has been shown to be a good way to summarise the combined effects of different risk factors on future cardiovascular risk,16 and this was confirmed in a preliminary survival analysis in the present population, where the risk score significantly increased the risk of myocardial infarction (n=72) by 31% per unit increase after controlling for age and sex.

SOCIOECONOMIC STATUS

The commonest indicators of socioeconomic status in epidemiological studies are education, occupation, income, or indices of social class.12-17 Our hypothesis was that an educational difference in cardiovascular risk factors increased during the 1980s, and therefore this study used education as the main indicator of socioeconomic status. Educational attainment was assessed with questions about the highest grade or year of regular schooling and the highest degree earned.

STATISTICAL METHODS

All data analyses were performed separately for men and women. We used multivariate regression models to analyse the influence of education and time on risk factors. Linear regression models were used for continuous variables, and logistic regression models for categorial variables. All the continuous biological risk factors and the summary index were approximately normally distributed. Educational level were divided into four categories: 0–7, 8–9, 10–11, and over 12 years. Time was measured in days calculated as the number of years between 1 November 1982 and examination of the participant.

Firstly, risk factor trends were analysed for each educational group, in univariate models with time as the independent variable of interest. Age was associated with education and most risk factors (data not shown), and consequently age was included in all models. Secondly, to test whether the time trend was the same in the different educational groups, data were analysed in multivariate models with age, education, time, and a time education interaction term as covariates. As the three surveys were not conducted in fully comparable seasonal periods, seasonal variations were considered as potential sources of bias. Hence, for variables with potential seasonal variation (weight, BMI, blood pressure and serum lipids) all analyses were repeated with data on a subsample of participants (n=3852) examined in the six months represented in all three surveys (February to April and August to October). All the data reported are based on the total sample, and whenever the regression coefficient based on the season specific data differed by more than 15% from the coefficients based on all the data, we have reported the signifcance level from both analyses. Data analyses were performed using SPSS version 7.0.

Men were better educated than women, and educational level increased over time in both sexes (table 2). The analyses of time trends from 1982 to 1992 in biological risk factors in each educational group showed that HDL-cholesterol concentrations declined in women in all educational groups, and in the highest educated men. Increases in height, and diastolic blood pressure were significant in men and women with less than 10 years of schooling, while an increase in weight and BMI was significant for men with 10–11 years of schooling and in women with 8–9 years. Systolic blood pressure, and total cholesterol, did not change significantly in any educational group (table 3).

Smoking and heavy smoking decreased most in the highest educated groups, and the proportion of heavy smokers in fact increased in the lowest educated women (table 4). The changes in the highest educated groups were attributable mostly to an increased prevalence of never smoking from the first to the second survey, whereas the changes from the second to the third survey were attributable to an increased prevalence of ex-smokers (data not shown). The proportion of physically inactive men did not change, but this frequency decreased in women with more than 11 years of schooling. The proportion with the most frequent intake of wholemeal bread, fruit, and vegetables increased in all educational groups in women, and in the least educated men. However, changes in the women were seen mostly from the first to the second survey. The summary index based on the seven cardiovascular risk factors improved with time, but this development was only significant in the highest educated men and women. This result did not retain significant when smoking was excluded from the index. In the regression models across educational groups, with the interaction between time and education testing the hypothesis that the trend was the same in the different educational groups, the interaction term was not significant in any of the models for each of the nine biological risk factors as the dependent variables (table 5). In the regression models across educational groups, test for interaction between time and education showed significance in models with smoking, heavy smoking, and the summary index.