Study population
The study population comprised of two subsamples of participants from the Inter99 study, which is a population-based randomised controlled trial, investigating the effect of non-pharmacological intervention on cardiovascular disease [22]. The study was performed at the Research Centre for Prevention and Health, Denmark, approved by the Copenhagen County Ethical Committee (KA 98155) and registered in the Clinical Trials.gov (NCT00289237).
The Inter99 study was initiated in 1999 with a subsequent five-year follow-up. The study population comprised all individuals born in 1939-40, 1944-45, 1949-50, 1554-55, 1959-60, 1964-65, 1969-70 and living in 11 municipalities in the southern part of Copenhagen County, and who were identified in the Civil Registration System. The initial entire study population consisted of 61,301 participants and was pre-randomised into three groups: a high intervention group (Group A, N = 11,708), a low intensity intervention group (Group B, N = 1,308) and a control group (N = 48,285). The inclusion criteria for the study population in the present article were: (1) randomised to group A; (2) information on dietary habits and physical activity at baseline and five-year follow-up; (3) high blood pressure (>140 mmHg) (N = 557) or high total cholesterol (>7 mmol/L) (N = 314) measured at baseline.
The intervention
The baseline examination took place between November 1999 and January 2001. The participants were invited for an initial health examination and their risk of ischemic heart disease was assessed at baseline. Using the Copenhagen Risk Score, the 10-year absolute risk of IHD was calculated (mean of sex, age, heredity, former IHD, diabetes, height, weight, smoking habits, cholesterol and blood pressure) by the computer program “PRECARD®” [23]. At baseline, 52% of the initially invited individuals in group A, corresponding to 6,091 men and women, participated in the baseline examination. At baseline, all participants at risk received a lifestyle consultation encouraging them to healthy behaviour (focusing on smoking, diet, physical activity and reduction in alcohol consumption). Participants in group A were offered group-based counselling in relation to smoking cessation and diet/physical activity, but participation was relatively low at baseline, and adherence was poor. All participants were re-invited after one-, three- and five years. At five-year follow-up, the health examination programme was repeated. Furthermore, all participants with high blood pressure, cholesterol or glucose received a print-out of the results for their general practitioner and were recommended to contact him/her.
Assessment of dependent variables
The Registry of Medical Product Statistics contains information on all outpatient prescriptions in Denmark using the Anatomical Therapeutic Chemical Classification System (ATC codes). By linking survey data with registry data using Danish unique personal identification numbers, we identified data on prescriptions related to lipid-lowering (code C10A and C10B) and antihypertensive medications (code C02, C03, C7, C8, C09). We used the 2007 version of the ATC/DDD classification. Initiation was defined as a least one redeemed prescription at a pharmacy in the period from baseline participation to the end of the 5-year follow-up. Participants who reported using the abovementioned medications at baseline were excluded from the analyses.
Health behaviour
Information on physical activity and dietary habits was obtained by questionnaire. Total physical activity was calculated on the basis of two questions, on commuting physical activity and leisure time physical activity [24]. The variable was calculated by summing responses to commuting physical activity (converted into minutes per week using a five day working week) and leisure time physical activity (converted into minutes per week). Physical activity was defined as: decreased/unchanged level and increased level. Change in dietary intake (nine-class variable) was measured by a dietary quality score (intake of fish, vegetables, fruit and fat) using a self-administrated 52-item food questionnaire (reference period: the last week). The questionnaire had been validated using a 28-day diet history and biomarkers as reference methods [25]. The nine-class variable was dichotomised into two groups: healthier and unchanged/more unhealthy. None of the participants in this study population received a maximum score of 9 (the healthiest) at baseline. Particularly, the groups that were more unhealthy or had decreased values at follow-up were relatively small, and therefore both health behaviour variables were dichotomised, due to small sample size.
Other covariates
Information on age and sex was taken from the Civil Registration System. From the Integrated Database for Labour Market Research at Statistics Denmark, information on educational attainment at baseline was obtained. Educational attainment was dichotomised into “high” (at least secondary education) and “low” (primary education only). Education is likely to be a good proxy for income level, as expenditures for medication require co-payments, while visits to the general practitioners are free of charge in Denmark. Information on contacts with general practitioners was obtained by linking to the National Health Insurance Service Registry. At the health examination, participants provided fasting blood samples for assessment of total cholesterol. Total cholesterol was measured by enzymatic procedures (Boeringer Mannhein, Germany). Blood pressure (BP) was measured twice with a mercury sphygmomanometer after 5 min of rest in a lying position. Height was measured without shoes to the nearest cm, weight was measured without shoes and overcoat to the nearest 0.1 kg and body mass index (BMI) was calculated. Risk of ischemic heart disease was assessed by using the “Copenhagen Risk Score” [23]. Smoking habits at baseline were assessed by answers to self-report questions at baseline.
Statistical analysis
Data were analysed using SAS statistical software, version 9.2 (SAS Institute., Cary, NC, USA).
The associations between initiation of preventive medications and changes in health behaviour were explored by logistic regression models. We excluded those with diabetes, as guidelines for patients with diabetes differ from those without diabetes. Regarding analyses for physical activity, those who reported low physical activity at baseline due to illness or handicap were excluded (N = 14). We examined three models of the relationship between change in health behaviour or initiation of lipid-lowering and antihypertensive medications. In model 1, we included age, sex and baseline values for physical activity (in analyses of change in physical activity) or dietary habits (in analyses of change in dietary habits). In model 2, we included covariates of health behaviour and educational level, and in model 3, we adjusted for risk score, BMI and in models regarding lipid-lowering medications, we adjusted for cholesterol at baseline, and in models regarding antihypertensives were adjusted for blood pressure level. As a model control, we carried out Hosmer and Lemeshow goodness-of-fit tests for the logistic regression models. Linearity of age, cholesterol and blood pressure were tested, and when needed, we squared the number, or raised it to the third power. We tested plausible interaction terms between the primary explanatory variable and sex and education.