Subjects
The HUNT Study is a large population-based health survey in Nord-Trøndelag County in Norway. Between 1984 and 1986, all inhabitants aged 20 years or older (85,100) were invited to participate in the first wave of the study (HUNT 1) and a total of 77,216 (90.7%) accepted the invitation, filled in a questionnaire and attended a clinical examination. The participants were given a second questionnaire at the clinical examination to complete at home and return in a pre-stamped envelope. All participants in the HUNT Study gave a written informed consent upon participation and the study was approved by the Regional Committee for Ethics in Medical Research.
For the purpose of the present study, a total of 21,046 participants were excluded at baseline; 4,413 who reported prevalent cardiovascular disease (i.e. angina, myocardial infarction, and/or stroke), 1,003 without sufficient information on diabetes (diagnosis, treatment or duration), 15,227 without information on leisure time physical exercise and 403 without information on potentially confounding factors (i.e. systolic blood pressure, body mass index). After these exclusions, 56,170 participants (27,321 men (mean age= 47.4 years old) and 28,852 women (mean age= 48.5 years old)) were available for follow up on cause of death.
Follow-up
Individual person time at risk of death was calculated from the date of participation in the HUNT 1 study (1984–86) until the date of death or until the end of follow-up 31st December 2008, whichever occurred first. The mandatory reporting of death to the Cause of Death Registry in Norway constitutes the basis for the coding of underlying cause of death. Deaths were classified according to the International Classification of Disease (ICD-9 and ICD-10). Cardiovascular disease was defined by ICD-9: 390–459 and ICD-10: I00-I99, and ischemic heart disease by ICD-9: 410–414 and ICD-10: 120–125.
Study variables
A detailed description of selection procedures, questionnaires, and measurements can be found at http://www.ntnu.edu/hunt and in a report by Holmen and colleagues [15]. Briefly, information was collected on a range of lifestyle and health related factors, including medical history, leisure time physical exercise, smoking status, alcohol consumption and education. Height was measured to the nearest centimetre and weight to the nearest half kilogram, and body mass index (BMI) was calculated as weight (kg) divided by the square value of height (m). Blood pressure was measured two times using a mercury manometer and the mean of the two measures was used.
Diabetes
Participants who answered ‘Yes’ to the question ‘Do you have or have you had diabetes?’ in the first questionnaire, were defined as having diabetes (N=1,105).
The second questionnaire included additional questions to those with known diabetes. People who responded ‘No’ to the questions ‘Do you take tablets for your diabetes?’ and ‘Do you take insulin injections?’ were categorized as people without medication, while people who responded ‘Yes’ to one of the questions were categorized as people with medication. We did not have sufficient power in the statistical analyses to assess the two questions separately. The question ‘When were you first diagnosed with diabetes?’, provided information on diabetes duration.
Leisure time physical exercise
Information on leisure-time physical exercise was obtained from the second questionnaire. The participants were asked ‘How often do you exercise (on the average)?, By exercise we mean going for walks, skiing, swimming and working out/sports’ with five mutually exclusive response options: 0, <1, 1, 2–3, ≥4 times per week. Those who reported exercising once a week or more were also asked ‘for how long do you exercise each time (average)?’, with four mutually exclusive response options: <15, 15–30, 31–60, >60 min. Based on the information of both frequency and duration of exercise we calculated ‘hours of leisure time physical exercise per week’. The frequency response option 2–3 times per week was counted as 2.5 times and ≥4 per week counted as five times, whereas the duration response options <15, 15–30, 31–60, >60 min were counted as 10, 25, 45 and 75 min, respectively. For the purpose of the statistical analysis, we categorized this variable into three groups. This resulted in 22,885 physically inactive participants, 24,140 participants performing 0.1-1.9 hours of leisure time physical exercise per week and 8,040 participants performing ≥2.0 hours of leisure time physical exercise per week.
Statistical analyses
A Cox proportional hazard model was used to estimate adjusted hazard ratios (HRs) of death from cardiovascular disease and ischemic heart disease associated with diabetes treatment (with or without oral hypoglycemic drugs or insulin) and duration, and in a separate analysis, to assess the combined association of leisure time physical exercise and diabetes treatment with risk of death. Precision of the estimated hazard ratios was assessed by a 95% confidence interval (CI). All estimated associations were adjusted for possible confounding by attained age (as the time scale) and birth cohort (five years strata). In multivariable models we adjusted for smoking status (never, former, current, unknown), alcohol consumption (0, 1–4, ≥ 5 times last 14 days, total abstainer, unknown), education (<10 years, 10–12 years, >13 years, unknown), body mass index (kg/m2) and systolic blood pressure (mmHg).
Analyses of the independent association of diabetes treatment were stratified by sex, whereas analyses of the combined association were adjusted for sex in a pooled sample to increase statistical power. The latter was justified by likelihood ratio tests of interaction with sex in the independent analyses of diabetes treatment (P = 0.11) and in the analyses combining diabetes treatment and leisure time physical exercise (P = 0.37).
Additional to the analyses combining leisure time physical exercise and diabetes treatment, we conducted analyses of physical activity stratified by diabetes treatment and tested for possible statistical interaction between the two variables in a likelihood ratio test. Moreover, a likelihood ratio test was used to assess whether the results were modified by age at baseline (< 70 years and ≥ 70 years).
We also conducted three sensitivity analyses; first, people with probable type 1 diabetes (i.e. who were diagnosed with diabetes before 40 years of age and reported insulin injections); second, we excluded the first five years of follow up to evaluate possible bias that could arise if people with the most severe diabetes were unable to be physically active; third, we excluded people who reported a moderate or high degree of movement disability.
Departure from the proportional hazards assumption was evaluated by Schoenfeld residuals and graphical procedures (log-log plots). All statistical tests were two-sided, and all analyses were conducted using Stata for Windows, version 11.2 (StataCorp LP, Texas 77845, USA).