The HUNT Study is a large population-based health survey in Nord-Trøndelag County in Norway. Between 1995 and 1997, all inhabitants aged 20 years or older (94,194) were invited to participate in the second wave of the study (HUNT2) and a total of 64,961 (70%) accepted the invitation, completed questionnaires and attended a clinical examination. All participants in the HUNT Study gave 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 9,427 participants were excluded at baseline: 2,443 with prior AMI; 40 without information on diabetes status; 6,357 without information on leisure-time physical activity; 342 without information on BMI and 245 without information on potentially confounding factors (i.e. systolic blood pressure or total serum cholesterol). After these exclusions, 55,534 participants (26,229 men and 29,305 women) were available for statistical analyses.
From the date of participation in HUNT2 (1995–1997), the participants were followed-up until a diagnosed first AMI or until the end of follow-up (31 December 2008). AMIs were primarily identified through medical records from the two hospitals in Nord-Trøndelag County, but also through a linkage with the National Cause of Death Registry. The mandatory reporting of death to the Cause of Death Registry at Statistics Norway constitutes the basis for the coding of underlying cause of death. Deaths were classified according to the ICD-9 (www.icd9data.com/2007/Volume1/240-279/250-259/250/default.htm) and ICD-10 (www.who.int/classifications/icd/en/). AMI was defined by ICD-9 code 410 and ICD-10 code I21, when identified through the Cause of Death Registry. The diagnosis of myocardial infarction followed international recommendations (first definition of myocardial infarction  and second universal definition of myocardial infarction ). Central to the diagnosis is the rise and fall of cardiac biomarkers above the 99th percentile of detection. Troponin I and T was included in the two hospitals from 1998 and from 2001 creatine kinase (CK) and the CK-MB test have been excluded from the routine diagnostic kit. In addition to the myocardial cell death detected by the biomarkers, at least one of the following was necessary: ischaemic symptoms (mostly chest discomfort or dyspnoea) or changes in the ECG indicative of ischaemia (development of pathological Q-waves, dynamic ST-segment changes, T-wave inversions or left bundle branch block). The cardiac units at the two hospitals validated most of the infarctions, and only about 5% of the infarctions were validated at departments outside a cardiac unit (i.e. internal medicine or surgery).
A detailed description of selection procedures, questionnaires and measurements can be found at www.ntnu.edu/hunt and in a report by Holmen and colleagues . Briefly, information was collected on a range of lifestyle- and health-related factors, including medical history, physical activity, smoking status, alcohol consumption and educational attainment. At the clinical examination, standard anthropometric measures were obtained in standing shoeless participants (height to the nearest centimetre, weight to the nearest half kilogram and waist and hip circumference to the nearest centimetre). BMI was calculated as weight (kg) divided by the squared value of height (m). For the purpose of the statistical analysis individuals were defined as being normal weight (<25.0 kg/m2), overweight (25–29.9 kg/m2) or obese (≥30 kg/m2). In the normal-weight group, 2.5% had a BMI < 18.5 kg/m2 and were excluded from the analysis of BMI due to possible bias by pre-existing conditions. Blood pressure was measured three times using a Dinamap 845XT (Critikon, Tampa, USA), and the mean of the second and third measure was calculated. A non-fasting whole-blood sample was drawn from all participants at the screening site. Blood was separated by centrifugation and serum samples were transported in a cooler to the Central Laboratory at Levanger Hospital and analysed on a Hitachi 911 Auto-analyzer (Hitachi, Mito, Japan). Glucose was measured using an enzymatic hexokinase method and total cholesterol using an enzymatic colorimetric cholesterol esterase method. All exposures were measured once, at baseline, without any updated information throughout the follow-up period.
Diabetes status was defined by two methods. First, participants who answered ‘Yes’ to the question ‘Do you have or have you had diabetes?’ were defined as having diabetes (n = 1,200). Second, persons who answered ‘No’ to this question but who presented with a non-fasting glucose level of ≥11 mmol/l at the examination were classified as having newly diagnosed diabetes (n = 166). A similar procedure has also been used in previous studies [20–22]. Ideally this criterion should be accompanied by information on symptoms of diabetes (e.g. polyuria)  but this information was not available. Those confirming diabetes in the questionnaire were invited to a follow-up investigation. A total of 926 individuals (77% of those invited) took part in these investigations, where blood glucose, serum C-peptide and autoantibodies to glutamate decarboxylase (GADA) were measured in a fasting state. In addition, GADA was analysed in individuals who declared having diabetes but who did not attend the follow-up and from whom serum was available from a non-fasting state (n = 228). Among those attending the follow-up visit, those with a GADA level >8.0 or with a GADA level <8.0 plus a C-peptide level <150 pmol/l were classified as having type 1 diabetes. Among those not attending the follow-up, only the former diagnostic criteria (GADA level >8.0) was used to classify type 1 diabetes. This resulted in a total of 185 people being classified as having type 1 diabetes.
Leisure-time physical activity
Information on leisure-time physical activity was obtained from the first questionnaire. Participants were asked to report their usual number of hours per week of light and/or hard leisure-time physical activity during the past year, with four response options (0, <1, 1–2, and ≥3 h) for light activity and the same response options for hard activity. In the questionnaire, light activity was defined as ‘not sweating/being out of breath’, whereas hard activity was defined as ‘sweating/out of breath’. For the purpose of the statistical analysis, a new variable was constructed based on the number of hours of both light and hard activity undertaken during a week, providing information on the level of total leisure-time physical activity. The participants were classified into the following four categories of leisure-time physical activity: inactive (no light or hard activity); low (<3 h light activity and/or <1 h hard activity per week); medium (≥3 h light activity and/or <1 h hard activity per week) and high (any light activity and ≥1 h hard activity per week). To increase the statistical power, participants with a low and medium level of physical activity were collapsed into one group (‘low/medium’).
Cox proportional hazard model was used to estimate adjusted HRs of first AMI associated with diabetes and, in separate analyses, to assess the combined association of physical activity and diabetes, and BMI and diabetes, with the risk of AMI. We also combined the information on BMI and physical activity to examine risk of AMI among people with diabetes. The precision of the estimated HRs was assessed by a 95% CI. All estimated associations were adjusted for possible confounding by attained age (as the time scale) and birth cohort (5 years strata). In multivariable models we adjusted for smoking status (never, former, current, unknown), alcohol consumption (never, not the last 4 weeks, one to three units in the last 4 weeks, more than four units in the last 4 weeks, unknown), education (<10 years, 10–12 years, ≥13 years, unknown), BMI (kg/m2), systolic blood pressure (mmHg) and total serum cholesterol (mmol/l). Additionally, we controlled for leisure-time physical activity level (inactive, low, medium, high) when estimating the associations between diabetes and risk of AMI. The analyses of the combined associations of diabetes and physical activity were conducted sex-specific and in the pooled sample adjusting for sex in the regression model. We used a likelihood ratio test of a product term in the model to assess statistical interaction (i.e. departure from a multiplicative effect) between diabetes and sex and between diabetes and physical activity. Additionally, the association between leisure-time physical activity level and risk of AMI was assessed in analyses stratified by diabetes status. In this analysis, among people with diabetes, we adjusted for diabetes duration (<5 years, 5–10 years, 10–15 years, >15 years, unknown) and reporting of ulcers on feet that had taken more than 3 weeks to heal (yes, no, unknown) in addition to the factors mentioned above. Finally, in an additive model we estimated the relative excess risk due to interaction (RERI) between physical activity and diabetes, as well as between BMI and diabetes. We used a method described by Anderrson et al  to calculate 95% CIs around the RERI estimate. A RERI larger than zero may suggest biological interaction between two or more risk factors.
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, College Station, TX, USA).