The Malmö Diet and Cancer (MDC) cohort was used to examine snuff use and its effect on incidence of CVD. All subjects who were 45–73 years old, and who lived in the city of Malmö, Sweden were invited. The health examination was performed at the Malmö University Hospital from March 1991 to February 1996. The subjects were invited by letters and advertisement in newspapers. Of the eligible population of approximately 70 000 citizens, 28 449 subjects, i.e. 17 203 women and 11 246 men, respectively participated. The participation rate was approximately 40% [14, 15]. The MDC cohort and the characteristics of the non-participants have been described previously . The representativeness of the MDC study has been evaluated by a comparison with subjects, in corresponding age groups, who participated in 1994 in a mailed questionnaire health survey in the city of Malmö [15, 16]. In that survey, non attendance was associated with a lower prevalence of smoking and a higher rate of mortality. However, prevalence of smoking and obesity in the MDC was rather similar compared to the mailed health study (prevalence of smoking 39.9% vs.37.5% and 12.4% vs. 11.8% for prevalence of obesity, respectively).
After exclusion of subjects with a history of MI or stroke, and subjects with missing information about BMI, blood pressure, diabetes or tobacco habits, the cohort consisted of 27 227 individuals, 16 754 women and 10 473 men, respectively. Mean age was 57.4 +7.6 years in women and 59.1+7.0 years in men.
Measurements of height (in cm), weight (in kilograms) and blood pressure (in mmHg) were performed. The participants also completed a comprehensive self-administered questionnaire about life style habits and medical history and treatment .
Smoking habits and use of snuff
Smoking was assessed in a self-administered questionnaire, and the subjects were categorized as never smokers, ex-smokers and current smokers (regular or occasional smokers). Daily cigarette consumption among current smokers was assessed by the questions "how many cigarettes do you smoke per day"?
Snuff use was assessed by the question "Do you use snuff?" In subjects confirming the use of snuff the weekly consumption of snuff was obtained by the question "How many packages of snuff do you consume every week?" We dichotomized weekly snuff consumption into three groups: a) 1–2 packages per week as low consumption, b) 3–5 packages per week as medium consumption and c) more than six packages per week as high consumption.
Chewing tobacco and use of nicotine gum
The habits of tobacco chewing and using nicotine gum were obtained by the questions: ""Do you chew tobacco" and "Do you use nicotine chewing gum?"
Blood pressure was measured twice in the supine position after 10 minutes rest using a mercury sphygmomanometer. Subjects who had systolic blood pressure (SBP) _140 Mm/Hg or diastolic blood pressure (DBP) _90 mmHg or blood pressure-lowering drug treatment of hypertension were considered to have hypertension .
Participants who reported that they had diabetes or who used anti-diabetic medication was considered to have diabetes. As fasting blood glucose and lipids were available only on subjects belonging to the cardiovascular project (n = 5 500) , It was not possible to define diabetes on all study subjects according to present or previous recommendations for classification of diabetes mellitus .
Body mass index (BMI)
BMI was calculated as weight/height2 (kg/m2).
Physical activity during leisure time was revealed through 18 questions covering a range of activities in the four seasons. The number of minutes per week for each activity was multiplied by an intensity coefficient, and an overall leisure time physical activity scores was created. Scores were divided into four quartiles and subsequently categorized as a) low (quartile 1), b) moderate (quartile 2 to 3), and c) high (quartile 4) levels .
Information on socio-economic circumstances
Information on occupation and marital status was assessed by the questionnaire. The categorization into occupational groups based on the Swedish socio-economic index (SEI)  has been presented previously [22, 23]. This classification takes into consideration the educational level needed for the job, the level of responsibility in the work organization as well as the actual work task . In this study data were dichotomized into four groups only, due to few cases. High and medium levels non-manual employees was dichotomized into the same group as "High occupational level" (SEI groups 46–69)", and contained the following groups: high level (i.e., business executives, university-engineers and teachers) and medium level employees (i.e., reg. nurses, employees, computer operators and high school teachers).
Low-level non-manual employees (SEI groups 33–36, i.e. office assistants, sales staff and secretaries, etc) were categorized as "Medium occupation level".
Skilled manual workers, unskilled workers and unspecified occupational groups (i.e. early retired women, housewives, students and unemployed) (SEI groups 11–22) were categorized as "low level occupational". The group "others" contained self employed* and farmers i.e. (SEI groups 70–99) [22, 23].
The questionnaire had the following four categories: Married, unmarried, divorced and widowed. In accordance with previous studies we dichotomized the cohort into the status living alone or not .
Education level was assessed in the questionnaire as previously described and divided into low (9 years education or less) and high education level (at least secondary graduation). The dichotomized variable education was used in the analyses .
Incidence of stroke and MI
Each participant was followed from the baseline examination until the first incident MI or stroke event, death, emigration out of Sweden, or Dec 31, 2004, whichever came first.
Incident stroke was defined as ICD-9 codes 430, 431, 434 and 436 [24, 25]. MI was defined as nonfatal MI (ICD-9 code 410) or fatal ischemic heart disease (ICD-9 code 410–414). The National hospital discharge register, the Stroke register of Malmö  and the Swedish hospital discharge register were used for case retrieval.
The Cox proportional hazards model was used to assess the relationship between snuff use and incidence of MI and stroke, respectively. The Cox model was also used to adjust the relationships for other risk factors. One way analyses of variance and logistic regression was used to compare risk factor in users and non-users of snuff. A general linear model was used to adjust the mean values (i.e., blood pressure and BMI) for other risk factors