The present study confirms that smoking increases the incidence of myocardial infarction and supports the notion that the relationships are stronger in women than in men [2–4]. Furthermore, this prospective cohort study found that passive smoking, at least in women, was an independent risk factor for myocardial infarction. The age-adjusted relationship in women between years living with a smoker after the age of 20 and myocardial infarction risk was as least as strong in ever smokers as in never smokers. In men, most of the association with passive smoking was probably explained by confounders, particularly their own active smoking.
Comparison with previous studies and explanations for the findings
Previous studies concerning the impact of passive smoking on myocardial infarction risk have shown that passive smoking or living with a smoker increases the risk of myocardial infarction [5–9]. Our results (at least before adjustments for active smoking were undertaken) are in accordance with this. Furthermore, we find that the effect of passive smoking was more pronounced in women. The latter seems to be in variance with most previous studies [6, 8, 9].
The findings of the large INTERHART case–control study indicated that the effect of passive smoking was weaker in current than in ex- or never smokers [7], but other studies have not found major differences [5, 8, 9].
We have no obvious explanation for the probably stronger relationship in women for both active and passive smoking with regard to the risk of myocardial infarction, and not much is known about how smoking may have different impacts in men and women [12]. It is possible that there is an interaction between some female specific risk factors (e.g., polycystic ovarian syndrome (PCOS), the use of oral contraceptives) and smoking. PCOS affects 5–10 % of women in a general population [13, 14] and is associated with several known risk factors for myocardial infarction (e.g., obesity, insulin resistance). The use of oral contraceptives in combination with smoking increases the risk of myocardial infarction [12], but this is of marginal relevance in this study as there were few cases among the females in fertile age. Some studies have shown an inverse relationship between age at menopause and myocardial infarction risk [15], and smoking lowers age at natural menopause by 1–2 years [16]. This could contribute to the higher myocardial infarction rate in female smokers. One likely explanation for the fact that the association of passive smoking with MI incidence was found predominately in women may be that men tend to smoke more than women. In current smokers, 28 and 12 % of men and women, respectively, smoked 20 cigarettes or more per day. Hence, a woman living with a male smoker might have a higher exposure to passive smoking than a man living with a female smoker.
Our data also support that the impact of smoking (both active and passive) on the MI incidence is stronger in younger than in older subjects. This was not unexpected as the relative risk associated with smoking is higher in relatively young individuals than in older subjects [2].
The INTERHEART study showed that in subjects who never had smoked cigarettes, passive smoking accounted for 10.8 and 18.6 % of all myocardial infarctions in women and men, respectively [7]. Another study indicated that passive smoking accounted for 9 % of the female cases and 7 % of the male cases [8]. The present study found that the PAR for passive smoking was 18 % in women, but it is difficult to compare PAR estimates from previous studies because of different threshold for being considered exposed for passive smoking as well as different prevalence of passive smoking.
Limitations and strengths
Population studies like this one may be prone to selection bias, as there will be the people who are healthy that are able to come to the clinical examination and participate in the study. We have previously reported lower mortality in attendees to the Tromsø 4 survey according to attendance to previous surveys they were invited to [10] and similar studies conducted in Norway have demonstrated higher morbidity and mortality in non-attenders than in subjects who took part in the surveys [17, 18]. The higher morbidity in non-attenders does, however, influence our findings only if the relationship between smoking and myocardial infarction is different in the large majority (77 %) who chose to attend the Tromsø 4 survey and in those who did not attend.
Data concerning smoking habits in the study were from a self-administered questionnaire which may have introduced an information bias. Some previous studies have used cotinine measurements as an objective measure of passive smoking [5, 9], only one of them demonstrated an association between serum cotinine level and myocardial infarction risk [5]. We have no objective measures of tobacco exposure like cotinine or thiocyanate. A previous Norwegian study showed that the relation between self-reported smoking habits and the measure of serum thiocyanate was strong if the question was asked in a neutral setting [19]. As the questions about active and passive smoking were included in a self-administered questionnaire, in a neutral setting, we believe that the validity was good, although we recognize that the tobacco use stated in this study probably is underreported.
In the present study we used two variables to measure passive smoking: How long they had been living with a smoker after their 20th birthday and how many hours per day they have spent in a smoke-filled room. The two variables were not highly correlated (Spearman correlation was approximately 0.3). Thus, the two variables probably reflect different types of exposure. We found that the relationships with myocardial infarction were stronger with years lived with a smoker than with number of hours per day in smoke-filled rooms. It is possible that the “smoke-filled room” question is somewhat ambiguous as it may include smoke filled rooms at workplaces (due to ambient air pollution) as well as passive smoking at home and at the workplace. This ambiguity might constitute a possible information bias, and we believe that it is correct to rather focus on the “living with smoker” variable which clearly reflects exposure to cigarette smoking.
Another possible limitation is the fact that the survey was administered in 1994, and we have no information regarding changes in smoking habits, both active and passive smoking, during follow-up. As the smoking prevalence in the Norwegian population has been reduced, we assume that some subjects classified as “current smokers” were ex-smokers by the time they had their infarction. Thus, our estimates of the risk of myocardial infarction associated with current smoking are probably underestimated because some in the “current smokers” group truly were “ex-smokers”, hence contributing a lower risk than a true “current smoker” to the results. Similar effects may have taken place with regard to exposure to passive smoking although the information concerning long duration of living with a smoker has probably not changed much during follow-up. It is a common finding in prospective studies that the relationship between the exposure and end-point becomes weaker with increasing duration of follow-up. This has also been seen in a study of the effect of passive smoking [5].
The study has, however, also some important strengths. This population based prospective study included 24,968 participants and the participation rate to the survey was high (77 %). The analyses are prospective, including incident cases, which rules out the effect of disease on the reporting of smoking habits. The analyses include all fatal and non-fatal cases in the study population, not only the non-fatal cases as in case–control studies [7, 8]. Out of the 411 men who died during follow-up, 194 (47 %) died within 2 days of the MI, i.e., most probably due to the MI. The corresponding figure in women was 50 % (144 of 289 deaths). Thus, nearly 30 % of the cases would have not been available for interview in a case–control study. Furthermore, we were able to adjust for baseline levels of the other traditional risk factors for myocardial infarction (total cholesterol, HDL cholesterol, blood pressure and physical activity), and we have also adjusted for active smoking when assessing the effect of passive smoking and vice versa.