Our study of postmenopausal women is population based with a high proportion of smoking participants. Our main finding was that current active smoking is related to early menopause, and that smoking cessation prior to menopausal age seems to protect against early menopause.
The most important limitation of our study is that the women were 59–61 years when they participated in the survey and were asked about health and related habits in their current situation. Many of the factors examined at that age, including BMI, coffee and alcohol consumption and marital status, may not reflect the status of these factors in the years before and around menopause. The reports on coffee and alcohol consumption or current smoking at the time of the survey were, however, probably representative for the status around menopausal age, and any misclassification of these variables would most probably be non-differential. In the Tromsø Heart Study the relative validity of the question on coffee consumption per day was compared with a dietary history survey two years later and the answers agreed well for coffee as well as other food items used every day in easily recorded unites [21]. Also, the reproducibility of the Tromsø survey questionnaire was studied by comparing it with a new questionnaire assessment one year later, and the concordance of coffee consumption was 67% for exact agreement and 99% for agreement within one category (Rho = 0.65). In the same study the concordance of alcohol consumption (beer, wine and spirit were recorded separately) was 66–76% for exact agreement and 96–100% for agreement within one category [22].
Previous studies have reported that high body weight is associated with later menopause [8, 9]. However, we found no association between BMI and menopause, possibly because of changes in BMI between the onset of menopause and the survey. It has been reported that the menopause increases the body weight [23] whereas other authors state that menopause is not affecting the BMI, merely the body composition and fat distribution [24]. Earlier smoking habits were reported and if misclassification occurred, it was probably not related to menopausal age. Any effect of these limitations would be to reduce the strength of the associations.
A potential limitation is recalling the exact age at which menopause occurred. Studies have shown that the reported age of menopause may be inaccurate [25, 26], although other studies have concluded that the event is so important in a woman's life that she will remember it clearly [27, 28]. However, even if the reporting is inaccurate, there is no systematic age misclassification [25] and any recall problem should reduce the strength of the associations. The use of hormone-replacement therapy was low in Norway until 1990, and then rose gradually to 16% of women aged 45–69 years in 1994 [29]. If started before any menopausal symptoms, hormone use may conceal the cessation of ovarian function and may lead a woman to not correctly estimate her age at menopause. However, this possible source of bias is of little relevance for the subgroup of women with menopause at the age of 40–44 years, and the low rates of hormone treatment make it unlikely that the menopausal age was misclassified in this cohort.
The response rate was 57.3% and selection bias is expected with this rate. However, it was conducted a study of self-selection based on all invited to The Oslo Health Study linking the socio-demographic data from public registers in Statistics Norway and data from The Oslo Health Study [19]. The response rate was positively associated with educational attainment, total income, married status and a western country of birth, and negatively associated with receiving disability benefit. Self-selection according to socio-demographic variables had little impact on the prevalence estimates of self-rated health, smoking and BMI [19]. There is no evidence or indication of connection between survey participation and menopausal age, so that self-selection for participation had probably little or no impact on the associations we found. Moreover, it has been shown that less educated persons were less likely to participate in the Oslo Health Study than were higher educated, but further that this had probably little impact even on prevalence data for risk factors such as smoking [19]
There may also be a certain survival bias. It can be speculated that individuals who are genetically the most vulnerable for smoking-related disease will be overrepresented among those dying before the age of 60. However, this most likely concerns few individuals and should reduce the association.
We lacked information about whether the menopause was natural or surgical. We excluded women who reported reaching menopause before the age of 40 years because it is likely that their menopause resulted from surgery, but it is also likely that some women in the sample underwent surgery after the age of 40 years. However, the rate of hysterectomy and oophorectomy is low in Norway, by a factor three to four, compared with the United States [30]. A large cohort study in Norway in 1995–97 reported that 10% of menopausal women aged 50–59 years had undergone a simple hysterectomy [31], and we consider 10% as a maximum estimate of the prevalence of women that might have been misclassified into the early menopause group in our study.
We have adjusted for most of the variables thought to influence age at menopause. It is possible that adjusting for variables such as stress, menstrual cycle patterns and family history of age at menopause could be relevant, but we had no information on these factors.
The results concerning current active smoking, number of cigarettes smoked daily and total smoking exposure differed by whether we adjusted for education alone or for all variables associated with menopausal age. The latter model may be an over-adjustment as not all of these variables have been shown to really influence menopausal age. In light of the present evidence one can argue that the adjustment for education only gives the least chance for a false result.
Similar to Cooper et al. [13], we did not find that passive exposure to smoking was associated with early menopause. This contrasts with the study by Everson et al[18] and with Cooper's earlier findings [32]. We included never-smokers only in the analysis of passive smoking. Cooper et al. also excluded current smokers in the analyses of passive smoking in the later paper [13], whereas former smokers were included in the earlier paper [32]. This may partially explain the different results. Everson et al. included only never-smokers, but still found an association between passive smoking and early menopause.
We found no association between early menopause and alcohol or coffee consumption. However, the crude OR showed that coffee consumption is associated with an early onset of menopause, which is consistent with Nagata et al.'s findings [9]. Nagata et al. adjusted for a number of factors, but not educational level or smoking, which we found to be important confounders of coffee consumption. When we adjusted for these factors, coffee was no longer associated with the onset of menopause, which is consistent with the findings of other studies [17, 33].
Other studies have reported that a moderate intake of alcohol is associated with later menopause [10, 12, 33]. However, only one of these studies adjusted for both smoking and education. Our analysis showed that education is an important confounder of the association between alcohol consumption and early menopause. If adjusted only for smoking and coffee consumption, alcohol was associated with early menopause (adj. OR, 0.56; 95% CI, 0.32–0.98, data not shown), but this association was no longer significant when adjusted for educational level.
Comments
View archived comments (1)