Our study shows that compared with never smokers, both former and current smokers at cohort enrollment in the 1990s have an increased risk of dying overall, as well as dying from the three main outcome categories; total cancer, total CVD and total respiratory diseases. The increased risk was more pronounced for current than for former smokers indicating the effects of quitting smoking. One in three deaths in this female population may be attributed to smoking and could have been avoided if the women did not smoke.
As we do, recent studies [13–15, 22–27], including Caucasian women born in the 1940’s or later, found a more than doubling in the all-cause mortality rate, when comparing current with never smokers. Corresponding figures have for the first time emerged among Japanese women [28].
In our study, except for cancers not established to be smoking-related [5], current smoking was associated with an significantly increased risk of mortality from all categories of cause-specific mortality examined (lung cancer, other smoking-related cancers, circulatory diseases, MI, cerebrovascular diseases, COBP, and other respiratory diseases).
We find a three-fold mortality rate among current smokers compared with that of never smokers for cerebrovascular diseases and MI, similar to those displayed in the Nurses’ Health study [25]; and in the Million Women Study [13], while our mortality rate from COPD and from lung cancer was much lower than in the other two studies. The US study [25], had more than 12,000 deaths during a follow-up time of 24 years, while the UK study [13], had 66,000 deaths after 12 years. Both studies [25] had a RR of mortality from lung cancer of more than 20, while the mortality rate from COPD was 40 and 35, respectively. We expect that the deaths related to lung cancer and COPD, which have a long natural history, will increase when we have a longer follow-up time. Also, Kenfield et al. [26] showed in a later follow-up study that exposure assessed only at enrollment, underestimated the mortality risk due to smoking, compared with a second assessment during follow-up, especially for COPD and lung cancer.
The PAF of mortality due to smoking among Norwegian women has previously been estimated to be less than 20 % when mortality rates from lung cancer were used as an indirect measure of cigarette smoking [4, 12, 29], when cohort studies from the mid-70 s were used [23] and when the smoking prevalence surveys published by the Norwegian Central Bureau of Statistics were used [8, 30] Thun et al. [12], estimated that in 2009, the PAF of mortality due to smoking was highest (26–30 %) among women aged 35–69 years, in The Netherlands, Denmark, Hungary and Canada. In our study, based on individual data on both exposure and outcome, the corresponding PAF was estimated to be 34 % in Norway in 2008, with the upper limit of the 95 % CI as high as 39 %.
The life expectancy for Norwegian females was more than 83 years in 2008 [31]. The low background mortality rate for women, which is the denominator of our RRs, leads to greater PAF’s of mortality due to smoking. Our finding of a high proportion of deaths that could have been avoided if nobody was smoking is in accordance with the rapidly increasing lung cancer incidence for Norwegian women. The age adjusted world lung cancer incidence rate more than doubled from 11.7 per 100,000 person-years in the 5-year period (1986–1990) immediately before the commencement of our study to 25.1 per 100,000 person-years for the current 5 year period including the year which our follow-up ended [10]. This increasing lung cancer incidence also tells us that the PAF of mortality due to smoking among Norwegian women has not peaked yet.
As pointed out by Jha [3], the full effects of smoking can take 50 years to measure in individuals, and up to 100 years to measure in populations. Our study shows the PAF due to smoking among females already to be higher than the peak in the gender specific model recently developed by Thun et al. [12].
Strengths
The most important strength of our study is that it is a nationally representative prospective cohort study allowing us to calculate the PAF of mortality due to smoking for middle-aged women. We know from our previous studies that the smoking exposure [17, 32, 33] and the cancer incidence [19] reflect known smoking patterns [8, 9] and cancer incidence [10] for Norwegian women. Thus, we are confident that our cohort is representative of the Norwegian female population, born between 1926 and 1965, both according to exposure and outcome of this study.
Other major strengths of our study are; we have a high proportion of both current and former smokers, virtually complete follow-up through the National population based registries and the possibility of examining the association with smoking according to both all-cause and cause-specific mortality. Another force is that we focus our PAF estimates on the comparison between ever versus never smokers. Thus, it is only never smokers that could possibly change smoking status during follow-up. Since very few Norwegians start to smoke after the age of 30 and the mean age at enrolment for our study is more than 40 years, we are confident that the possible changes in smoking status among the never smokers during follow-up did not influence our PAF estimates. Furthermore, we have detailed information on, and were able to control for alcohol consumption and BMI, which are established risk factors for mortality.
Limitations
One major limitation of this report is that we have a limited number of deaths. Furthermore, 9 % of the women had missing values for physical activity levels and therefore this variable was not included in the multivariable analyses as there was no difference in physical activity levels according to smoking status. As in our study, neither of the seven [13–15, 25, 26, 34, 35] recently published studies did find much effect of confounding when they compared the effect of smoking on mortality. Five of these studies were conducted in the US [14, 15, 25, 26, 34]; one in the UK [13], and one in China [35]. We found a small non-significantly increased risk of deaths from cancers not established to be associated with smoking. This may be due to chance or that breast cancer should be in the category for smoking-related cancers. According to the most recent IARC monograph cigarette smoking is possibly carcinogenic to the human breast [5]. After this monograph was published two large cohort studies, one from the US [36], and our own Norwegian study [37] based on more than 300,000 women, conclude that smoking initiation before the first childbirth increase the risk of breast cancer.
Other limitations are that we have not validated self-reported information on height and body weight and that the current PAF formula does not provide a fully adjusted estimate of PAF due to lack of adjustment of the prevalence. We chose not to exclude women with prevalent disease as a previous study from the same cohort which examined physical activity and mortality found basically the same results when they did their analyses with and without women with prevalent disease at enrolment [38]. This has most likely deflated or RR estimates of mortality. Given the reduction in life expectancy associated with smoking we cannot rule out that part of the cause-specific associations is hidden or obscured by the competing causes of death with increasing age. We do not expect that the changes in BMI or alcohol consumption during follow-up would influence our results to any great extent. Nevertheless, we cannot rule out the possibility of residual confounding due to the above described factors, or other factors we did not measure.
In summary, one in three deaths among middle aged women in Norway could have been prevented if the women did not smoke. More middle-aged women, than ever before, are dying prematurely due to smoking in Norway.