Smoking-related general and cause-specific mortality in Estonia

Tobacco smoking is a single major cause of premature death worldwide [1, 2]. Despite substantial reduction in the use of tobacco, smoking still causes more deaths globally than diseases like tuberculosis, HIV and malaria combined, making it the largest preventable health risk factor [3, 4]. The use of tobacco is legally allowed and therefore the prevalence of smoking behaviour is still very high. While in the developed countries tobacco use is reduced and restricted, in developing countries the tobacco epidemic is still in the growing phase [1, 5]. It has been estimated that tobacco smoking causes globally 6 million deaths in a year [6]. 80% of these deaths are premature and affect lower income individuals [5]. This all makes tobacco smoking the largest single avoidable cause for mortality. Reducing the prevalence of smoking increases the health of the general population by avoiding premature disability and death. Smoking cessation and support to stop smoking are the easiest tools to improve the quality of life in population [7, 8].

While it is clear that smoking increases mortality, the association between smoking and mortality is different across specific causes of death. Cancers, chronic obstructive diseases of respiratory system and cardiovascular diseases are the most commonly referred smoking-induced causes of death.

The association between smoking and lung cancer has been described already in 1950 and following years [9‐12]. Later studies found clear association with other cancers as well [13]. Tobacco smoking is known to cause cancers of upper gastrointestinal tract, pancreas, prostate [14]. While the association is conclusive for lung and upper respiratory tract cancers, the connection between smoking and other cancers does not have conclusive evidence.

Tobacco smoking is reported to cause 90% of all chronic obstructive pulmonary disease (COPD) mortality and smoking cessation is the best measure to avoid this outcome [15]. The relative risk for smokers to die from COPD is 25.6 in men and 22.3 in women, compared to non-smokers [16]. Therefore, one of the most common health problems in smokers is COPD [16, 17]. The development of COPD is dependent on the chronic persistent inflammation in the lung tissue that induces tissue-remodelling [18]. COPD is the most common and the most prevalent disease in smokers and smokers have very highly increased risk for COPD.

Cardiovascular diseases have been found to be the third major cause of death caused by smoking [16, 19]. The main causes for death are ischaemic heart disease, cerebrovascular disease and other heart diseases [16]. Interestingly the risks caused by smoking of death due coronary heart disease are remarkably different between men and women [20]. Women who smoke have four times increased risk of death from ischaemic heart disease, while male smokers only have twofold increased risk [20]. For cerebrovascular diseases, this sex difference in risk has not been found [21]. This influence of sex on the risk of ischaemic heart disease is striking and its mechanism could be biological or behavioural.

Increased prevalence of chronic diseases suggests increased inflammatory activity in smokers. Indeed, several studies indicate chronic inflammation in the smokers and different immunomodulating toxins in the smoke probably cause increased immune reactivity [22, 23]. A recent study analysed RNA expression profiles in the blood of smokers and non-smokers and identified a highly significant increase in the expression of GPR15 gene [24]. GPR15 is an orphan receptor that is responsible for the homing of tissue-specific immune cells [25].

Several previous studies have shown that the association between smoking with different causes of deaths is variable [26]. There are some large-scale population studies where causes of death and association with smoking have been addressed [7, 13, 16]. However, more detailed differential risk analyses for smoking and causes of death are not that common [27]. The goal of present study was to analyse the smoking-caused mortality in the Estonian Genome Center cohort and to describe the differential sensitivity for the different specific causes of death.

Based on the epidemiological analysis of tobacco smoking in the cohort of the Estonian Genome Center, this study concluded that Estonia still belongs to the group of countries with high number of smokers. On average, there are two times more smokers in Estonia than in Western-European countries. The high number of smokers also is characteristic for other Eastern-European countries. This statement is particularly true in men. In Estonia and other Eastern-European countries around 40% of men are smokers, whereas in Western-European countries only 10–20% of men are smokers. In the case of female smokers, Estonia is similar to Western-European countries, where 20% of women smoke [1]. This fits the typical epidemiologic pattern seen in Europe and developed countries. Differences between different countries are smaller in case of women and are prominent in the case of men. These differences also explain the significantly worse health of men in Eastern-European countries. Men smoke too much and they pay for that with their health and quality of life as seen by the reduced life span.

The smoking prevalence in the Estonian Genome Center cohort is similar to the prevalence described in other study based on general population in three Baltic states [30]. In addition, the age-group prevalence of the Estonian Genome Center cohort is similar to the general Estonian population [28]. Therefore, our results are generalizable for the entire Estonian population.

While the proportion of smokers in Estonia has been stable for 20 years, the age at which individuals start smoking tobacco has decreased remarkably. At present, people start to smoke at the median age of 16 and this is similar for men and women. This trend is worrisome and has an important impact on the health of the Estonian population.

The age-group analysis indicated that men start to quit smoking at the age 46–65, women start quitting 26–45. This indicates that women smoke for a shorter period of their life. Taken together, men start smoking earlier, they smoke more cigarettes per day (mostly 20 cigarettes per day) and they quit smoking later. At the same time, women smoke less (mostly 10 cigarettes per day) they start smoking later and give up smoking earlier. This difference could also be detectable as the difference in survival and causes of death. Therefore smoking has a significantly larger impact on men’s health than in women.

There present study analysed the impact of tobacco smoking on the hazard of all-cause and cause-specific mortality. Survival analysis clearly indicated shorter median survival in smokers compared to non-smokers. Survival analysis also indicated that smoking cessation is beneficial for the survival. This means that toxic effects of smoking are reversible. The positive effect of quitting on the survival has been described in many previous studies [16, 17, 19].

Interestingly, in women the general survival differences caused by tobacco smoking were less prominent. For example, men who smoked lived on average 11.4 years less than non-smokers, bud women who smoked only lost 5.8 years on average. In some previous studies, similar sex-differences in smoking effects have been found [31]. This sex-difference may be explained by the lower smoking activity in women. In our sample women smoked 9 cigarettes per day while men smoked 15 cigarettes per day on average. Women also start smoking later and quit earlier that makes their smoking duration shorter and this explains the reduced number of cigarettes smoked by women. Previous studies support this idea. If women smoke more they have mortality and morbidity profile similar to men and the sex differences disappear [32, 33].

The loss of years found from the survival analysis in the present study was similar to the findings from previous studies where loss of at least 10 years was found [17]. This indicates that the study design used and the results are comparable with other studies. For example, the study of Jha et al. analysed survival from the age 25 [17]. The present study included people from the age 18 and mortality data were retrieved from the National Death Registry. This approach has been used for many previous studies.

The impact of smoking duration on general mortality gave somewhat controversial results as in men we found increased mortality in both groups compared to non-smokers. In case of men, the difference in general mortality between smoking more than 20 years and less than 20 years was very small (HR 2.1 versus 2.3) and statistically not significant. In case of women, the difference related to smoking duration was significant. Women smoking more than 20 years had HR 2.7 (CI 2.4–3.1) and women smoking less than 20 years have HR 1.4 (CI 1.3–1.6). This finding is similar to the previously published findings [34]. Similarly, we found that number of cigarettes per day does not have impact on the all-cause mortality.

The cause-specific analysis of death should describe the most common reasons why smokers lose years of their life. The comparison of causes of death induced by smoking in men and women identified very high similarities between both sexes in the pattern of cause of death. Namely, the most significantly smoking-associated causes of death were diseases of the respiratory and cardiovascular systems both in men and women. This indicates that there is a common biological mechanism behind the smoking that increases morbidity and mortality. This mechanism is the same in men and women. One very good candidate for this common pathway is the GPR15 that was recently found to be upregulated in both male and female smokers [24].

In addition to the lung and heart diseases, the one of the largest hazard ratio was related to external causes of death. In men, this cause was death from accidental poisoning by an exposure to alcohol (X45). In women, these causes were alcoholic liver disease (K70) and intentional self-harm by hanging, strangulation, suffocation (X70). Alcoholic liver disease (K70) is closely related to a poisoning by an exposure to alcohol (X45) and should be taken as an external cause of disease. The significantly increased risks for external causes of death have been described in few previous studies and more detailed analysis is difficult to find [14]. The prevalence of alcohol poisoning and self-harm in former smokers as causes of death were the same in former smoker and non-smokers. This means that the increased incidence of suicide and alcohol intoxication in current smokers could reflect particular biologically determined behavioural endophentoype and personality characteristics of current smokers. Smokers could be more impulsive or more prone to risk taking. One side of this behaviour is increased alcohol consumption, binge drinking and suicide. Not many studies indicate causes for this kind of association, but most probably there is common genetic variation between smokers and suicide victims. One recent study found an association between the polymorphic variable number tandem repeat (VNTR) in the SLC6A4 gene and increased risk for tobacco use disorder [35]. The association between polymorphisms in the BDNF gene and nicotine dependence also has been described [36]. Moreover, as tobacco smoking behaviour is often explained in conjunction with mood disorders, the link between suicide and smoking is not that surprising. At least one deletion in the GSTM1 and GSTT1 genes has been described in anxious smokers while no association with mood disorders have been found [37].

One of the limitations of the study is the small number of several specific causes of death that could make our analysis imprecise. This small number is caused by the short time for observation. Currently it is impossible to get better estimations and this study can be a starting point for the repeated mortality analysis of the Estonian Genome Center cohort. Extended observation will eventually increase the number of specific causes and improve the precision of the analysis. In order to improve the regression analysis, we merged some cases (like J44, J45 and J47) that increased the power to estimate proportional hazard for smokers and ex-smokers. This helped to get more reliable HR estimations.

In conclusion, the prevalence of tobacco smoking is still high in Estonia and men smoke at least two times more than women. This high prevalence is reflected in substantial loss of years of life and shorter survival of men. Smoking cessation can reverse the increased mortality, but the time needed for this reversal is not known. The two main types of cause of death in smokers were diseases of lung and cardiovascular system. The third group of cause of death is self-destructive behaviour. This finding suggests a common biological substrate in the brain of nicotine addicts and risk-taking behaviour. Better knowledge about the link between psychiatric illnesses and smoking could help to develop more effective interventions. Further studies are required to understand the genetic and biological background of this association.