Abstract

Introduction

One of the first of several studies 2-7 to use this method was conducted by Collishaw,3 who calculated the number of deaths attributable to smoking in Canada for 1979. Unfortunately, the estimates of the proportion of smokers and of the relative risk of smoking, necessary to calculate the PAR, were not age-specific.

A theoretically more valid estimate of the number of smoking-related deaths in Canada for 1979 was provided soon after by Collishaw and Myers.4 Not only did they control for age, but, realizing that ex-smokers were still at increased risk of mortality relative to never-smokers, they included a selected group of ex-smokers to be part of their exposed population.

Collishaw, Tostowaryk and Wigle 5 estimated the mortality attributable to smoking in Canada for 1983 and 1985, using two different sources of information on relative risk. The first source,8 while more recent, could not provide age-specific estimates. The second source 9 provided age-specific estimations of relative risk, but they were based on studies conducted during the late 1960s.

Mao, Gibbons and Wong 7 estimated the number of smoking-related deaths in Canada for 1989, calculating deaths separately for current and ex-smokers using a recent source of information for risk estimates. The present study uses similar methods to estimate smoking-attributable mortality in Canada, by province, for 1991. To our knowledge, this study is the first to present provincial results.

Assuming that smoking-related relative risk remains constant, it is possible to estimate the future burden of mortality due to smoking. Based on overall smoking prevalence rates for 1969, 1979 and 1989, Mao et al.7 estimated decreases of 10% for males and 2% for females from 1989 to 1999. Assuming these decreases in smoking prevalence would occur at the same rate throughout all age groups, they projected the number of smoking-attributable deaths for 1999.

This paper presents the results of a model-specific projection of smoking prevalence rates in the year 2000. Trends in age-specific prevalence rates are allowed to vary independently of each other. Projected smoking-attributable mortality, based in part on the above results, is also presented. By projecting mortality among subgroups, we identify those who are at the highest risk of contributing to the future burden of smoking-related deaths. Such information can then be used to implement public health policies more effectively.

Methods

Sex-specific estimates of relative risk for white current and ex-smokers were obtained from the American Cancer Society's Cancer Prevention Study (CPS-II) [personal communication, 1994]. Relative risk, mortality and population data were all collected by five-year age groups (25-29, 30-34, ... , 85+). We retrieved all-cause mortality rates from the Canadian Mortality Database of Statistics Canada (unpublished tabulations). Population estimates for 1991 were also acquired from Statistics Canada,10 along with projected population estimates for 2000, 11 based on medium migration and fertility rates.

To estimate all-cause mortality as well as the prevalence of current and never-smokers in the year 2000, we created a general linear model for each sex and province stratum. This model consisted of terms corresponding to year, age group and the effect of the interaction between year and age group. The interaction term allowed trends in age-specific prevalence rates to vary independently of each other. The projected prevalence of ex-smokers in 2000 was determined by adding two terms corresponding to the prevalence of current and never-smokers to the model. This approach, as opposed to the independent modelling of each prevalence measure, was used to better ensure that in each case the sum of the three projected prevalences for 2000 was close to one.

Projected estimates of mortality, smoking prevalence and population, as well as current estimates of relative risk, were then used to compute the number of deaths attributable to smoking in the year 2000 1,5 by age and sex, for each province and for Canada. Due to the non-linearity of the data, log transformations were performed on each of the smoking prevalence variables prior to modelling.

As a basis of comparison, we also determined the number of smoking-related deaths in 1991 for the various categories. Since these results were to be used as reference points and because age- and sex-specific mortality rates can vary greatly from year to year in the less populous provinces, predicted 1991 mortality values were used in these calculations. Predicted mortality rates for both 1991 and 2000 were based on mortality data from 1980 to 1992.

In order to allow comparisons among provinces, smoking-attributable death rates were calculated and age-standardized by the direct method to the 1991 Canadian population.

Results

We estimated that a total of 45,064 deaths in Canada were attributable to smoking in 1991 (Table 3). By the year 2000, the number of such deaths was projected to reach 46,910. This growth was the result of a 24% increase in smoking-related deaths among females, which offset a 4% decrease among males. Between 1991 and 2000, projected percentage changes in smoking-related deaths among men ranged from declines of about 11% in Manitoba and Saskatchewan to increases of 1-2% in Newfoundland and Prince Edward Island. Among females, increases in smoking-attributable mortality were projected in every province, ranging from 8% in Saskatchewan to 44% in Newfoundland.

While the Canadian rate of smoking-related deaths for males was projected to be twice that of the corresponding female rate in 2000, this represented a sizeable narrowing of the gap between the sexes when compared to estimated figures for 1991 (Table 4). Projected declines in male rates were highest in the more populous provinces (Ontario, Quebec, Alberta and British Columbia) and lowest in Prince Edward Island. With the exception of Newfoundland, where a 17% rise was projected, female rates of smoking-attributable deaths were also projected to decline in every province. The greatest decrease was projected in Quebec.

After calculating the age-specific rates of mortality attributable to smoking, projected to the year 2000, we found a clear west-east gradient in provincial rates among men 55 years and over (Table 5). The eastern provinces, particularly Prince Edward Island and Newfoundland, were projected to have higher rates, whereas British Columbia was projected to have the lowest rates. Among the younger men, high age-specific rates were projected for Quebec, Nova Scotia and Prince Edward Island, while low rates were projected for New Brunswick, Newfoundland and Alberta.

Age-specific rates of smoking-attributable mortality were higher in males than females, except in the 85+ age category. Between the ages of 45 and 84, male rates were projected to be two to three times those in females. Before the age of 45, smoking-related deaths among females were rare.

The provincial west-east gradient in projected age-specific rates of smoking-related deaths was not evident in females. We projected that women living in the prairie provinces of Manitoba, Saskatchewan and Alberta would experience the highest provincial rates in both the 45-49 and 50-54 age groups, whereas women in Newfoundland and Nova Scotia would experience the highest rates in the 55-59 and 60-64 age groups. While the projected female rates of smoking-related deaths for Ontario, Quebec and British Columbia were relatively low in comparison to the other provinces prior to the age of 65, they were dominant in the older age groups. Projected female rates for Saskatchewan were the lowest for all ages after 65, and projected female rates for Prince Edward Island were consistently low, especially in the 55-59 and 60-64 age categories.

Discussion

In using all-cause mortality data to calculate the mortality attributable to smoking, it was necessary to assume that the excess mortality in smokers versus non-smokers was entirely due to their smoking. It is conceivable that those who smoke have a greater tendency to engage in activities that are injurious to their health apart from their smoking. Were such a bias present, it would serve to overestimate the number of deaths attributable to smoking.

While we used all-cause mortality data in this study, other researchers 12-16 have taken a cause-specific approach, determining which causes of death are "causally" related to smoking. The contribution of smoking to each of these cause-specific death totals was calculated and summed to provide an overall estimate. This approach provides the reader with a disease-specific breakdown of smoking-related deaths that can be quite informative. However, this method may underestimate the total deaths attributable to smoking due to the necessity of using summary relative risk estimates and because weak associations between smoking and diseases might remain undetected.17

Our estimate of 45,064 deaths attributable to smoking in 1991 was considerably larger than estimates for 1989. 6,14 While our estimate likely reflected at least a small real increase in smoking-related deaths, we believe the majority of the disparity to be the result of methodological differences. In particular, we used relative risk estimates within five-year age groups up to and including age 85 and over. This was especially relevant after the age of 65, where relative risk estimates and mortality rates (for current smokers and for former smokers) were known to vary by five-year age groups. To emphasize the importance of using relative risk estimates by five-year age groups, we determined weighted estimates of relative risk for the five age groups for which smoking prevalence data were consistently available and recalculated the number of deaths related to smoking in 1991. The resulting estimate was approximately 5000 fewer, with virtually all of the difference accounted for in the 65+ age group.

We projected a 24% increase among females and a 4% reduction among males in the number of deaths attributable to smoking in Canada from 1991 to 2000. In comparison, Mao et al.7 projected a lower increase in females (18%) and a higher reduction in males (7%) from 1989 to 1999. The growth in smoking-related deaths for females was present despite the general decline in current smoking prevalence among females in Canada since the late 1970s. This can be attributed to a projected increase in current smoking prevalence among those aged 65 and over, believed to be a cohort effect of the growth in the popularity of smoking among younger women in the late 1940s and 1950s.18 The reasoning is that the effect of changes in smoking prevalence on the number of deaths attributable to smoking will be most pronounced among the elderly due to their higher levels of mortality.

With regard to the rates of smoking-attributable deaths, we projected a decrease for both males and females in Canada from 1991 to 2000. A corresponding reduction in the future burden of smoking mortality did not follow owing to an expanding and aging Canadian population. The extremely low rate of smoking-related deaths calculated for women under age 45 was likely the result of the low rate of cardiovascular disease mortality among women in this group.

Provincial rates of deaths attributable to smoking provide a useful description of regional disparities as they have existed and are likely to become. While it is difficult to ascertain which factors are important in creating these disparities, it is conceivable that regional differences in public health education and its ability to reach the appropriate target audiences have played a role and will continue to do so. It is crucial that smoking cessation messages are delivered to the right audience in the most effective way. The diversity of cultures that exist across this country means that these messages must be specifically designed for the local communities to which they are targeted.

Acknowledgements

References

2. Armstrong B, de Klerk N. A comparison of premature mortality due to cigarette smoking and road crashes in Australia. Community Health Stud 1981;5:243-9.

3. Collishaw NE. Deaths attributable to smoking-Canada, 1979. Chronic Dis Can 1982;3(1):1-5.

4. Collishaw NE, Myers G. Dollar estimates of the consequences of tobacco use in Canada, 1979. Can J Public Health 1984;75:192-9.

5. Collishaw NE, Tostowaryk W, Wigle DT. Mortality attributable to tobacco use in Canada. Can J Public Health 1988;79:166-9.

6. Wigle DT, Mao Y, Semenciw R, McCann C, Davies JW. Premature deaths in Canada: impact, trends and opportunities for prevention. Can J Public Health 1990;81:376-81.

7. Mao Y, Gibbons L, Wong T. The impact of the decreased prevalence of smoking in Canada. Can J Public Health 1992;83:413-6.

8. Johansen H, Semenciw R, Morrison H, et al. Important risk factors for deaths in adults: a 10-year follow-up of the Nutrition Canada Survey Cohort. Can Med Assoc J 1987;136:823-8.

9. Godley FH. Cigarette smoking, social factors, and mortality: new estimates from representative national samples [Ph D dissertation]. University Microfilms International, Ann Arbor (Michigan): University of Maryland, 1974.

10. Statistics Canada. Annual demographic statistics, 1993. Ottawa, 1994; Cat 91-213.

11. Statistics Canada (Demography Division, Population Projections Section). Population projections 1990-2011 based on recent changes in fertility levels and revised immigration targets. Ottawa, 1991.

12. Ouellet BL, Romeder J-M, Lance J-M. Premature mortality attributable to smoking and hazardous drinking in Canada. Am J Epidemiol 1979;109:451-63.

13. King DR, Smith AH, Salter DM. Mortality attributable to smoking in New Zealand. NZ Med J 1983;96:195-9.

14. Collishaw NE, Leahy K. Mortality attributable to tobacco use in Canada, 1989. Chronic Dis Can 1991;12(4):46-9.

15. Morin M, Kaiserman MJ, Leahy K. Regional mortality attributable to tobacco use in Canada, 1989. Chronic Dis Can 1992;13(4):64-7.

16. Cigarette smoking-attributable mortality and years of potential life lost-United States, 1990. MMWR 1993;42:645-9.

17. Burch PRJ. Period and cohort trends for mortality and cigarette consumption in England and Wales, 1946-1980, with emphasis on sex ratios. J Clin Epidemiol 1988;41:373-84.

18. Thomson ME. Statistics of smoking in Canada. Ottawa: Canadian Council on Smoking and Health, 1984.

Author References

Larry F Ellison, Yang Mao and Laurie Gibbons, Cancer Division, Bureau of Chronic Disease Epidemiology, Laboratory Centre for Disease Control, Health Canada, Tunney's Pasture, Postal locator: 0601E2, Ottawa, Ontario K1A 0L2

[Previous] [Table of Contents] [Next]