Background: In 2010, the prevalence of tobacco use in France was 33% and reached 39% in the population aged 18–44. The purpose of this article is to describe the trends in tobacco-attributable mortality in France between 1980 and 2010. Methods: Using data from the national mortality statistics and relative risks of death, we estimated the tobacco-attributable fractions (AF) by sex and age using the method developed by Peto et al. and used recently by the World Health Organization with improved relative risk estimates. The tobacco-attributable mortality by age and sex is obtained by multiplying the AFs by the number of deaths. They are estimated in 5-year intervals from 1980 to 2010. Results: In 2010, a total of 78 000 deaths were attributable to tobacco use in France. The number of deaths attributable to tobacco use among men decreased from 66 000 deaths in 1985 to 59 000 deaths in 2010, and the tobacco-AF decreased from 23% in 1985 to 21% in 2010. The number of deaths attributable to tobacco use among women increased from 2700 in 1980 (1% of all deaths) to 19 000 in 2010 (7% of all deaths). In the population aged 35–69, one in three deaths among men and one in seven deaths among women are attributable to tobacco use. Conclusion: While tobacco-attributable mortality among men has been declining during the past three decades, it has increased dramatically among women. Thus, effective preventive measures are urgently needed to stem the tobacco epidemic.
In 2010, the prevalence of tobacco use in France was 33% and reached 39% in the population aged 18–44.1,2 Tobacco smoking is an important risk factor for cancer and certain cardiovascular and respiratory diseases. Several estimations of tobacco-attributable mortality in France have been published. Peto et al.3,4 estimated mortality from smoking in developed countries, including France, in 5-year intervals from 1950 to 2005 and in 2009, the estimation for France in 2009 was 66 000 deaths. The World Health Organization (WHO) estimated the tobacco-attributable mortality in 2004 using improved relative risk estimates and extrapolation to 2004 of the mortality statistics based on previous years.5 Hill has estimated that 73 000 deaths were attributable to tobacco on the basis of the observed 2004 mortality and the same relative risks as WHO.6
The aim of this article is to estimate the 30-year trend in tobacco-attributable mortality in France by sex and cause of death. We describe the data and the method used to estimate mortality attributable to tobacco in the first part of the article, and the results and discussion are presented in the latter part of the article.
The WHO validated list of diseases with risks increased by tobacco use is presented in Supplementary Table S1.5 The annual number of deaths in France by cause, sex and 10-year age-group were extracted from the national mortality statistics online database (http://www.cepidc.inserm.fr). The causes of deaths between 1980 and 1999 were coded using the International Classification of Diseases 9th revision (ICD-9), while the 10th revision (ICD-10) was used for deaths from 2000 forward. The codes used to classify the diseases associated with tobacco are presented in Supplementary Table S1.
Though the number has decreased over time, the cause of death on a number of death certificates (on 6% of death certificates in 2010) is coded as ill-defined or unspecified or a cancer of an ill-defined or unspecified site (see Supplementary Table S1 for the ICD-10 codes). To avoid underestimating the tobacco attributable mortality, these deaths were redistributed. Deaths from cancer of unspecified sites were distributed among the specified sites according to the proportions of each of the specified sites by year, sex and age group, while the deaths of ill-defined or unspecified causes were distributed among the specified causes according to the proportion of each of the specified causes. These computations are presented in detail for lung cancer in Supplementary Table S2.
The relative risk of each disease incurred by smokers as compared to never-smokers has been estimated by Thun et al.7 and Danaei et al.8 from the Cancer Prevention Study II (CPSII), the American Cancer Society second cohort study9 in which 1.2 million individuals interviewed in 1982 were followed through 2006. The risks were adjusted for age, race, education, marital status, current or most recent occupation, weekly consumption of vegetables and citrus fruit, vitamin (A, C and E) use, alcohol use, aspirin use, body mass index, exercise and dietary fat consumption. These relative risks are presented in Supplementary Table S1.
We used the method proposed by Peto and Lopez10 with improved relative risk estimates from Thun et al.7 and Danaei et al.8 as used by the WHO to estimate the tobacco-attributable mortality.5 The estimation of the tobacco-attributable mortality is done differently for lung cancer and for all the other diseases.
Lung cancer tobacco-attributable mortality
The lung cancer tobacco-attributable mortality is estimated directly as the difference between the observed lung cancer mortality and the expected lung cancer mortality in the absence of exposure to tobacco. This expected mortality in the absence of tobacco is calculated by multiplying the number of individuals in the French population by the risk of lung cancer death among non-smokers in the CPSII (Supplementary Table S3). This assumes that the age and sex-specific risks of lung cancer in non-smokers are similar in France and USA.
Tobacco-attributable mortality for diseases other than lung cancer
The tobacco-attributable fraction (AF) is the proportion of deaths which would be avoided in the absence of exposure. For a disease other than lung cancer associated with tobacco use, one can compute the tobacco-AF using Levin’s formula:11
with P the proportion of smokers and RR the relative risk of disease among smokers as compared to non-smokers.
Consistency of the estimations of RR and P
The estimation of the relative risk RR and of the proportion of smokers P must be consistent. If the relative risks have been estimated in a population with a different history of tobacco consumption in terms of dose, duration, age at entry etc. than the population where the AF is computed, the result will be biased. To solve this problem, Peto et al.10 suggested to use lung cancer rates observed in France as an indirect indicator of tobacco exposure. The principle is to estimate a hypothetical prevalence calibrated to explain the lung cancer mortality observed in France as a mixture of lung cancer risks of death observed among US smokers and US non-smokers. For example, the risk of lung cancer in France in males aged 40–44 is F = 294/2 152 000 and the risks in US smokers and US non-smokers are, respectively, S* = 23 per 100 000 and NS* = 3 per 100 000, therefore one needs a proportion of French males smoking like the US population equal to 53% to explain the risk observed in France. This pseudo prevalence is called by Peto et al. the ‘smoking impact ratio’. It is computed for each year, sex and age group as:
where F is the age- and sex-specific lung cancer mortality rate observed in France, S* is the age- and sex-specific lung cancer mortality rate among smokers in CPS II, and NS* is the age- and sex-specific lung cancer mortality rate among non-smokers in CPS II.
Estimation of the attributable mortality
Having defined the prevalence of exposure in the French population on the basis of the risks of lung cancer in USA, one can use the relative risks observed in USA for other diseases.
The AF for each disease is then estimated using Levin’s formula with the relative risk observed in the US population, and replacing P by the smoking impact ratio.
The tobacco-attributable mortality for one disease is the product of the observed mortality for this disease and of the AF, for each year, sex and agegroup.
The age groups used in the analysis were ten 5-year age groups from 35–39 to 80–84 and one group for ages 85 years and older. No deaths under age 35 have been attributed to smoking.
Table 1 presents the tobacco-attributable mortality in France in 2010. Overall, 78 000 deaths were attributable to tobacco, representing 14% of the total mortality. The tobacco AF is 21% in the male population with 59 000 deaths and 7% in the female population with 19 000 deaths. Tobacco is an important cause of premature mortality with AFs equal to 27% in the population aged 35–69 vs. 11% in the population aged 70 or older. In the population aged 35–69, 33% of all male deaths and 15% of all female deaths are attributable to tobacco.
↵a: Taking into account the cancer deaths of unspecified site and deaths of ill or unspecified causes.
↵b: Stomach, liver, pancreas, lung, cervix, kidney, bladder and leukemia.
Cancer is the main contributor to tobacco-attributable mortality, causing 47 000 of the 78 000 tobacco attributable deaths. The tobacco-attributable cancer deaths represent 44% of the total male cancer mortality and 58% of the cancer mortality among men aged 35–69 years. Thirteen percent of female cancer mortality was attributable to tobacco in 2010, and this fraction reaches 20% among women aged 35–69 years. The most common of the tobacco-associated cancers is lung cancer, causing 28 000 deaths (23 000 among males and 5000 among females), followed by head and neck cancers (mouth, pharynx, esophagus and larynx) with 7400 deaths (6500 among males and 900 among females).
Cardiovascular and respiratory diseases are the main tobacco-attributable causes of death other than cancer, causing 20 000 and 11 000 deaths, respectively.
Table 2 and figure 1 present the number of tobacco-attributable deaths and the proportion of tobacco-attributable deaths in France between 1980 and 2010. In the male population, the total number of tobacco-attributable deaths decreased from 66 000 deaths in 1985 to 59 000 deaths in 2010, and the tobacco-AF also decreased from 23% in 1985 to 21% in 2010. In the female population, there were few tobacco-attributable deaths in 1980 (2700 deaths), but the number increased dramatically to 19 000 deaths in 2010. Furthermore, the fraction attributable to tobacco rose from 1.0 to 7% in the 30 years between 1980 and 2010. In the population aged 35–69, the same trends are observed but the proportion of deaths attributable to smoking are larger, with one death in three deaths due to tobacco in the male population and one in seven deaths in the female population in 2010.
Tobacco-attributable mortality in France by sex and year every 5 years
Number of deaths (thousands)
Tobacco-attributable fraction %
Number of deaths (thousands)
Tobacco-attributable fraction %
Number of deaths (thousands)
Tobacco-attributable fraction %
In 2010, a total of 78 000 deaths in France were attributable to tobacco, and in the population aged 35–69 one in three deaths among men and one in seven deaths among women age 35–69 were due to tobacco. The tobacco-attributable mortality has been decreasing in the male population over the last several decades, whereas it has increased dramatically in the female population. Public health policies on the price of tobacco products, advertising restrictions and antismoking campaigns have reduced tobacco consumption among French males. However, females who were aged 20 during the 1970s started smoking heavily, and the younger generations that followed have also adopted this same behavior.2 Therefore, the risk among females is expected to increase until 2050 (the expected extinction of the first generation of women who smoked heavily), even if the present younger generations stop smoking rapidly.
Previous estimates of tobacco-attributable mortality in France for the years 1975, 1985, 1990 and 1995 have been published,3 and updated for 2000, 2005 and 2008.4 These estimations are lower than ours because the relative risks for causes other than lung cancer were derived from the second survey of the American Cancer Society without taking into account potential confounding factors. They were considered to be overestimations and were systematically divided by 2. Following the WHO,5 we have used the excess risk incurred by smokers estimated on the basis of the same survey but taking into account various potential confounding factors, demographic and behavioral characteristics.8,12
To solve the problem of the consistency between the estimation of the relative risks and the definition of exposure, we have used the smoking impact ratio instead of the proportion of smokers observed in French general population surveys. We would have had to use surveys conducted between 1965 and 1995 by different institutions using different methods, and this would add heterogeneity,13 more importantly the consistency between the risk and the prevalence estimations would remain a major problem.
We did not study tobacco-associated morbidity because no national estimations of the burden of cardiovascular or respiratory diseases are currently available. Finally, the recent introduction of e-cigarettes on the French market may affect the mortality trends, however, very little is currently known about the risks of vaping. Future studies should integrate this new type of nicotine consumption in the estimation of the risks associated with the use of tobacco and/or nicotine products.
This work was supported by Gustave Roussy and by a grant from the “Cancéropôle Ile de France” and the French Cancer Institute “Institut National du Cancer” (INCA) grant number 2012-1PL SHS-06-IGR-1.
Conflicts of interest: None declared.
Our work presents the 30-year trend in tobacco-attributable mortality in France, by sex and cause of death.
A total of 78 000 deaths in France were attributable to tobacco in 2010, and in the population aged 35–69, one in three deaths among men and one in seven deaths among women aged 35–69 were due to tobacco.
The mortality attributable to tobacco has been decreasing among men, whereas it has been dramatically increased among women. Such mortality trends in women underline the need for urgent public health action to stem the tobacco epidemic among females.
The authors thank Nita-Hanh Nguyen for her assistance.