Cancers attributable to tobacco smoking in France in 2015

Abstract

Introduction

The evidence on the carcinogenicity of tobacco smoking has grown vastly since the reports of an association between cigarette smoking and carcinoma of the lung in 1950.^1–3 The International Agency for Research on Cancer (IARC) has included cancers of the lung, larynx, oral cavity, pharynx, nasal cavity and paranasal sinuses, oesophagus, stomach, pancreas, liver, kidney, ureter, bladder, uterine cervix, colon and rectum, acute myeloid leukaemia and the mucinous tumours of the ovary in the list of cancers caused by tobacco smoking.^4,5 Exposure to second-hand smoke has also been found to increase the risk of lung cancer in never smokers and has been classified as carcinogenic.⁴

France is one of the high-income European countries with the highest daily smoking prevalence: in 2014, 28% (32% in men and 24% in women) of the French population were daily tobacco smokers.^6,7 In order to reduce the smoking prevalence in France, the public health authority has launched a national programme of tobacco control (Programme national de réduction du tabagisme), aiming to reduce the number of daily smokers by 10% in 2019.⁸ It is therefore of great public health interest to provide the up-to-date estimates of the proportion and number of cancers attributable to active and passive smoking.

This article provides, for 2015, estimates of the number and proportion of cancer cases attributable to active smoking in France for all cancers identified by IARC to be caused by smoking. It also estimates the number and proportion of lung cancer cases among never smokers due to exposure to passive smoking in their home.

Methods

Estimation of population-attributable fractions

Active smoking

We estimated the population-attributable fractions (PAFs) and numbers of cancer cases attributable to cigarette smoking using an indirect method of Peto and Lopez (1992),⁹ modified by Parkin (2011) and used in similar projects for the UK and Australia.^10,11 This method relies on two main assumptions for high-income populations: (i) active tobacco smoking is the most important risk factor for lung cancer; (ii) lung cancer incidence among never smokers is fairly small and roughly constant across populations of a given sex and age. The number of lung cancers due to active smoking can be calculated as the difference between the observed number of lung cancers and the expected number if the population had the incidence rate of never smokers.

Data

The data required for the analysis include, from the same population, the sex and age-specific incidence rates of lung cancer among never smokers and relative risks (RRs) for individual cancers in smokers in comparison to never smokers.^12–14 We used the lung cancer incidence rates in never smokers reported in Parkin (2011), which are estimated from lung cancer death rates in the second Cancer Prevention Study (CPS II study) for the period 1982–02.^10,13 The CPS II is an American prospective cohort study, involving roughly 1.2 million participants, over 30 years old at the baseline survey in 1982 with a 20-year follow-up.^12,13

Similar to previous studies, we used for most cancer sites the RRs of death from cancer among current smokers as compared to never smokers from analyses based on the CPS II study. We chose to use the RRs of cancer death instead of incidence, because they are estimated from cohort studies with long duration of follow-up and sample-size large enough to produce accurate risk estimates even for less common cancers. However, for mucinous ovarian cancer, the RR estimate for cancer incidence from a meta-analysis by Jordan et al. (2006) was used, as no RR for mortality or incidence have been estimated for the CPS II cohort.¹⁵ And for colorectal cancer, we used the RR for cancer incidence estimated by Hannan et al. (2009) from the follow-up of the CPS II Nutrition Cohort.¹⁶ Most other RRs are derived from the CPS II study re-analysis conducted by Ezzati et al. (2005) adjusted for important covariates (including age, race, marital status, education, employment, food and alcohol consumption).¹⁴ The RRs used are listed in table 1 with their sources.

The size of the French population in 2015 by sex and age group was obtained from the French National Institute for Statistics and Economic Studies datasets (INSEE).¹⁷ The number of cancer cases in France in 2015 by age, sex and site was estimated by applying the observed incidence rates for 2013 reported by the national network of cancer registries (FRANCIM) to the 2015 population.¹⁸ To estimate the number of mucinous ovarian cancers, we applied the age-specific proportions of mucinous ovarian cancers reported by the FRANCIM for 2009–13 (shown in Supplementary table S1) to the age-specific numbers of ovarian cancer cases estimated for 2015.

Analysis

Having calibrated the prevalence of smoking on the risk in the CPS II population, one can use Equation (2) to estimate the corresponding PAFs. The numbers of cancer cases attributable to smoking for each site are the products of the PAFs and the total incidence for each site.

Domestic passive smoking

Because never smokers exposed to second-hand smoke are at increased risk of lung cancer, we estimate the PAF and the number of lung cancer cases attributable to passive smoking. Due to lack of exposure data for passive smoking in workplaces and other public areas, we restrict the analysis to domestic passive smoking.

Data

The RR estimates of passive smoking on lung cancer are most robust for never smokers who have lived with a partner who smokes. Because data on exposure to passive smoking at work and other public places in France are not readily available, we restrict the analysis to never smokers who were living in the same household as a smoking partner.

Relative risks

The RRs for lung cancer incidence in never smokers who were exposed to tobacco smoke from a smoking partner were obtained from the meta-analysis published in the IARC Monograph volume 83 (2004).⁴ The RRs for males and females never smokers were 1.37 (95% CI 1.02–1.82) and 1.24 (95% CI 1.14–1.34), respectively.

Exposure

Because there is no data available in France on the proportion of never smokers living with a smoker, it was estimated following the methods described by Parkin (2011).¹⁰ Briefly, we used the marital status data in 2005 published by INSEE and the tobacco smoking prevalence (categorized in never, former and current smoker) from a 2005 national health survey (Baromètre Santé, 2005),²⁰ allowing a 10-year latency time for the occurrence of lung cancer. Both data are available by sex and five-year age groups in 2005. Recognizing that couples tend to be concordant for smoking status and as proposed by Wald et al. (1986)²¹ and used by Parkin (2011),¹⁰ we assumed that ever smokers and never smokers were three times more likely to live with a partner with the same smoking status than with one having different smoking status. We also assumed that couples were in the same five-year age-group. The estimation is detailed in Supplementary table S3.

Analysis

The number of lung cancer cases is the product of the estimated PAF and the total number of lung cancer cases estimated among never smokers. The latter is obtained by multiplying the proportion of never smokers by the number of lung cancer cases not attributed to active smoking.

Sensitivity analysis for active smoking

We additionally performed sensitivity analysis to test different assumptions for active smoking, by applying alternative smoking prevalence and alternative RR estimates, and by including female breast cancer for which there is limited evidence of tobacco carcinogenicity. Detailed descriptions of methods and corresponding results are provided in the Supplementary methods and Supplementary tables S5 and S6.

Results

Active smoking

Overall, 66 150 cancer cases (54 142 in men and 12 008 in women) were attributable to active smoking, accounting for 28 and 8% of all cancer cases diagnosed in 2015 among French men and women over 30 years old (table 2).

In 2015, 32 686 (25 494 in men and 7192 in women) of the 40 450 lung cancer cases (29 097 in men and 11 353 in women) diagnosed in the French population aged 30 years and above were attributable to active smoking, corresponding to PAFs of 88% for men and 63% for women (table 3). Table 2 shows, in men and women respectively, the estimated numbers and proportions of cancer cases attributable to active smoking, by cancer sites. Detailed estimates by age are provided in Supplementary tables S2a and b. Besides lung, cancer sites that have the highest PAFs include larynx (84% in men and 68% in women), oral cavity and pharynx (80% in men and 42% in women) and oesophagus (69% in men and 45% in women). Cancer sites (besides lung) that account for the largest number of cases due to smoking include oral cavity and pharynx (8177 in men and 1281 in women), bladder (4046 in men and 287 in women) and oesophagus (3101 in men and 497 in women).

Domestic passive smoking

Table 4 shows, by age and sex, the proportion and number of lung cancer cases attributable to domestic passive smoking in 2015. Detailed estimates, including the estimated prevalence of never smokers living with an ever smoker in 2005 and the number of lung cancer cases among never smokers in 2015, are provided in Supplementary tables S4a and b. In sum, 36 and 142 lung cancer cases among male and female never smokers were attributable to exposure of smoking from a partner, corresponding to 4.2 and 6.7% of lung cancer occurring in male and female never smoker respectively.

Discussion

Overall, 66 150 (or 19%) of all new cancer cases among adults in France in 2015 can be attributable to active smoking, representing 28 and 8% of all cancer cases diagnosed in men and women over 30 years of age. Almost half of those cases (32 686) were lung cancer. Among never smokers, past exposure to domestic passive smoking caused 178 cases of lung cancer.

Our results are consistent with previous studies. In 2000, 27% of the cancer cases in men and 6% in women in France were found to be attributable to active smoking²² vs. 28 and 8% in 2015 in this study. While prevalence of current smoking in men has declined progressively from about 60% in early 1970s to around 40% in 2000 and remained relatively stable thereafter, it has increased progressively in women from about 25% to over 30% during the same period.²³ In 2014, the prevalence estimates in men and women were 38 and 30%, respectively.⁷ This overall male-downward and female-upward trends in smoking prevalence have been reflected in smoking-attributable deaths. Deaths due to smoking in French men decreased from 66 000 in 1985 to 59 000 in 2010, corresponding to PAFs of 23 and 21%, respectively. In contrast, deaths in French women increased from 4700 in 1985 to 19 000 in 2010, corresponding to PAFs of 1 and 7%.²⁴ Consequently, smoking-related cancer burden is expected to increase among females in the foreseeable future, given the ascending trend in smoking observed in French women. Also worrisome is the high and rising prevalence of adolescent smoking, e.g. the prevalence of 16-year olds daily smokers increased from 17% in 2007–08 to 23% in 2011.²³

Several limitations should be considered. First, we used the lung cancer incidence rates among never smokers in the US to estimate the PAFs for lung cancer in France, assuming the factors that cause lung cancer among never smokers and the exposure to these factors among never smokers to be identical in both populations. Second, the sensitivity analyses using RRs from CPS II and smoking prevalence in France (with different latency times) are not ideal, as it assumes that the smokers in France and in the CPS II cohort have the same past smoking history. Results from sensitivity analysis (see Supplementary material) using RRs from the European Prospective Investigation into Cancer and Nutrition (EPIC) study suggest that this may be a concern for the PAF estimates for some cancer sites (e.g. oesophagus) where the RRs from the CPS II and the EPIC study differ markedly. Third, for most cancer sites, we used the RRs of deaths instead of incidence. This is based on (i) the assumption that the death rates are the same among cancer patients whether their cancer is tobacco related or not and (ii) the clear advantage of using RR estimates from a study with large sample-size and long duration of follow-up i.e. 20 years in CPS II.

As for domestic passive smoking, the RRs were derived from never smokers currently living with smokers. The corresponding PAF is therefore likely to be underestimated since the estimation fails to account for: (i) never smokers who have previously lived with a smoking partner and separated only recently; (ii) never smokers currently living with a former smoker who quit smoking only recently. To reduce this bias, following Parkin (2011), we estimated the PAF using the proportion of never smokers who were living with someone who had ever smoked in 2005. However, we may still have underestimated the true burden of passive smoking. First, the best data source we could find for marital status does not include unmarried partners living together, a common demographic status in France. Second, the current analysis does not include exposure to smoking from cohabiting members other than their partners, as such data are lacking. All these limitations are potential sources of underestimation.

Additionally, the choice of 10-year latency for lung cancer in never smokers is arbitrarily based on data availability and the assumption that the prevalence of living with an ever-smoking partner has been constant over the past 2–3 decades. This could also lead to underestimations, as the level of passive exposure to indoor smoking was likely higher in the 1990s due to limited awareness of the harmful effects of passive smoking on health and tobacco control regulations in place. Finally, we were unable to assess the PAFs for exposure to passive smoking at workplaces and other public spaces due to lack of data.

In conclusion, smoking was responsible for a significant share of cancer cases in France in 2015. Over the past decades, legislation has been passed to reduce tobacco consumption and exposure to second-hand tobacco smoke (e.g. loi Veil, 1976; loi Evin, 1991; décret, 2006).²⁵ Comprehensive smoking-free policy was further strengthened in February 2007 for workplaces, shopping centres, transportation hubs, hospitals and schools.²⁶ Subsequently, the bans were extended to hospitality places, such as bars, restaurants, hotels and casinos in January 2008.²⁶ Despite France’s commendable initiatives to de-normalize tobacco use, including the recent adoption of plain packaging, tobacco taxation and advertizing restrictions, a sizeable proportion of the country’s population are currently tobacco smokers. France also remains one of the highest tobacco-consuming developed countries, falling behind the U.K., Canada, Australia and New Zealand which all have reduced smoking prevalence to below 20% (vs. over 30% in France).^27,28 Accordingly, the present and predicted smoking pattern in France and its impact on population health should be alarming to the pro-tobacco control stakeholders.

Due to the long latency time between smoking and the occurrence of cancer, and the high historical smoking prevalence in France, cancer incidence and mortality are not likely to decline in the foreseeable future, particularly for female cohorts that smoked most heavily and are still young. More tobacco control policies specifically targeting females should be devised. Given the recent rise in tobacco consumption in 2005–10 (has stabilized since) and the rising smoking prevalence among adolescents, the long-term smoking-related cancer burden may continue to increase.^7,29 Hence, more in-depth research must be done to better understand the reasons for the present failures to reduce tobacco consumption despite generally strong tobacco control policies.²⁸ Specifically, we need to investigate how increases in the retail price of tobacco products and their magnitude affect consumption, why its positive effects on consumption have often not been sustained over time in France, and whether periodic adjustment by changing the affordability will address this. As 17% of smokers in France used the less expensive roll-your-own cigarettes in 2010, monitoring the consumption and regulating the price of cheap cigarettes and other substitute products are also necessary.³⁰ Meanwhile, reinforcement and adoption of more effective regulations on smoking are needed to achieve sustained reduction in smoking and thus lower the cancer burden in France.

Supplementary data

Supplementary data are available at EURPUB online.

Acknowledgements

The authors would like to thank Drs. D. Max Parkin, Lesley Rushton and the Steering Committee of the project entitled ‘Définition des priorités pour la prévention du cancer en France métropolitaine: la fraction de cancers attribuables aux modes de vie et aux facteurs environnementaux’. The authors would also like to thank the French Cancer Registries Network (FRANCIM) for providing cancer incidence data for France.

Funding

This work was funded by the French National Cancer Institute (INCa, grant nr. 2015-002). The funding source had no role in the collection, analysis or interpretation of the data or in the decision to submit the manuscript for publication.

Conflicts of interest: None declared.

References