Active smoking and secondhand smoke increase breast cancer risk: the report of the Canadian Expert Panel on Tobacco Smoke and Breast Cancer Risk (2009)
- Kenneth C Johnson1,
- Anthony B Miller2,
- Neil E Collishaw3,
- Julie R Palmer4,
- S Katharine Hammond5,
- Andrew G Salmon6,
- Kenneth P Cantor7,
- Mark D Miller8,9,
- Norman F Boyd10,
- John Millar11,
- Fernand Turcotte12
- 1Science Integration Division, Centre for Chronic Disease Prevention and Control, Public Health Agency of Canada, Ottawa, Ontario, Canada
- 2Dalla Lana School of Public Health, University of Toronto, Toronto, Ontario, Canada
- 3Physicians for a Smoke-Free Canada, Ottawa, Ontario, Canada
- 4Boston University School of Public Health, Slone Epidemiology Center at Boston University, Boston, Massachusetts, USA
- 5School of Public Health, University of California, Berkeley, California, USA
- 6Air Toxicology and Risk Assessment Division, Office of Environmental Health Hazard Assessment, California Environmental Protection Agency, Oakland, California, USA
- 7Division of Cancer Epidemiology and Genetics, US National Cancer Institute, Bethesda, Maryland, USA; currently of KP Cantor Environmental LLC, Silver Spring, Maryland, USA
- 8Office of Environmental Health Hazard Assessment, California Environmental Protection Agency, Oakland, California, USA
- 9Paediatric Environmental Health Specialty Unit, University of California, San Francisco, California, USA
- 10The Campbell Family Institute for Breast Cancer Research, Ontario Cancer Institute, University of Toronto, Toronto, Ontario, Canada
- 11Population Health, Provincial Health Services Authority, Vancouver, British Columbia, Canada
- 12Département de médecine sociale et préventive, Université Laval, Québec City, Québec, Canada
- Correspondence to Kenneth C Johnson, Science Integration Division, Centre for Chronic Disease Prevention and Control, Public Health Agency of Canada, 785 Carling Avenue, Ottawa K1A 0K9, Canada;
Contributors KJ drafted the current manuscript based on the full report. All authors were involved in the preparation of and editing of the main report and approved the final version of the main report on which the summary report is based. All authors reviewed and approved the summary manuscript.
- Received 27 January 2010
- Accepted 14 June 2010
- Published Online First 8 December 2010
Four authoritative reviews of active smoking and breast cancer have been published since 2000, but only one considered data after 2002 and conclusions varied. Three reviews of secondhand smoke (SHS) and breast cancer (2004–2006) each came to different conclusions. With 30 new studies since 2002, further review was deemed desirable. An Expert Panel was convened by four Canadian agencies, the Ontario Tobacco Research Unit, the Public Health Agency of Canada, Physicians for a Smoke-Free Canada and the Canadian Partnership Against Cancer to comprehensively examine the weight of evidence from epidemiological and toxicological studies and understanding of biological mechanisms regarding the relationship between tobacco smoke and breast cancer. This article summarises the panel's full report (http://www.otru.org/pdf/special/expert_panel_tobacco_breast_cancer.pdf). There are 20 known or suspected mammary carcinogens in tobacco smoke, and recognised biological mechanisms that explain how exposure to these carcinogens could lead to breast cancer. Results from the nine cohort studies reporting exposure metrics more detailed than ever/never and ex/current smoker show that early age of smoking commencement, higher pack-years and longer duration of smoking increase breast cancer risk 15% to 40%. Three meta-analyses report 35% to 50% increases in breast cancer risk for long-term smokers with N-acetyltransferase 2 gene (NAT2) slow acetylation genotypes. The active smoking evidence bolsters support for three meta-analyses that each reported about a 65% increase in premenopausal breast cancer risk among never smokers exposed to SHS. The Panel concluded that: 1) the association between active smoking and breast cancer is consistent with causality and 2) the association between SHS and breast cancer among younger, primarily premenopausal women who have never smoked is consistent with causality.
Several lines of evidence suggest that environmental factors can influence breast cancer risk: (1) several fold differences in risk between low-risk and high-risk regions internationally, (2) higher risks in the more industrialised nations and (3) changes in risks observed over time and in migrant studies.1 Concern about exposure of women to tobacco smoke stems from the knowledge that: (a) smoking is an established cause of at least 15 types of cancer,2 (b) smoking prevalence among women was over 20% in more than 25 countries at the turn of the 21st century,2 and (c) that exposure to secondhand tobacco smoke (SHS) was epidemic in many countries over much of the 20th century.2
Three authoritative reviews of active smoking and breast cancer considered data published to 20021–3 and one considered data until 2005,4 but their conclusions were not consistent. Similarly, the International Agency for Research on Cancer (IARC) in 2004,2 the California Office of Environmental Health Hazard Assessment in 2005,4 and the US Surgeon General in 2006,5 reported on the relationship between breast cancer and SHS. Each came to a different conclusion, the first concluding there was no relationship of SHS to breast cancer, the second that the evidence was consistent with causality for breast cancer in younger primarily premenopausal women and the third that the evidence was suggestive of a causal relationship.2 4 5
Since 2002, more than 30 more original epidemiological studies and at least 6 meta-analyses have been published on various aspects of tobacco smoke and breast cancer,6 and further review of the evidence was deemed desirable.
The mandate of the Expert Panel was to provide an up-to-date synthesis of current knowledge of breast cancer and exposure to tobacco smoke, focusing on the extensive new research in the area and examining active smoking and exposure to SHS and their association with premenopausal and postmenopausal breast cancer.
An Expert Panel was convened by four Canadian agencies, the Ontario Tobacco Research Unit, the Public Health Agency of Canada, Physicians for a Smoke-Free Canada and the Canadian Partnership Against Cancer. This article provides a summary of highlights of the Expert Panel's full 75-page report6 which is available free online from the Ontario Tobacco Research Unit (http://www.otru.org/pdf/special/expert_panel_tobacco_breast_cancer.pdf).
The work of the Expert Panel
Prior to the deliberations of the panel, the secretariat prepared a background document that summarised the breast cancer and tobacco literature, including detailed tables of risks reported for specific metrics of smoking from more than 100 studies. The document also highlighted key original research, meta-analyses and reviews from the year 2000 until November 2008. Studies were located from: existing reviews and reports;1–5 PubMed searches (using the terms smoking, cigarette, tobacco, tobacco smoke, secondhand smoke (SHS) and environmental tobacco smoke each combined with breast cancer, breast neoplasm, breast tissue, cancer, female cancer and breast and risk); and references in the located papers. The document and key papers were circulated to panel members for review in advance of the formal panel meeting that took place on 10–11 November 2008 in Toronto, Canada.
The Expert Panel based their assessment of breast cancer risk related to tobacco smoke exposure on the weight of evidence from epidemiological studies, toxicological studies and current understanding of biological mechanisms, using standard criteria for judging evidence as described by the IARC,2 the US Surgeon General3 and the California Environmental Protection Agency (CalEPA).4
Methodological issues considered in the panel's review included the number and quality of individual studies, the extent to which the design or analysis took into account potential confounders, selection bias, the potential for exposure misclassification, prospective or retrospective assessment of exposure and study power. The evidence from cohort studies was given more weight than that from case-control studies if other important aspects of the quality of the studies were similar but risks differed, because the cohort design avoids the possibility of recall bias.
Accurate ascertainment of tobacco smoke exposure was considered fundamental to the quality of case-control and cohort studies. The studies deemed of highest quality with respect to exposure were those providing timing, duration and intensity of smoking for active smokers. For SHS exposure assessment, studies that collected a quantified lifetime assessment of SHS exposure, including childhood, adult residential and adult occupational histories of SHS exposure and that used the never SHS exposed group as referent were preferred.
We report first on biological reasons to suspect that tobacco smoke may cause breast cancer. Then, we summarise the epidemiological evidence related to active smoking and breast cancer risk, focusing on earlier summarisations of the evidence and then on new evidence compiled by the panel in particular from recent cohort studies. Next, we examine research on active smoking and breast cancer risk in genetic subgroups, focusing on studies of women who were carriers of mutations in the breast cancer 1, early onset gene (BRCA1) and BRCA2, familial breast cancer and meta-analyses of risk associated with genetic subgroups. Finally, we examine the evidence on SHS and breast cancer risk among never smokers with a focus on recent meta-analyses.
There are persuasive biological reasons to suspect that exposure to the carcinogens in tobacco smoke may lead to breast cancer:
There are at least 20 known or suspected human carcinogens identified by IARC that are present in tobacco smoke that have been demonstrated to induce mammary tumours in rodents (see the first table in the full report). Many of these carcinogens are fat soluble and act by genotoxic mechanisms.7
These carcinogens can be activated into electrophilic intermediates by enzymes active in human breast epithelial cells.8
Several carcinogenic components of tobacco smoke are known to reach the breast and are secreted into the breast milk.9–12
Genes coding for activation/detoxification enzymes such as N-acetyltransferase 1 (NAT1), NAT2 and CYP1A1, have been reported to modify the relation of tobacco smoke to breast cancer risk.13–17
Electrophilic metabolites of tobacco compounds bind to DNA and form DNA adducts that can be detected in human breast cancer epithelial cells and in normal and cancerous breast tissue biopsies from women who are current or former smokers or who are passively exposed to tobacco smoke.18–21 (See reviews by the Office of Environmental Health Hazard Assessment,4 Miller et al22 and Hecht).23
Genomic alterations observed in vitro after exposure of human breast cancer epithelial cells to tobacco carcinogens resemble those seen in familial breast cancer.24
Epidemiology: active smoking and breast cancer
Palmer and Rosenberg27 identified 47 studies of breast cancer and smoking published from 1960 through 1992. Only 10 case-control and 5 cohort studies published between 1984 and 1992 met their quality criteria for inclusion. They concluded that there was ‘little evidence to suggest that cigarette smoking materially increases risk. Most studies have found no association or very small positive associations for ever smoking, current smoking, or heavy smoking’.27
In 2002, Terry and Rohan1 cited 67 additional studies of breast cancer and smoking published since the Palmer and Rosenberg review, and identified 4 areas where findings were suggestive of an increased breast cancer risk: (1) smoking of long duration, (2) smoking before the first full-term pregnancy, (3) SHS and (4) increased risk among women of certain genotypes who smoke.
Since that review, at least 30 more original studies, 6 meta-analyses and 3 major reports have been published. Appendix 1 in our full report6 presents summaries of seven epidemiological reviews and four meta-analyses that have addressed active smoking, five reviews and four meta-analyses of SHS, and three meta-analyses of genetics and active smoking.
Collaborative Group on Hormonal Factors in Breast Cancer: meta-analysis of smoking, alcohol and breast cancer, 2002
The Collaborative Group on Hormonal Factors in Breast Cancer meta-analysis of smoking, alcohol and breast cancer28 was highlighted in the evaluations by the Surgeon General3 5 and IARC2 of active and passive smoking and breast cancer. The Collaborative Group analysis involved individual data from 53 studies of breast cancer, estimated by the authors to be 80% of the world literature on the subject at the time. They concluded that: ‘the relationship between smoking and breast cancer was substantially confounded by the effect of alcohol. When analyses were restricted to 22 255 women with breast cancer and 40 832 controls who reported drinking no alcohol, smoking was not associated with breast cancer (compared to never smokers, RR for ever smokers 1.03; 95% CI 0.98 to 1.07 and for current smokers 0.99; 0.92 to 1.05).28
The panel considered comparisons of never, ever and current smokers as too crude to carefully assess lifetime breast cancer risks because women's smoking histories vary dramatically. Metrics examining the relative smoking dose (years of smoking and pack-years) allow for targeted assessment of risk among the women with the longest and the highest exposure to tobacco smoke. Age at smoking initiation, smoking before first pregnancy and years of smoking before first pregnancy allow targeted assessment of exposure at potentially critical times when changes in the breast are occurring either because of puberty or first pregnancy. An ever/never or ex/current metric can easily obscure these risks. Thus, if the Collaborative Group had restricted their analysis of alcohol to ever/never drinker they would have failed to find an increased risk for alcohol.28 Instead, they considered numbers of drinks per day and found a dose-response relationship (an increase in breast cancer risk of 7% for each drink per day) and concluded the relationship was causal.
Therefore, we compared the ever smoker versus never (active) smoker risk estimates in the Collaborative Group's report with risk estimates for longer duration smoking for the subset of those studies which had controlled for alcohol consumption. Three of the four cohort studies had borderline statistically significant or statistically significant risk increases for longer smoking duration: 1.18 (95% CI 1.00 to 1.38), 1.38 (1.05 to 1.38) and 1.50 (1.19 to 1.89). (For details see the second table in the full report.)
Summaries of risk associated with duration and pack-years
The panel noted the consistency of the cohort studies in observing increased risk estimates associated with longer duration and higher pack-years of smoking, whereas a number of the case-control studies reported increased risks and some did not. Selection or recall bias in some of the case-control studies and random variation are likely the main explanations for inconsistency in the case-control study findings.
Six of the seven cohort studies reporting on longer duration smoking showed increased risk for the highest duration category; two had borderline significance (RR 1.14 (95% CI 1.0 to 1.3); 1.18 (95% CI 1.0 to 1.4)) and four statistically significant increases (RR 1.21 (95% CI 1.06 to 1.45); 1.36 (95% CI 1.1 to 1.7); 1.38 (95% CI 1.08 to 1.83); and 1.50 (95% CI 1.19 to 1.89) (table 1). For risk associated with pack-years of smoking (table 1), all five cohort studies report increases in risk associated with the highest pack-year category, and four reach statistical significance (RR 1.17 (95% CI 1.02 to 1.34); 1.25 (95% CI 1.06 to 1.47); 1.46 (95% CI 1.11 to 1.93) and 1.74 (1.15 to 2.62)).
Cigarette smoking before a first full-term pregnancy
The period between the onset of puberty and first full-term pregnancy may be a time of higher risk of cancer initiation29 30. In addition, adolescence may be a critical period of risk, as shown by studies of atomic bomb survivors and women medically exposed to ionising radiation at young ages.31
Lawlor et al32 performed a meta-analysis of 12 studies of smoking and breast cancer that had examined smoking before first pregnancy. The meta-analysis resulted in a summary risk estimate of 1.07 (95% CI 0.94 to 1.22) and led the authors to conclude that there was no relationship between smoking before/during first pregnancy and breast cancer.
Since the Lawlor et al meta-analysis, at least 11 more studies, including 4 cohort studies (table 1), have been published. Overall, taking into account the relative sizes of the studies and the magnitude of the associations observed, the available data are suggestive that active smoking before a first full-term pregnancy is associated with an increase in risk of breast cancer.
Age at initiation of smoking
All eight cohort studies reporting on age of initiation of smoking have risks greater than one, four are statistically significant and two close to statistical significance (table 1). The three studies with the lowest age cut-off (age less than 15 at initiation) had higher risk estimates of 1.29 and two of 1.48. There is likely to be considerable overlap between women who began smoking at a young age and women who smoked for ≥5 years before their first full-term pregnancy. Most studies have not had the power to disentangle these closely related exposures.
Epidemiological studies of active smoking and genetics
Several different approaches to genetic predisposition have yielded increased risks for higher exposure smokers, which are summarised below.
Smoking and BRCA1 and BRCA2
In a 2008 report on collaboration among familial breast cancer registries in the USA, Australasia and the Ontario Cancer Genetics Network, a case-control study of women under 50 who were carriers of mutations in BRCA1 and BRCA2, an increased risk of breast cancer associated with as a little as 5 pack-years of smoking was seen.33 Compared to non-smokers, the risk associated with ≥5 pack-years of smoking was 2.3 (95% CI 1.6 to 3.5) for BRCA1 carriers and 2.6 (95% CI 1.8 to 3.9) for BRCA2 carriers. In both groups, risk increased 7% per pack-year (p<0.001).
The results differed from five previous studies examining BRCA carriers that had not found an increased risk associated with smoking34–38 and two that had noted a significant inverse association. However, the authors of the 2008 study suggested that because the other studies had used prevalent cases, many of which had been diagnosed many years before study, the results could have been biased towards the null or even towards a protective effect because of selective mortality associated with smoking.
Motivated by the 2007 collaborative case-control study, researchers reanalysed the earlier Ghadirian et al study.39 They did not find increased risk for smokers versus never smokers among the 764 case-control pairs of women with BRCA1 or BRCA2 that had been interviewed within 2 years of diagnosis, but for this subset of their dataset the authors presented risks only for ever versus never smoking and no dose-response results. BRCA1 carriers who were past smokers had an increased risk (OR 1.27; 95% CI: 1.06 to 1.50).
Smoking, familial history of breast cancer and breast cancer risk
The effect of smoking and familial history of cancer on cancer risk was examined in a case-control study of 18 836 incident cancer cases (3861 breast cancer cases) and 28 125 age and sex matched controls aged 20–79 collected between 1988 and 2004 in Japan.40 A significant interaction between smoking history and a family history of breast cancer in first-degree relatives was observed (interaction p=0.01). These findings reinforce earlier findings by Couch et al41 who reported that the risk associated with smoking increased the most among women with the strongest familial predisposition to breast cancer.
Smoking, genotypes and breast cancer meta-analyses
Terry and Goodman13 performed meta-analyses summarising the findings of the approximately 50 epidemiological studies that have evaluated a role for genetic polymorphisms related to carcinogen metabolism, modulation of oxidative damage, and DNA repair and the risk of breast cancer related to smoking. Inconsistent results have complicated interpretation. Most of the meta-analyses of specific gene/smoking interactions had the limitations of a small number of studies, studies with small sample sizes and varying measures of smoking. However, a fairly consistent positive association was found with long-term smoking among women with a NAT2 slow acetylation genotype, especially among postmenopausal women. (For a full summary of meta-analyses results, see the full report and table 2.)
A meta-analysis of NAT2 status and smoking was published in 2008 by Ambrosone et al,14 which included nine case-control studies and four nested case-control studies within cohorts. Among women who smoked, the risk of breast cancer was elevated among those with NAT2 slow acetylation genotypes (meta-RR 1.27; 95% CI 1.16 to 1.39), but not for those women with rapid NAT2 genotypes (meta-RR 1.05; 95% CI 0.95 to 1.17). Pack-years were significantly associated with a dose-dependent increase in risk for slow acetylators, with a RR for higher pack-years (≥20 pack-years) of 1.44, (95% CI 1.23 to 1.68) (see table 2).
Ambrosone et al14 also conducted a pooled analysis using raw data for 9 of the 13 studies (5201 cases and 5829 controls); the results were consistent with those from the meta-analysis (table 2). Increased risks for high pack-years among slow acetylators were similar for premenopausal (OR 1.49; 95% CI 1.08 to 2.04) and postmenopausal women (OR 1.42; 95% CI 1.16 to 1.74) (table 3). These analyses demonstrate a clear pattern of increased breast cancer risk associated with active smoking among women who have the NAT2 slow acetylator genotypes.
Epidemiology: SHS and breast cancer
A critical issue in the SHS and breast cancer literature is the quality of the SHS exposure assessment. Although many studies of SHS and breast cancer have been published, only six have collected a lifetime history of SHS exposure. When exposure assessment is inadequate, some exposed cases and controls will be classified as unexposed, thus contaminating the referent group. When an exposure such as that to SHS is considered, the vast majority of the population may be exposed (lifetime assessments typically find 80 to 95% of Western female populations with regular residential and/or occupational SHS exposure). In this case, inadequate SHS exposure assessment (eg, ignoring occupational exposure) can result in the majority of those categorised as ‘unexposed’ actually having been exposed to SHS, thus seriously contaminating the referent group and leading to underestimates of risks, if they should exist.
A working group of IARC2 concluded that there was no association between either active or passive smoking and breast cancer. For active smoking they relied on the Collaborative Group meta-analysis of 53 studies.28 For SHS they noted the lack of increased risk reported in two large US cohort studies,42 43 and that the lack of an active smoking risk made a risk from SHS unlikely.
The CalEPA issued a report on the health effects of environmental tobacco smoke.4 Expanding on a meta-analysis by Johnson,44 they calculated a RR of 1.68 (95% CI 1.31 to 2.15) for breast cancer among younger, primarily premenopausal women who had never smoked, associated with regular exposure to SHS, based on 14 studies. They concluded that ‘the evidence for exposure to environmental tobacco smoke and breast cancer is consistent with causality in younger, primarily premenopausal women’.
The Surgeon General reported a pooled estimate of risk of 1.64 (95% CI 1.25 to 2.14) for the 11 studies reporting on premenopausal breast cancer and SHS (table 3). They concluded ‘The evidence is suggestive but not sufficient to infer a causal relationship between secondhand smoke and breast cancer’.5
Miller et al22 published a summary of the CalEPA report breast cancer findings and presented an extended discussion of potential biases and other concerns expressed during CalEPA's formal public consultation period for the large report. The paper concluded that these further assessments of the evidence did not change the CalEPA conclusion and the authors reiterated the conclusion that SHS risk for younger primarily premenopausal women and breast cancer was consistent with causality.
Since the CalEPA report was completed, four studies of SHS and breast cancer that present results in younger women have been published. They continued to show the same patterns of breast cancer risk related to SHS exposure as the CalEPA and US Surgeon General's meta-analyses demonstrated. The one study that collected lifetime SHS exposure data and analysed the SHS risk in comparison to the group reporting no SHS exposure45 suggested increased risk among premenopausal women for SHS exposure and for active smoking, while those with poorer exposure assessment or analysis did not observe increased risk.
Based on the weight of evidence from current understanding of toxicological and biological mechanisms; increased risks observed in most cohort studies for early initiation, longer duration and/or higher pack-years of smoking; the increased breast cancer risks reported for NAT2 slow acetylators in pooled and meta-analyses; and studies suggesting higher breast cancer risk with smoking among genetically susceptible familial and BRCA1 and BRCA2 subgroups, the panel concluded that the relationship between active smoking and breast cancer is consistent with causality.
Based on the weight of evidence from current understanding of toxicological and biological mechanisms, increased risks reported by the CalEPA and the Surgeon General, and evidence of an active smoking breast cancer risk, the panel concluded that the relationship between SHS and breast cancer in younger, primarily premenopausal women is consistent with causality. The evidence was considered insufficient to pass judgement on SHS and postmenopausal breast cancer.
An unresolved issue is why the risks associated with SHS appear to be similar to those associated with active smoking when SHS is properly controlled. One possible explanation is based on a dose-related difference in antioestrogenic effects between active and passive smoking.46 Additionally, SHS risk could reflect the situation proposed for colon cancer47 where the modifying effect of a genotype might be more apparent at low doses. It is also possible that activating pathways with a low threshold become saturated at a relatively low level of exposure to tobacco smoke (in the SHS dose range) so further exposure does not result in further risk.47
The panel recommends that research be undertaken to further consolidate and synthesise knowledge of the relationship between breast cancer and tobacco smoke, especially among women known to be genetically susceptible to breast cancer.
The panel noted that comprehensive health communication strategies have repeatedly been found to be vital parts of comprehensive tobacco control. Young women in particular, should understand that available evidence suggests that the relationship between breast cancer and active smoking and SHS is consistent with causality. A call for more education, communication and increasing of public awareness about the dangers of tobacco are key features of the WHO Framework Convention on Tobacco Control.48 Well designed comprehensive health communications strategies could be effective at communicating the risk of breast cancer from exposure of women, particularly girls and young women, to tobacco smoke, whether through active smoking or exposure to SHS.
The Expert Panel would like to thank the Public Health Agency of Canada and the Ontario Tobacco Research Unit for financial support. The Expert Panel is especially grateful to the Ontario Tobacco Research Unit for including the Expert Panel's 10–11 November 2008 meeting as part of the larger conference, Tobacco Control for the 21st Century: Challenges in Research and Evaluation. The Expert Panel is also grateful to Physicians for a Smoke-Free Canada and the Canadian Partnership Against Cancer for their contributions of staff time and support of the project. We would also like to acknowledge Jodi Wilson for extensive background research and technical support, Meagan Loftin, Robert Burton and Howard Morrison for editorial input and Marilyn Pope and Sonja Johnston for final formatting of the full report.
Funding The Public Health Agency of Canada provided financial support for travel and accommodation to bring the Expert Panel together for their meeting in November 2008. The Ontario Tobacco Research Unit provided financial support of the Expert Panel by paying for the meeting space and funds for production and printing of the Expert Panel report. Views expressed in this report represent those of the panel members and do not necessarily represent the views of the respective institutions they work for.
Competing interests None to declare.
Provenance and peer review Not commissioned; externally peer reviewed.