Does Tobacco Smoke Cause Breast Cancer?

“It is biologically plausible that tobacco smoke could increase breast cancer risk, given its many carcinogenic constituents, their association with (smoking-related) changes in DNA, and their detection in the breast tissue of smokers.”

More than 1 billion people currently smoke tobacco [1], and tobacco consumption is the leading cause of cancer worldwide. Even more people, especially women in developing countries, continue to be exposed to second-hand smoke. Consequently, efforts to determine whether tobacco smoke affects breast cancer risk, and if so, the nature and strength of the relationships with active and passive smoking, have important public health as well as scientific implications. Some researchers have posited that tobacco smoke increases breast cancer risk as much or more so among women who are exposed passively to tobacco smoke as in those who smoke actively. If this were true, then reductions in active and passive smoking would be among the most important interventions for the prevention of breast cancer, in addition to their numerous other health benefits. However, to date, no scientific consensus has been reached on these issues, despite more than 20 years of debate and over 150 epidemiological studies [1–3].

The views of breast cancer epidemiologists on these issues have fluctuated over the past 20 years. In the early 1990s, Baron and colleagues [4] proposed that active smoking might protect against breast cancer through its presumed antiestrogenic effects, as is thought to be the case with endometrial cancer [5]. However, subsequent studies failed to demonstrate this inverse relationship, showing either no association or a small increase in risk associated with active smoking [3]. The idea that smoking might increase breast cancer risk was later tested in a pooled collaborative analysis of 53 epidemiological studies with 83,000 breast cancer cases [6]. No association was observed between breast cancer and ever having smoked actively in the subgroup of women who reported no alcohol consumption. Therefore it was concluded, that residual confounding by alcohol consumption, an established risk factor for breast cancer, likely introduces a spurious positive association between smoking and breast cancer [6]. Although the authors suggested restricting future analyses of this issue to lifelong nondrinkers, few studies have been sufficiently large for this purpose.

The null findings reported by the collaborative group have been challenged more recently since the crude definition of active smoking used in that analysis did not consider duration of smoking or time periods of potential vulnerability, such as exposure during the period between menarche and first full-term pregnancy [2,7]. Indeed, there are now several examples in the literature of studies in which long-term cigarette smoking or years of smoking before first full-term pregnancy was more strongly associated with risk than was ever-smoking [2,3]. As yet, no pooled analysis has tested these hypotheses in large studies of women who report no regular consumption of alcohol.

Although the majority of epidemiological studies have demonstrated no clear association with risk [3], several investigators and consensus reviews have concluded that a causal association between active smoking and breast cancer cannot be ruled out, given recent updates from large prospective studies reporting: positive associations (overall 10–30% increased risk) in women who actively smoked large numbers of cigarettes over many years; increased risk in women who smoked prior to a first full-term birth; and evidence from studies suggesting that alcohol consumption may not explain positive associations with smoking, including recent analyses in nondrinkers [8,9]. There is also conflicting evidence regarding whether the association with active smoking may be confined to genetically susceptible subgroups, particularly women with the NAT2-slow or very slow acetylation genotypes [10–12].

Several scientific consensus groups have evaluated this evidence repeatedly between 2004 and 2009 and have reached different conclusions, although the assessments are generally converging for active smoking (Table 1). Both the International Agency for Research on Cancer (IARC) and the US Surgeon General Report reviewed the evidence for active smoking in 2004 and concluded that that there was evidence suggesting a lack of carcinogenicity [13,14]. Several years later, in 2009, another review by IARC changed the designation of the evidence for active smoking to ‘limited‘, based on updated evidence from large prospective studies previously mentioned [1]. A 2009 report of the Canadian Expert Panel on Tobacco Smoke and Breast Cancer Risk [2] concluded that the association between active smoking and breast cancer risk was ‘consistent with causality‘, particularly in premenopausal women. This designation follows the terminology of the California Environmental Protection Agency (EPA). Although it appears to be stronger than the IARC category of limited evidence, the two are actually synonymous with respect to the criteria that define them.

Table 1.

Conclusions of various scientific consensus groups on the association between smoking and breast cancer.

Agency (year)	Active smoking	Passive smoking	Ref.
IARC (2004)	Evidence suggesting lack of carcinogenicity	–	[14]
Surgeon General (2004)	Evidence suggestive of no causal relationship	–	[13]
California EPA (2005)	–	Data are consistent with causality in premenopausal women	[101]
Surgeon General (2006)	–	Evidence is suggestive but not sufficient to infer a causal relationship	[15]
Canadian Panel (2009)	Data are consistent with causality	Data are consistent with causality in premenopausal women	[2]
IARC (2009)	Limited evidence	Evidence is inconclusive	[1]

EPA: Environmental Protection Agency; IARC: International Agency for Research on Cancer.

Regarding passive smoking, IARC considered the evidence to be conflicting and unclassifiable in its 2004 and 2009 evaluations, due to concerns regarding bias in the case–control studies and lack of conclusive mechanistic evidence to account for the stronger than expected (i.e., nonlinear) association with second-hand smoke reported in these studies. The 2006 US Surgeon General Report on passive smoking [15] concluded that “the evidence is suggestive but not sufficient to infer a causal relationship”. By contrast with these evaluations, a 2005 report from the California EPA Air Resources Board concluded that the weight of the evidence was consistent with a causal relationship [101]. This conclusion was based predominantly on case–control studies that these reviews considered to have the most comprehensive evaluation of lifetime exposure to second-hand smoke. Both IARC and the US Surgeon General had reviewed many of the same studies and considered them to be highly susceptible to reporting bias. The varying opinions of these consensus groups largely reflect differences in emphasis placed on the results of different types of studies.

There are several critical reasons why researchers have failed to reach a consensus on whether active and/or passive smoking affects breast cancer risk. First, the association of active smoking with breast cancer risk is (at most) considerably weaker than the associations observed with other established smoking-related cancers in large prospective studies. Better mechanistic information is required to document whether active smoking inhibits or potentiates the development of breast cancer through hormonal effects that have been hypothesized, but are as of yet not well defined. Second, case–control studies that use as their referent group women who report no lifetime exposure to either active or passive smoking are vulnerable to reporting and other biases that can undermine validity. Third, it is implausible that second-hand smoke is as strongly or more strongly associated with breast cancer risk than active smoking, unless there is a definable mechanistic explanation to account for a nonlinear relationship between tobacco smoke exposure and breast cancer risk.

Other issues have limited the interpretability of the published literature. There have been many reports of the results of single studies, and many reviews that present selected results from these studies, but there have been very few systematic efforts to test specific hypotheses in pooled or parallel analyses that control adequately for other established risk factors for breast cancer. This is particularly true with respect to various parameters of exposure that have been postulated to be important. For example, it is plausible that years of exposure between menarche and age at first pregnancy could be important. However, unless this can be shown to be the case in parallel or pooled analyses of multiple studies that control adequately for potentially confounding factors, it remains an hypothesis rather than an established observation. It is also reasonable to be concerned that residual confounding by alcohol may introduce a spurious association between active smoking and breast cancer. Large pooled analyses in nondrinkers are best able to resolve this uncertainty as well. Furthermore, it continues to be more difficult to publish null results than positive associations between smoking and breast cancer risk. For example, the findings of the meta-analysis of Ambrosone and colleagues [10] that observed effect modification by NAT2 genotype received considerable attention in the Canadian Expert Panel's report [2]. These findings were not replicated in a larger pooled analysis based on prospective cohorts in the National Cancer Institute Breast and Prostate Cancer Cohort Consortium (BPC3) [102], but the null results have been difficult to publish.

It is biologically plausible that tobacco smoke could increase breast cancer risk, given its many carcinogenic constituents, their association with (smoking-related) changes in DNA, and their detection in the breast tissue of smokers [3]. Active smoking is already considered causally related to at least 17 different sites or subsites of cancer [1,16]. Given this, it would actually be surprising if the 4000 or more chemicals in tobacco smoke did not affect breast cancer risk. The potential relationships between breast cancer and both active and passive smoking have strong implications for advocacy and risk communication. Women are generally more concerned about breast cancer than about lung cancer or the host of other deleterious effects of smoking on health [103]. A credible link to breast cancer could be a more potent deterrent than other health warnings in preventing adolescent girls from starting to smoke, and could motivate stronger policies restricting second-hand smoke.

Research on the association of smoking with breast cancer has continued in the 2 years since the Canadian and IARC panels met to address this issue, during which time the results of six cohort studies [17–22] and eight case–control studies [23–30] have been published. The results of these studies have been mixed, with some demonstrating positive associations with active [18,21,23,25,27,30] or passive [22,25] smoking and others showing no clear association with active [19,20,24,26,28,29] or passive [17,29] exposure. Given the issues previously discussed, it seems premature either to dismiss the possibility of a causal association between breast cancer and active and/or passive smoking or to accept the current evidence as definitive.

That cigarette smoking can cause breast cancer in genetically susceptible women remains a plausible hypothesis and a potentially important public health concern, especially given the continued high smoking rates in young women and widespread exposure to second-hand smoke [31]. However, it is unlikely that this controversy will be resolved by continuing to emphasize single studies. A new collaborative effort is required to bring together all of the prospective information to test clearly defined hypotheses in pooled analyses, using a consistent definition of exposure and properly controlling for, or stratifying on, all of the established risk factors for breast cancer.