Article Text

other Versions

Research Letter
Prevalence of rules prohibiting home and workplace smoking correlates with state-specific breast cancer outcomes: an ecologic analysis
  1. Kenneth M Piazza,
  2. Andrew Hyland
  1. Department of Health Behavior, Roswell Park Cancer Institute, New York, New York, USA
  1. Correspondence to KM Piazza, Department of Health Behavior, Roswell Park Cancer Institute, CH A313, Elm & Carlton Streets, Buffalo, New York, NY 14263, USA; kenneth.piazza{at}

Statistics from

Some review panels have concluded that second-hand tobacco smoke (SHS) causes breast cancer in premenopausal women. The California Environmental Protection Agency's Office of Environmental Health Hazard Assessment did a meta-analysis in 2005 of 19 studies on breast cancer and SHS exposure and concluded that regular exposure to SHS is ‘causally related to breast cancer diagnosed in younger, primarily premenopausal women.’ The meta-analysis, using a weight-of-evidence approach and specific criteria to deal with methodological issues common in the epidemiologic literature, yielded a combined OR of 1.68 (95% CI 1.31 to 2.15) for the 14 studies with analyses that included younger, primarily premenopausal women.1

In spite of such reviews, a firm agreement on the causal nature of an association between SHS and breast cancer remains elusive. In 2006, the Surgeon General's Report examined the same issue and concluded that the evidence was ‘suggestive but not sufficient’ to infer causality. They found a pooled risk estimate of 1.45 (1.04 to 2.01).2 While more well-designed longitudinal studies with adequate exposure assessment are certainly to come and add to those already in the literature, we hypothesise that if a causal relationship between SHS and breast cancer does exist, then breast cancer incidence and mortality rates would be lower in those states with a greater fraction of never-smoking women with rules not allowing smoking in their homes and workplaces.

From 1990 to 2000, the prevalence of rules not allowing smoking in the home has risen 40%.3 The prevalence of smoke-free workplaces has likewise risen dramatically over the last two decades. In roughly the same period, combination therapies and better screening have contributed to a decline in breast cancer mortality, a better detection and a shift in incidence to lower stage tumours. The intention of this ecologic study is to explore the hypothesis that the prevalence of rules not allowing smoking in the home and workplace correlates with state-specific rates of breast cancer incidence and mortality.

We used data from the 1992 to 1993 Tobacco Use Supplements of the Current Population Survey (TUS-CPS)4 in order to assess the prevalence of these rules in 73 204 never-smoking, non-Hispanic Caucasian women, for each state and the District of Columbia. The prevalence of rules was defined as the percentage of never-smoking women answering that ‘smoking was not allowed anywhere in the home’ and ‘smoking was not allowed anywhere’ in the workplace versus the responses that smoking was allowed ‘in some’ or ‘in all’ areas of these places. Pearson correlations were used to examine the associations between the prevalence of these rules in 1992–1993 with state-specific rates of pre- and postmenopausal breast cancer after a plausible lag period. The outcome variables were invasive breast cancer incidence and mortality rates for each state per 100 000 women of the same racial subgroup averaged over the 7-year period from 1999 to 2005, first using all-age data and then data stratified by women of age ≤50 years and >50 years using age-specific rates.5 Moderate inverse correlations were observed between the percentage of women in each state, who stated smoking was not allowed anywhere in their home and workplace in 1992–1993 and breast cancer incidence and mortality rates in women aged 50 years and younger at outcome and for mortality among women of all ages (see figure 1A,B,E).

Figure 1

Correlations between prevalence of smoke-free rules in the home and workplace in 1992–1993 and breast cancer outcomes in 1999–2005, stratified by age.

Because states that have higher percentages of rules not allowing smoking may have healthier values for other risk factors for breast cancer incidence and mortality, multivariate linear regression models were used to adjust for potential confounders for each outcome presented in figure 1. Data on potential confounders of age (continuous), race (categorical), household income (categorical: <$25 000, $25 000–$74 999, $75 000+) and education (categorical: <HS grad, some college and degree or higher) were from TUS-CPS census data of 1992 and were specific for each individual reporting rules not allowing smoking in their homes and at their workplaces.4 State-specific percentages of fruit and vegetable consumption (≥5 per day), physical exercise (>90 min per week), body mass index (BMI) (>30) and alcohol consumption (>1 drink per month) were not specific for age, gender or race.6 State-specific data on age-at-first-live-birth (percentage >30 years) were used from the National Vital Statistics System.7 All potential confounders were included in stepwise, backward selection regression models for both incidence and mortality.

Since state-specific mammography rates correlate significantly with mortality and more so incidence, especially in younger women <50 years, age-specific rates were used for the age-stratified models: mammography within 2 years for younger women aged 40+ and within 2 years for older women aged 50+. For the mortality models, rates for both age groups used the 1995 data to allow a plausible lag for mortality outcome; for incidence models, rates for both age groups used the 2000–2001 data since any change in breast cancer incidence attributable to mammography would occur in a short time span.6

Even after adjusting for these covariates, the correlations between the prevalence of rules not allowing smoking in both the home and workplace in 1992–1993 and all-age mortality and incidence in younger women remained strong and statistically significant.

While causality cannot be inferred from this ecologic analysis, the findings are consistent with the hypothesis that SHS causes breast cancer in younger, primarily premenopausal women.


View Abstract


  • Kenneth M Piazza and Andrew Hyland contributed equally to this project.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.