Similar to the entire CTS cohort, subjects in this analysis were predominantly non-Hispanic white (86%) and almost half were post-menopausal at baseline. Compared to those who did not have a breast cancer diagnosis, cases tended to be older and more frequently report a family history of breast cancer. The vast majority (76%) of tumors diagnosed among women in this analysis were hormone responsive.
The distributions of women according to their passive smoking exposure by timing and setting are presented in . Overall, 86% of study subjects reported some passive smoking exposure during their lifetime. Most reported household exposure (71%), half reported workplace exposure and more than one-third reported exposure in a social setting (37%). Approximately two thirds of respondents reported exposure before the age 20 and nearly three-quarters reported exposure at age 20 or older. Among those with exposures before age 20, the household was the predominant setting (58%), followed by considerably less exposure in social (20%) and workplace (17%) settings. In contrast, adult exposures (age 20 or older) were most commonly reported for the workplace (47%), followed closely by household (43%) and social settings (34%).
Passive smoke exposure by timing and setting among study participants (n=57,523).
The distributions of adult exposures in the three setting are shown in . The distributions of exposures in the workplace and home, among those exposed, were virtually identical, and both were lower than the distribution of exposures in social settings. The final box plot in shows the distribution of the sum of the intensity-years measures across the three settings, where the large upper tail of the “All Settings” box plot reflects exposure levels that would not be captured by examining household exposures only. Moreover, the workplace exposures appear to be at least as intense as home exposures.
Figure 1 Distributions of intensity-years among adults for three settings. The box shows the 25th and 75th percentiles, the dot shows the median, the whiskers above and below the box show the range of the bulk of the distribution, and the dots show extreme values. (more ...)
Exposures by setting and age group were not mutually exclusive, although correlations were modest. The Spearman correlation coefficient(r) for the intensity-years measure, for instance, was 0.25 for overall childhood and adulthood exposures (among those exposed in both ages). For adult exposures, the Spearman r was 0.23 for home and work exposures, and 0.39 for work and social settings (among those exposed in all three settings).
The point estimates for risk of breast cancer associated with any exposure (ever/never) to passive smoking reported by study participants did not substantially differ from the null in any of the age group or setting categories, and risk estimates from age stratified and race/birthplace adjusted models were similar to those from the models fully adjusted for personal risk factors, albeit with wider confidence intervals (). Although not statistically significant, the highest point estimates were observed for lifetime exposure opportunity, rather than for any particular setting or age group.
Invasive breast cancer Hazard Ratios (HRs) and 95% Confidence Intervals (95% CI) associated with never/ever passive smoking exposures, by setting and by age of exposure, among 1,754 cases diagnosed 1997-2007 in 57,523 study participants.
Analyses evaluating the cumulative measures for each component of exposure (years and intensity) contributing to the summary measure of intensity-years and the summary measure itself generally did not show consistent and significant associations for particular settings or for overall lifetime exposures or for those under age 20 years (). Point estimates were significantly elevated for the highest exposure categories among women exposed at ages 20 years or older for years and intensity-years in the age/race adjusted models, but remained statistically significant in the fully adjusted models only for the combined summary measure, intensity-years (HR=1.18, 95% CI=1.00,1.40). A test for trend, however, was not statistically significant (p=0.30).
Invasive breast cancer Hazard Ratios (HRs) and 95% Confidence Intervals (95% CI) associated with years, intensity, and intensity-years of passive smoke exposure for 1,754 cases diagnosed 1997-2007 among 57,523 study participants.
When attention was confined to those subjects with a modest to large exposure to environmental tobacco smoke (total intensity-years > 4 for all exposures), a statistically significant nonlinear dose response curve emerged for exposure modeled on a log scale (). After stratifying by age and adjusting for race, log2 intensity-years had a marginally statistically significant linear term (HR=1.06, 95% CI=1.01, 1.11, p=0.02), corresponding to a 6% increase in hazard for every doubling of intensity-years. shows how the estimated slope of a linear dose response curve in log2 years changes as we vary the threshold used to confine our attention. As the lower-exposed subjects are eliminated from consideration by increasing the threshold for “modest to large” exposures, the linear part of the dose-response increases as well, reaching an asymptote of around 1.13 at a threshold of approximately 20 intensity-years. Hence, for exposures exceeding 20 intensity-years, a doubling of exposure results in an approximately 13% increase in hazard. At that point, fitting becomes more and more unstable as the number of subjects diminishes. To relate this result back to , note that the thin vertical line in , corresponding to a threshold of 4, intercepts the smoothed hazard curve at a value of approximately 1.06, corresponding to the 1.06 calculated above. As expected with a dose-response curve that is concave at lower doses, restricting one's attention to dose-response curves that are linear in log2 intensity-years biases the effect estimate for doubling large exposures downward.
Figure 2 Smoothing spline representation of the dose response curve for total intensity-years in adults from all sources. The curve represents the logit of relative risk as a function of dose, was calculated using a binomial Generalized Additive Model, and arbitrarily (more ...)
Figure 3 The effect of changing the lower threshold for inclusion in the fit on the coefficient of the slope term in a proportional hazards model relating breast cancer to log2 intensity-years. Each dot represents an estimated hazard ratio for proportional hazards (more ...)
Risks stratified by menopausal status are presented in . Women who were peri-menopausal at baseline were excluded from this analysis, as the number of events in this group was too small for reliable point estimates. There was no evidence for elevated breast cancer risks from passive smoking among pre-menopausal women at baseline. Separate analyses for women who were pre-menopausal at entry but were followed for a cancer event only up to the age of 50 (as a proxy for menopause), produced slightly elevated but non-statistically significant risk estimates for the highest level of cumulative lifetime exposure (HR=1.20; 95% CI 0.70 – 2.05), based on 33 cases in this category of exposure. For postmenopausal women, hazard ratios for lifetime low, medium and high cumulative exposure were 1.17 (95% CI 0.91, 1.49), 1.19 (95% CI 0.93, 1.53, and 1.26 (95% CI 0.99, 1.60), respectively.
Invasive breast cancer Hazard Ratios (HRs) and 95% Confidence Intervals (95% CI), by menopausal status at baseline, associated with any exposure and intensity-years of passive smoke exposure in the California Teachers Study.
Analyses evaluating categories of exposure separately by setting (household, workplace and social) for those exposures occurring at ages 20 years or older yielded globally null results (data not shown). Those who reported workplace exposures, the predominant source of exposure during this age period, in the medium and highest categories had only slightly elevated risk estimates with confidence limits that included 1.0 (HR=1.05 in both categories) compared to those with no reported workplace exposure.
Passive smoking exposure in relation to a women's first pregnancy was categorized among parous subjects in the study (n=44,680). A slightly elevated risk of breast cancer was observed for women with known passive smoking exposure prior to a first pregnancy in age stratified and race/birthplace adjusted models (HR 1.21, 95% CI 1.00-1.47), as well as the fully adjusted models (HR 1.17. 95% CI 0.96-1.41).
Finally, we conducted a number of stratified analyses to evaluate whether risks differed by family history of breast cancer, or tumor hormone responsiveness. Results from these analyses did not vary by subgroup and were similar to those observed in the full study sample (data not shown).