Distributions of baseline characteristics for selected variables by smoking status are presented in . At study initiation 52.9% (1194/2258) of the study participants reported that they had never smoked, 39.5% (891/2258) were former smokers, and 7.7% (173/2258) were current smokers. The majority of the women were early-stage breast cancer survivors with 81% in stages I or IIa at the time of diagnosis. The median age was 58 years (SD=11.0 years; range 25–79 years) at the time of breast cancer diagnosis; 20% of the participants were non-white and nearly 27.4% had a high school level education or below. Never and former smokers tended to be older, and more educated than current smokers.
We noted a significant variation in the distribution of tumor stage (p-value=0.02): 83.2% of current, 81.9% of former and 78.7% of never smokers had stage I or IIa disease. Current and former smokers were more likely to have HER2 negative tumors compared to never smokers (78%, 79.7% and 71.9% respectively, p-value=0.002). Significant differences were also observed in menopausal status with 60.1% of never, 69.7% of former, and 56.6% of current smokers reporting post-menopausal status (p-value<0.0001). Former and current smokers were more likely to consume larger quantities of alcohol than never smokers: 28.3% of current smokers and 30.0% of former smokers consumed more than 6 grams of alcohol per day compared to 13.5% of never smokers (p-value<0.0001). Never smokers were also more likely to receive chemotherapy than former or current smokers (60.6% vs. 52.6% and 56.6% respectively, p-value=0.002).
The median follow-up in our analytic cohort of 2,258 women was 12.3 years (standard deviation=2.9 years; range 1.5–15.5 years). The median follow-up among current smokers was significantly shorter than the median follow-up among former and never smokers (11.9 vs. 12.2 vs. 12.4 years respectively; p=0.0005). During this period, a total of 485 deaths was observed: 215 among 1194 never smokers, 213 among 891 former smokers, and 57 among 173 current smokers. Of these 485 deaths, 241 of deaths were due to other causes (105 among never smokers, 105 among former smokers, and 32 among current smokers) and 244 were due to breast cancer (111 among never smokers, 108 among former smokers and 25 among current smokers).
Kaplan-Meier plots of survival by smoking status for all-cause survival (), competing-cause survival (), and breast cancer-specific survival () revealed differences by smoking status in all-cause, competing-cause, and to lesser extent in breast cancer survival. In , we present results indicating that current smokers had significantly higher risk of all-cause, competing-cause and breast cancer-specific mortality than never smokers. Current smokers had an increased risk of death from any cause in unadjusted (HR=2.03, 95% CI 1.51–2.72), and covariate-adjusted models (HR=2.63, 95% CI 1.93–3.58). Former smoking was also associated with increased risk of all-cause mortality in both unadjusted (1.38, 95%CI 1.14–1.67) and covariate-adjusted models (1.28, 95%CI 1.05–1.56). Both current and former smoking were associated with increased risk of competing-cause mortality, with covariate-adjusted hazard ratios of 3.84 (95%CI 2.50–5.89), and 1.33 (1.00, 1.78), respectively. Compared to never smokers, current smokers had an approximately two-fold increase in the risk of breast cancer death in unadjusted (HR=1.71, 95%CI 1.11–2.64), and covariate-adjusted models (HR=2.01, 95%CI 1.27–3.18). Former smoking was associated with increased risk of breast cancer mortality in unadjusted models (HR=1.35, 95%CI 1.04–1.76). After adjusting for all potential confounders, the association was attenuated and did not reach statistical significance (HR=1.24, 95%CI 0.94–1.64). We further examined the association between ever-smoking and all mortality outcomes and found similar magnitude and direction of associations (data not shown).
Kaplan-Meier plots of survival, by smoking status
Hazard ratios and 95% CIs for the effect of smoking on mortality among women with breast cancer
We next examined whether tumor characteristics such as ER, PR and HER2, menopausal status, and BMI at cohort entry modified the effect of smoking on breast cancer mortality. When interaction terms were included in multivariate models, we found significant interactions with normal BMI (p-value= 0.003), and HER2 positivity (p-value=0.02), but not with ER, PR and menopausal status. Stratified analyses revealed increased risk of breast cancer death in current smokers with both ER positive and negative tumors (); the association did not reach statistical significance among those with ER negative tumors, likely due to the small number of breast cancer deaths in this group. We found no evidence of effect modification by PR or HER2 status.
Covariate adjusted hazard ratios for the effect of smoking on breast cancer mortality stratified by tumor characteristics, body mass index, and menopausal status*
Possible variations in the effect of smoking status on breast cancer survival according to BMI were also evaluated (). Fully adjusted models suggest that both current and former smoking increases the risk of breast cancer death across all BMI strata. While the association was statistically significant among current smokers (HR=4.46, 95%CI 2.03–9.76) and former smokers (HR=1.70, 95%CI 1.00–2.90) of normal weight (BMI<25), the association did not reach statistical significance among overweight and obese women (BMI ≥25) who were former or current smokers.
We next evaluated whether the association of smoking status with breast cancer mortality varied according to menopausal status. We found an increased risk of breast cancer death in pre and postmenopausal women who were current smokers (). Among former smokers, we found an increased risk of breast cancer mortality among pre-menopausal women (HR=2.73, 95%CI 1.40–5.32), but not post-menopausal women.
For the systematic review, we identified six published cohort studies of the association, and these are summarized in together with our own study. In general, we observed increased a stronger risk of breast cancer mortality among women with breast cancer who were current smokers than among those who were former smokers. Notably, three of the seven studies in the systematic review reported risk ratios for estimates of breast cancer mortality and four reported hazard ratios for this outcome. Given these differences in the effect measures reported, we chose not to combine the studies quantitatively using meta-analysis.
Systematic review of smoking and breast cancer mortality