Abstract
BACKGROUND: NSABP P-1 provides an opportunity to examine the association of behavioral factors with prospectively monitored cancer incidence and interactions with tamoxifen. METHODS: From 1992 to 1997, 13,388 women with estimated 5-year breast cancer risk greater than 1.66% or a history of lobular carcinoma in situ (87% younger than age 65; 67% postmenopausal) were randomly assigned to tamoxifen versus placebo. Invasive breast cancer, lung cancer, colon cancer, and endometrial cancer were analyzed with Cox regression. Predictors were baseline cigarette smoking, leisure-time physical activity, alcohol consumption, and established risk factors. RESULTS: At median 7 years follow-up, we observed 395, 66, 35, and 74 breast cancer, lung cancer, colon cancer, and endometrial cancer, respectively. Women who had smoked were at increased risk of breast cancer (P = 0.007; HR = 1.3 for 15-35 years smoking, HR = 1.6 for ≥ 35 years), lung cancer (P < 0.001; HR = 3.9 for 15-35 years, HR = 18.4 for ≥ 35 years), and colon cancer (P < 0.001; HR = 5.1 for ≥ 35 years) versus never-smokers. Low activity predicted increased breast cancer risk only among women assigned to placebo (P = 0.021 activity main effect, P = 0.013 activity-treatment interaction; HR = 1.4 for the placebo group) and endometrial cancer among all women (P = 0.026, HR = 1.7). Moderate alcohol (>0-1 drink/day) was associated with decreased risk of colon cancer (P = 0.019; HR = 0.35) versus no alcohol. There were no other significant associations between these behaviors and cancer risk. CONCLUSION: Among women with elevated risk of breast cancer, smoking has an even greater impact on breast cancer risk than observed in past studies in the general population. IMPACT: Women who smoke or are inactive should be informed of the increased risk of multiple types of cancer.