Proportion of premenopausal and postmenopausal breast cancers attributable to known risk factors: Estimates from the E3N‐EPIC cohort

Abstract

Breast cancer is the most frequently diagnosed cancer among women worldwide. Breast cancer risk factors have been widely explored individually; however, little is known about their combined impact. We included 67,634 women from the French E3N prospective cohort, aged 42-72 at baseline. During a 15-year follow-up period, 497 premenopausal and 3,138 postmenopausal invasive breast cancer cases were diagnosed. Population-attributable fractions (PAFs) were used to estimate cases proportions attributable to risk factors under hypothetical scenarios of lowest exposure. We examined overall premenopausal and postmenopausal invasive breast cancers and tumour subtypes (ER status and HER2 expression). Premenopausal breast cancer was not significantly attributable to non-behavioral (61.2%, 215.5 to 91.88%) nor to behavioral (39.9%, 271.0 to 93.9%) factors, contrary to postmenopausal breast cancer (41.9%, 4.5 to 68.7% and 53.5%, 12.8 to 78.7%, respectively). Individually, the highest statistically significant PAFs were obtained in premenopause for birth weight (33.6%, 5.7 to 56.6%) and age at menarche (19.8%, 5.2 to 33.6%) for non-behavioral factors and in postmenopause for history of benign breast diseases (14.9%, 11.6 to 18.0%) and age at menarche (9.7%, 3.9 to 15.5%) for non-behavioral factors and for body shape at menarche (17.1%, 9.7 to 24.3%), use of hormone replacement therapy (14.5%, 9.2 to 19.6%), dietary pattern (10.1%, 2.6 to 17.4%) and alcohol consumption (5.6%, 1.9 to 9.3%) for behavioral factors. These proportions were higher for ER1, HER22 and ER1/HER22 postmenopausal breast cancers. Our data support the hypothesis that in postmenopause, never starting unhealthy behaviors can reduce the number of diagnosed breast cancers.

Comparison of methods for estimating the attributable risk in the context of survival analysis

“…This estimate was close to those obtained at the end of follow-up with the nonparametric methods and at the start of follow-up with the parametric and semiparametric approaches. In a recent publication, Dartois et al [12] reported a higher AR estimate of 14.5% (95% CI, 9.2 to 19.6%) for recent MHT use and postmenopausal invasive breast cancer from the E3N cohort data, using the approach proposed by Spiegelman et al [10] and a more refined, adjusted analysis with MHT exposure as a time-dependent covariate.…”

“…By June 2008 (for a maximal 16.4 years and mean 14.0 years of follow-up), 2,228 invasive breast cancers had been diagnosed (1, 106 in unexposed women). A recent work on the E3N cohort estimated a 14.5% postmenopausal breast cancer risk attributable to MHT use after 15 years of follow-up [12]. We estimated AR as a CDF-based function of time at four time points using nonparametric, semiparametric and parametric approaches, as well as the single overall AR measure proposed by Spiegelman et al [10].…”

Towards research-tested smartphone applications for preventing breast cancer

“…Modifiable risk factors for breast cancer that could be addressed in a smartphone app for preventing breast cancer in women are shown in (Table 1). Based upon attributable risks, about 30–35% of breast cancers could potentially be prevented by addressing obesity, physical inactivity, alcohol consumption, and HRT (29,30). …”

The preventable burden of breast cancers for premenopausal and postmenopausal women in Australia: A pooled cohort study

“…The burden of breast cancer preventable by modifications to certain causally related risk factors can be quantified by calculating population attributable fractions (PAFs). Only one cohort study has evaluated the individual and joint contributions of modifiable risk factors on the breast cancer burden for premenopausal women, and several have examined the breast cancer burden for postmenopausal women. All previous studies evaluated the complete removal of the exposure, even though a gradual reduction in exposure might be more realistic.…”