METHODS: We built two models, for ER+ (ModelER+) and ER- tumors (ModelER-), respectively, in 281,330 women (51% postmenopausal at recruitment) from the European Prospective Investigation into Cancer and Nutrition cohort. Discrimination (C-statistic) and calibration (the agreement between predicted and observed tumor risks) were assessed both internally and externally in 82,319 postmenopausal women from the Women's Health Initiative study. We performed decision curve analysis to compare ModelER+ and the Gail model (ModelGail) regarding their applicability in risk assessment for chemoprevention.
RESULTS: Parity, number of full-term pregnancies, age at first full-term pregnancy and body height were only associated with ER+ tumors. Menopausal status, age at menarche and at menopause, hormone replacement therapy, postmenopausal body mass index, and alcohol intake were homogeneously associated with ER+ and ER- tumors. Internal validation yielded a C-statistic of 0.64 for ModelER+ and 0.59 for ModelER-. External validation reduced the C-statistic of ModelER+ (0.59) and ModelGail (0.57). In external evaluation of calibration, ModelER+ outperformed the ModelGail: the former led to a 9% overestimation of the risk of ER+ tumors, while the latter yielded a 22% underestimation of the overall BC risk. Compared with the treat-all strategy, ModelER+ produced equal or higher net benefits irrespective of the benefit-to-harm ratio of chemoprevention, while ModelGail did not produce higher net benefits unless the benefit-to-harm ratio was below 50. The clinical applicability, i.e. the area defined by the net benefit curve and the treat-all and treat-none strategies, was 12.7 × 10- 6 for ModelER+ and 3.0 × 10- 6 for ModelGail.
CONCLUSIONS: Modeling heterogeneous epidemiological risk factors might yield little improvement in BC risk prediction. Nevertheless, a model specifically predictive of ER+ tumor risk could be more applicable than an omnibus model in risk assessment for chemoprevention.
METHODS: Using data from 272,098 women participating in the European Prospective Investigation into Cancer and Nutrition (EPIC) study, we assessed dietary intake of 92 foods and nutrients estimated by dietary questionnaires. Cox regression was used to quantify the association between each food/nutrient and risk of breast cancer. A false discovery rate (FDR) of 0.05 was used to select the set of foods and nutrients to be replicated in the independent Netherlands Cohort Study (NLCS).
RESULTS: Six foods and nutrients were identified as associated with risk of breast cancer in the EPIC study (10,979 cases). Higher intake of alcohol overall was associated with a higher risk of breast cancer (hazard ratio (HR) for a 1 SD increment in intake = 1.05, 95% CI 1.03-1.07), as was beer/cider intake and wine intake (HRs per 1 SD increment = 1.05, 95% CI 1.03-1.06 and 1.04, 95% CI 1.02-1.06, respectively), whereas higher intakes of fibre, apple/pear, and carbohydrates were associated with a lower risk of breast cancer (HRs per 1 SD increment = 0.96, 95% CI 0.94-0.98; 0.96, 95% CI 0.94-0.99; and 0.96, 95% CI 0.95-0.98, respectively). When evaluated in the NLCS (2368 cases), estimates for each of these foods and nutrients were similar in magnitude and direction, with the exception of beer/cider intake, which was not associated with risk in the NLCS.
CONCLUSIONS: Our findings confirm a positive association of alcohol consumption and suggest an inverse association of dietary fibre and possibly fruit intake with breast cancer risk.
METHODS: We used a panel of 34 putative susceptibility genes to perform sequencing on samples from 60,466 women with breast cancer and 53,461 controls. In separate analyses for protein-truncating variants and rare missense variants in these genes, we estimated odds ratios for breast cancer overall and tumor subtypes. We evaluated missense-variant associations according to domain and classification of pathogenicity.
RESULTS: Protein-truncating variants in 5 genes (ATM, BRCA1, BRCA2, CHEK2, and PALB2) were associated with a risk of breast cancer overall with a P value of less than 0.0001. Protein-truncating variants in 4 other genes (BARD1, RAD51C, RAD51D, and TP53) were associated with a risk of breast cancer overall with a P value of less than 0.05 and a Bayesian false-discovery probability of less than 0.05. For protein-truncating variants in 19 of the remaining 25 genes, the upper limit of the 95% confidence interval of the odds ratio for breast cancer overall was less than 2.0. For protein-truncating variants in ATM and CHEK2, odds ratios were higher for estrogen receptor (ER)-positive disease than for ER-negative disease; for protein-truncating variants in BARD1, BRCA1, BRCA2, PALB2, RAD51C, and RAD51D, odds ratios were higher for ER-negative disease than for ER-positive disease. Rare missense variants (in aggregate) in ATM, CHEK2, and TP53 were associated with a risk of breast cancer overall with a P value of less than 0.001. For BRCA1, BRCA2, and TP53, missense variants (in aggregate) that would be classified as pathogenic according to standard criteria were associated with a risk of breast cancer overall, with the risk being similar to that of protein-truncating variants.
CONCLUSIONS: The results of this study define the genes that are most clinically useful for inclusion on panels for the prediction of breast cancer risk, as well as provide estimates of the risks associated with protein-truncating variants, to guide genetic counseling. (Funded by European Union Horizon 2020 programs and others.).