Context: There is limited evidence on how the risk of breast cancer and its subtypes depend on low-penetrance susceptibility loci, individually or in combination.
Objective: To analyze breast cancer risk, overall and by tumor subtype, in relation to 14 individual single-nucleotide polymorphisms (SNPs) previously linked to the disease, and in relation to a polygenic risk score.
Design, setting, and participants: Study of 10,306 women with breast cancer (mean age at diagnosis, 58 years) and 10,393 women without breast cancer who in 2005-2008 provided blood samples for genotyping in a large prospective study of UK women; and meta-analysis of these results and of other published results.
Main outcome measures: Estimated per-allele odds ratio (OR) for individual SNPs, and cumulative incidence of breast cancer to age 70 years in relation to a polygenic risk score based on the 4, 7, or 10 SNPs most strongly associated with risk.
Results: Odds ratios for breast cancer were greatest for FGFR2-rs2981582 and TNRC9-rs3803662 and, for these 2 SNPs, were significantly greater for estrogen receptor (ER)-positive than for ER-negative disease, both in our data and in meta-analyses of all published data (pooled per-allele ORs [95% confidence intervals] for ER-positive vs ER-negative disease: 1.30 [1.26-1.33] vs 1.05 [1.01-1.10] for FGFR2; interaction P < .001; and 1.24 [1.21-1.28] vs 1.12 [1.07-1.17] for TNRC9; interaction P < .001). The next strongest association was for 2q-rs13387042, for which the per-allele OR was significantly greater for bilateral than unilateral disease (1.39 [1.21-1.60] vs 1.15 [1.11-1.20]; interaction P = .008) and for lobular than ductal tumors (1.35 [1.23-1.49] vs 1.10 [1.05-1.15]; interaction P < .001). The estimated cumulative incidence (95% confidence interval) of breast cancer to age 70 years among women in the top and bottom fifths of a polygenic risk score based on 7 SNPs was 8.8% (8.3%-9.4%) and 4.4% (4.2%-4.8%), respectively. For ER-positive disease the corresponding risks were 7.4% (6.9%-8.0%) and 3.4% (3.1%-3.8%), respectively; while for ER-negative disease they were 1.4% (1.2%-1.6%) and 1.0% (0.8%-1.2%). The findings did not differ materially according to the number of SNPs included in the polygenic risk model.
Conclusions: The polygenic risk score was substantially more predictive of ER-positive than of ER-negative breast cancer, particularly for absolute risk.