Hormone therapy (HT) is associated with an elevated risk of breast cancer (BC) in
postmenopausal women. Only limited studies have examined whether BC risk after HT
exposure varies by individual susceptibility. To identify loci that modify HT related
BC risk, four genome-wide case-only studies were analysed separately and combined
in a meta-analysis. We included data from the genome-wide association studies HEBCS
(Finland), MARIE (Germany), NHS (USA) and SASBAC (Sweden) which yielded 344, 742,
1,090 and 773 female, postmenopausal BC cases, respectively. Recruitment was hospital-based
for HEBCS and population-based for MARIE, SASBAC and NHS. The available SNPs (genotyped
using the Illumina 370k array in MARIE and 550 k in HEBCS, NHS and SASBAC) were used
to impute additional SNPs (software: MACH) which resulted in up to 2,800,000 SNPs
in total per study. The case-only approach is more powerful to detect gene-environment
interactions than the case-control approach if the assumption of gene-environment
independence is valid. We performed case-only logistic regression analysis for each
study separately (outcome: current versus never/past HT use) using the software ProbABEL
that takes into account the uncertainty introduced by imputation, i.e. the probabilities
of each genotype per SNP and individual are used rather than the most likely genotype.
A log-additive mode of inheritance is employed. The meta-analyses will be conducted
assuming fixed effects (software: Plink). Our study has 80% power to detect an interaction
effect of 1.25 assuming genetic and environmental main effects of 1.15, a moderate
risk allele frequency of 0.10 and a prevalence of the environmental exposure of 0.50,
Currently, analyses are ongoing and results will be presented at the conference. The
most significant results will be replicated in independent case-control studies from
the Breast Cancer Association Consortium (BCAC). We will thus rule out false positives
due to possible gene-HT dependencies.
