This study combined genome-wide data from the Breast Cancer Association Consortium (BCAC) and the UK Biobank to investigate genetic susceptibility and causal risk factors for early-onset breast cancer (EOBC). The analysis set included 141,952 breast cancer cases and 280,863 age-matched controls. Cases were stratified as EOBC (<45 years) and late-onset breast cancer (LOBC; ≥45 years). The authors performed genome-wide association studies (GWAS) for EOBC, LOBC and overall breast cancer, and derived an EOBC-specific latent factor for downstream analyses. They also applied linkage disequilibrium score regression (LDSC) and Mendelian randomisation (MR) across thousands of traits obtained from publicly available exposure GWAS datasets.
GWAS identified 21 genomic risk loci for EOBC, 123 loci for LOBC and 145 loci for overall breast cancer. Among the 145 loci for overall disease, three loci close to the genes FAM175A, IFLTD1 and ITGB6 were reported as novel in this analysis. These identifications expand understanding of common-variant architecture implicated in breast cancer and show that some loci are differentially associated by age at diagnosis.
Comparing effect sizes between EOBC and LOBC across the 145 overall breast cancer loci, the average association was 1.12 times stronger for EOBC than for LOBC (P = 3.82E-05). Eighteen loci showed nominally significant effect differences between EOBC and LOBC. Notably, the ESR1 locus demonstrated a substantially larger effect for EOBC, with an estimated 67.8% greater effect relative to LOBC (95% confidence interval: 36.4% to 106.3%), meeting a P threshold of <0.05/145 in the reported analyses. These findings indicate an age-decreasing gradient in effect sizes for common susceptibility variants at some loci.
LDSC was used to screen genetic correlations (r_g) between breast cancer phenotypes and a broad array of traits. Fifteen traits showed significant genetic correlation with breast cancer, with r_g values reported in the range from −0.63 to 0.56. The psychiatric trait schizophrenia was highlighted as the only trait with a stronger genetic correlation to EOBC than to LOBC in these analyses.
Mendelian randomisation analyses tested putative causal effects of thousands of traits on EOBC and LOBC. Nineteen traits were identified as having causal effects on EOBC; these included brain imaging phenotypes and gene expression traits implicated in neurodevelopment and neurodegeneration. Fifteen traits showed greater causal effects on EOBC than on LOBC in MR analyses. Schizophrenia was the only trait that appeared in both the LDSC genetic-correlation results and the MR causal-effect findings, strengthening the signal linking neuropsychiatric pathways and EOBC susceptibility in this dataset.
The investigators derived an EOBC-specific latent factor intended to capture genetic architecture unique to early-onset disease. Variants at the ESR1 locus and signals related to schizophrenia were associated with this latent factor. The EOBC-specific latent factor explained 27% of the SNP-based genetic variance of EOBC, as reported in the study.
The combined genome- and phenome-wide analyses indicate an age-related gradient in the effect of common susceptibility variants, with many loci showing stronger associations with EOBC than LOBC. The convergence of findings implicating neurodevelopmental and neurodegenerative gene expression traits, brain imaging phenotypes, and schizophrenia suggests potential involvement of neurocognitive pathways in EOBC susceptibility. The authors interpret these results as providing new insights into both genetic susceptibility and possible causal pathways for early-onset disease.
This report is a preprint and has not been peer reviewed; the authors note that the findings should not be used to guide clinical practice until validated. The manuscript states that BCAC studies obtained local ethics approvals and that UK Biobank has appropriate ethics approval; all participants provided written informed consent. Data access pathways are provided: UK Biobank data are available upon application, BCAC data access requests are to be directed to BCAC data access, and exposure GWAS datasets were sourced from public repositories such as OpenGWAS. The report does not provide additional validation or replication beyond the combined consortium and biobank datasets in the preprint. Readers should consult the supplementary material and data/code links provided by the authors for technical details and to assess reproducibility.
Notes: All numerical values, loci counts, statistical estimates and trait findings reported here are taken directly from the source preprint. The original article is a medRxiv preprint and the authors declared no competing interests.