Background
Alterations in blood DNA methylation are considered promising biomarkers for breast cancer (BC) risk. Epigenetic age estimators (“epigenetic clocks”) which are built from methylation levels at selected CpG sites quantify biological ageing. Several studies have evaluated the association between epigenetic age acceleration (EAA) and BC risk, suggesting that accelerated ageing may distinguish BC cases from non-cases. However, the findings remain inconsistent, possibly due to small sample sizes, heterogeneity between clock models and study designs, limited replication, and potential reverse causation.
Objectives
The objectives of this study were (1) to assess the quality and harmonization of DNA methylation data in the E3N cohort and (2) to evaluate whether EAA is associated with subsequent BC risk.
Methods
We conducted a case-cohort study within the French E3N-Generations cohort, including a random subcohort of women free of cancer at the time of blood collection (1995–1999), together with all incident BC cases with blood sample available identified until June 2018. A total of 2046 women were included, of whom 1059 developed incident BC during the follow-up. DNA methylation was measured using two versions of the Illumina HumanMethylationEPIC BeadChip (v1: 866,553 CpGs; v2: 937,690 CpGs). Data from both platforms were harmonized using a multi-step pipeline, including preprocessing, normalization, and batch correction.
EAA was computed for four epigenetic clocks (Horvath, SkinHorvath, Hannum, and Levine), adjusted for chronological age and blood cell composition. Associations between EAA and BC incidence were estimated using weighted Cox proportional hazards models. Models were further adjusted for established BC risk factors (age at first birth, education, body mass index, age at menarche, parity, menopausal status, smoking, and alcohol consumption). Analyses were repeated after excluding BC cases diagnosed within 5 years after blood sampling to address reverse causation.
Results
In association analyses, both Hannum-EAA and Levine-EAA were associated with increased BC risk, with hazard ratios indicating a 20–22% higher risk per 5-year increase in EAA (FDR < 0.05). Associations were strongest among women diagnosed within 5 years after blood sampling. After excluding these early cases, only the Levine-EAA remained significantly associated with BC risk, with an approximately 20% increase in BC risk per 5-year EAA increment (FDR < 0.05).
Conclusion
In this large case-cohort study, EAA derived from second-generation clocks, especially the Levine clock, was associated with subsequent BC risk. The attenuation of associations after excluding early BC cases suggests that part of the signal may reflect preclinical disease, while the persistence of the Levine association supports its potential relevance for risk stratification. These findings highlight the importance of clock selection and longitudinal designs when evaluating epigenetic ageing as a biomarker of cancer risk. The identification of robust methylation-based ageing signatures may contribute to future BC risk assessment strategies.