Abstract
BACKGROUND: Early-onset breast cancer (EOBC) is rising in incidence and poses unique clinical and biological challenges, necessitating reliable prognostic tools. Given that aberrant DNA methylation is an early and stable epigenetic alteration with prognostic relevance in tumorigenesis, this study aimed to construct a methylation-based prognostic model to improve risk stratification and personalized management strategies for EOBC patients. METHODS: We conducted an integrative bioinformatics analysis of multiple datasets to identify EOBC-specific methylation markers. Differentially methylated probes (DMPs) were screened, and survival-associated CpG sites (CpGs) were refined using Cox and least absolute shrinkage and selection operator (LASSO) regression analyses. A methylation signature for EOBC (MSEO) was developed and combined with clinicopathological factors to construct a prognostic nomogram. Predictive performance was evaluated by the area under the curve (AUC) and calibration in internal and independent external cohorts. RESULTS: A total of 19,343 DMPs were identified, of which 127 significantly associated with overall survival (OS). An 11-CpG-based methylation signature was developed and validated as an independent prognostic factor. Combining MSEO with clinicopathological variables yielded a nomogram with superior predictive accuracy, achieving AUC values of 0.873-0.912 in the training cohort, 0.835-0.971 in the test cohort, and 0.756-0.868 in the external validation cohort. Calibration analyses confirmed strong agreement between predicted and observed survival. CONCLUSIONS: These findings introduce a robust methylation-based prognostic model for EOBC, providing a clinically relevant tool for individualized survival prediction and patient management.