Later-generation epigenetic aging clocks outperform first-generation models in predicting survival in TCGA breast cancer.
Epigenetic aging bridges the gap between biological and chronological age by exploiting DNA methylation (DNAm) patterns. Over the past decade, successive DNAm-based clocks have been introduced, beginning with the first-generation Horvath and Hannum models and extending to second-generation PhenoAge and the GrimAge family; complementary measures include DNAm-estimated telomere length and mitotic indices such as epiTOC/pcgtAge. We previously conducted a side-by-side evaluation of these metrics in colorectal cancer using publicly available data from The Cancer Genome Atlas (TCGA) COAD and READ cohorts, but an equally systematic assessment in breast cancer has been lacking.
Here, using TCGA-BRCA tumor methylomes linked to clinical data (analytic n = 781), we compared seven metrics (Horvath, Hannum, PhenoAge, GrimAge1, GrimAge2, epiTOC/pcgtAge, DNAmTL) via Kaplan-Meier grouping (median and tertiles) and Cox models adjusted for menopausal status, age at diagnosis, receptor subtype, stage, race, and ethnicity, with overall survival truncated at 4000 days. Our analysis reproduced expected benchmark patterns: Triple Negative Breast Cancer (TNBC) had the worst outcomes, Luminal A the best, and higher stage and older age predicted poorer survival, supporting analytic validity. We found first-generation clocks did not separate survival, whereas PhenoAge and GrimAge2 stratified outcomes; in multivariable analyses, only GrimAge1 provided independent prognostic information. DNAmTL was inversely associated with mortality in univariate models, and epiTOC stratified tertiles but showed wide, nonsignificant Cox estimates.
Second-generation clocks demonstrated stronger prognostic signal than first-generation models in unadjusted analyses. Among them, GrimAge1 retained independent prognostic value beyond established clinicopathologic factors in breast cancer, supporting further external validation with richer covariates to refine clinical utility.
Here, using TCGA-BRCA tumor methylomes linked to clinical data (analytic n = 781), we compared seven metrics (Horvath, Hannum, PhenoAge, GrimAge1, GrimAge2, epiTOC/pcgtAge, DNAmTL) via Kaplan-Meier grouping (median and tertiles) and Cox models adjusted for menopausal status, age at diagnosis, receptor subtype, stage, race, and ethnicity, with overall survival truncated at 4000 days. Our analysis reproduced expected benchmark patterns: Triple Negative Breast Cancer (TNBC) had the worst outcomes, Luminal A the best, and higher stage and older age predicted poorer survival, supporting analytic validity. We found first-generation clocks did not separate survival, whereas PhenoAge and GrimAge2 stratified outcomes; in multivariable analyses, only GrimAge1 provided independent prognostic information. DNAmTL was inversely associated with mortality in univariate models, and epiTOC stratified tertiles but showed wide, nonsignificant Cox estimates.
Second-generation clocks demonstrated stronger prognostic signal than first-generation models in unadjusted analyses. Among them, GrimAge1 retained independent prognostic value beyond established clinicopathologic factors in breast cancer, supporting further external validation with richer covariates to refine clinical utility.