Associations of accelerated epigenetic aging with cancer and mortality risk in the USA.

Abstract

DNA methylation (DNAm) algorithms have been developed to assess biological aging and its association with cancer. Despite their potential, direct comparisons to identify the most accurate algorithm for predicting cancer risk and mortality remain limited. The study population (n = 2532) consisted of adults aged 50 years and older from the National Health and Nutrition Examination Survey, with 17-year follow-up mortality data. Twelve DNAm algorithms were evaluated using Illumina EPIC BeadChip array. Logistic regression models were used to assess both overall and site-specific cancer risk, while Cox proportional hazards models and Fine-Gray sub-distribution hazard model were employed to assess cancer mortality. Three hundred forty-three cancer cases were observed at baseline, and 271 cancer-caused deaths were recorded during the follow-up. GrimAgeMortAcc, GrimAge2MortAcc, and VidalBraloAgeAcc were positively associated with overall cancer risk, with multivariable-adjusted odds ratios per standard deviation increase of 1.44 (95% CI: 1.06-1.95), 1.32 (95% CI: 1.01-1.72), and 1.20 (95% CI: 1.01-1.44), respectively. PhenoAgeAcc, GrimAgeMortAcc, and GrimAge2MortAcc were associated with increased cancer risk in women (particularly non-Hispanic White women), while no significant associations were observed in men, including for prostate cancer specifically. Several DNAm algorithms showed strong inverse associations with skin cancer risk. In addition, higher HorvathAgeAcc was linked to an increased risk of cancer mortality, with multivariable adjusted hazard ratio 1.19 (95% CI: 1.04-1.37). This study reveals a close association between several DNAm algorithms (particularly GrimAge) and cancer risk and mortality. These algorithms offer promising tools for advancing precision medicine, with potential applications in cancer prevention and risk stratification.