The relationship between DNAm-based age acceleration measures in the prediction of cross-sectional and longitudinal health outcomes and mortality

By Pooja Toshniwal PahariaReviewed by Danielle Ellis, B.Sc.Feb 24 2023

In a recent study published in PNAS, researchers assessed the association between deoxyribonucleic acid methylation (DNAm)-based clocks or aging measures and demographic and serological parameters in estimating health-associated outcomes, including deaths.

Background

DNAm-based biomarkers are garnering increased interest as estimators of health-associated outcomes and deaths among older individuals. However, data on the correlation of epigenetic aging measures and known socioeconomic and behavioral age-associated variables regarding health-associated outcomes are limited.

About the study

In the present study, researchers evaluated the relationship between epigenetic aging clocks of the first-generation (Horvath and Hannum), second-generation (GrimAge and PhenoAge), and the third-generation (the DunedinPACE clock) and health-associated outcomes and evaluated the same associations after the epigenetic clocks underwent principal component (PC) training.

The team investigated probable improvements in the estimative ability of the epigenetic measures following PC training, performed to eliminate noise and increase measurement reliability. In addition, they evaluated the performance of DNA methylation-based estimators against known estimators of health-associated outcomes, including health behaviors, socioeconomic status (SES), and demographic variables.

Further, the estimative abilities of the epigenetic clocks were compared to a composite predictor of 22 blood-based biomarkers previously reported to estimate multimorbidity. Data from older United States (US) adults, who participated in the health and retirement study (HRS), were used to assess epigenetic aging in contexts of behavioral and socioeconomic relationships with age-associated health outcomes. Increases in the epigenetic age, relative to chronological age, were evaluated using all clocks, and the impact of biological age acceleration was evaluated regardless of individual age.

Results

Aging, determined using epigenetic clocks of the second-generation and the third-generation, was consistently a significant estimator of health-associated outcomes, including cognitive disorders, functional restrictions, and chronic medical conditions evaluated after two years of DNAm measurements and four-year mortality. For HRS participating individuals (n=3,581), epigenetic clocks generated using conventional techniques and principal component training yielded different epigenetic age estimates.

Epigenetic age ranged from 54 to 67 years using the non-principal component-trained measures and 64 to 77 years post-principal component training. Except for the Horvath aging measure, PC training increased the samples’ average epigenetic age, bringing the epigenetic and chronological mean age (68 years) values closer to each other. The DunedinPACE third-generation clock had an average of 1.0, indicating slightly swifter aging biologically compared to the chronological age.

The first- and second-generation cocks correlated moderately, and the correlations were strengthened by PC training. All clocks, except GrimAge, irrespective of PC training, showed that women aged at a slower pace than men. Lower educational status was related to accelerated aging, as determined by the DunedinPACE and GrimAge clocks and the PC-trained second-generation PhenoAge clock. Aging measures of the first generation showed that Blacks aged slowly, compared to Whites, whereas the original measures of the subsequent generations showed opposite findings.

All clocks showed that Hispanics aged slowly, whereas obese individuals aged swiftly. Smoking habits accelerated aging, using the original measures (except Hannum), and excessive alcohol intake accelerated aging according to the original second and third-generation clocks.

Depression was another factor that accelerated the aging process. The original and PC-trained first-generation clocks showed childhood financial hardship associated with slower aging. Faster aging, using the second-generation clocks, showed significant associations with increased difficulties in performing activities of daily living (ADL) or instrumental ADL (IADL), evaluated two years after DNA methylation measurements. Similar findings were obtained using the PC-trained clocks. Faster aging was related to greater cognitive decline, using the second and third-generation clocks, and the association was largely the same after PC training.

Three (except the non-PC-trained Hovarth clock) epigenetic clocks and all the principal component-trained clocks significantly estimated four-year mortality. Only the PhenoAge clock estimates differed significantly after PC training. Integrating all epigenetic clocks and adjusting for covariates showed that the second-generation clocks’ age acceleration could significantly and independently estimate ADLs/IADLs and deaths, whereas the PhenoAge, Horvath, DunedinPACE, and GrimAge clocks could independently estimate multimorbidities.

Only the PhenoAge and Hannum clocks could estimate cognitive decay. Demographic estimators significantly estimated the health outcomes, especially deaths. Previous depression was the largest contributor to estimating functional limitations after two years, and SES (educational status, ethnicity/race) contributed the most to estimating cognitive decay. Comparing epigenetic clock estimations with the biomarker analysis showed that the epigenetic measures were no longer statistically significant estimators of ADL/IAD difficulty and cognitive decay. Nevertheless, the capability of epigenetic clocks to estimate mortality remained largely unaltered and reduced slightly for later-age multi-morbidities.

Conclusion

Based on the study findings, PC-based epigenetic aging measures do not significantly change the association between DNAm-based age acceleration measures and health outcomes or deaths compared to previous versions of the measures. The utility of DNAm-based age acceleration as an estimator of health outcomes in later life is clear; however, other factors such as SES, demographics, health behaviors, and mental well-being remain comparable estimators of health outcomes.