Quantification of Epigenetic Aging in Public Health

Abstract

Estimators of biological age hold promise for use in preventive medicine, for early detection of chronic conditions, and for monitoring the effectiveness of interventions aimed at improving population health. Among the promising biomarkers in this field are DNA methylation–based biomarkers, commonly referred to as epigenetic clocks. This review provides a survey of these clocks, with an emphasis on second-generation clocks that predict human morbidity and mortality. It explores the validity of epigenetic clocks when considering factors such as race, sex differences, lifestyle, and environmental influences. Furthermore, the review addresses the current challenges and limitations in this research area.

INTRODUCTION

In the field of public health, accurately measuring aging is increasingly important. This pursuit is driven by the need to understand the factors that contribute to aging and to identify effective interventions that extend the healthy lifespan. Researchers historically relied on telomere length as a biological indicator of aging. However, recent studies have shifted to focus on a broader array of aging hallmarks.

The year 2013 was notable for the publication of not only the pan-tissue epigenetic clock but also a highly influential review article on the hallmarks of aging (99). These hallmarks were defined as molecular and cellular processes that are associated with changes across the lifespan and have the possibility to increase or decrease with treatment, interventions, or lifestyle changes (99). The hallmarks included genomic instability, telomere attrition, epigenetic alterations, loss of proteostasis, deregulated nutrient sensing, mitochondrial dysfunction, cellular senescence, stem cell exhaustion, and altered intercellular communication. In 2023, the hallmarks of aging were updated to include impaired macroautophagy, chronic inflammation, and dysbiosis (100). The same year saw the publication of the pan mammalian methylation clock that measures age in all mammalian species (102).

In this review, we focus on DNA methylation–based biomarkers of aging (epigenetic clocks), as they provide the most precise aging estimators across humans and other mammalian species (102). However, these measures of epigenetic aging likely capture only a portion of the overall biological aging process. The twelve hallmarks are not independent and are highly correlated with each other. Indeed, in vitro studies have revealed that epigenetic clocks are linked to nutrient sensing, mitochondrial activity, and stem cell composition but seem to be disconnected from cellular senescence, telomere attrition, and genomic instability (72). Thus, even though this review discusses epigenetic alterations, specifically DNA methylation, it is important to account for the other hallmarks. Epigenetic aging, an indicator of biological age, may provide a nuanced perspective of aging and has the potential to be used as a proxy for future risks for mortality and morbidity, to determine potential risk and aging consequences of various exposures, and to identify the impact of an intervention on someone’s risk profile.

DNA METHYLATION AND EPIGENETIC CLOCKS

DNA methylation of cytosine [5′ methylcytosine (5mC)] is the most common DNA modification and occurs mainly at cytosines in a CpG dinucleotide context in differentiated mammalian cells. The stability of 5mC in biological samples, including DNA stored over long periods, allows for extensive data collection for subsequent high-throughput analysis.

Over the past decade, we have made striking discoveries about the potential of epigenetic shifts to determine a person’s age. Epigenetics involves the chemical adjustments and structuring of the genome, which can impact or signify its activity. The effects of age on global methylation levels were first noted by Berdyshev et al. in 1967, as reviewed in Horvath & Raj (67). However, high-fidelity platforms that measure DNA methylation at the single nucleotide level were needed to develop highly accurate age prediction models that deserved the label “clocks.” These DNA methylation–based age estimators have revolutionized how we measure age across all tissues and nucleated cells by tracking the abundance of molecules (methyl groups) that attach to DNA cytosines (11, 56, 61, 65, 153).

We define epigenetic clock as a multivariate age (or mortality risk) predictor, based on a multivariate regression model involving multiple cytosines. The first epigenetic clock, developed in 2011, estimated age from human saliva samples (11), followed by the pan-tissue clock in 2013, which applied to all human DNA sources except sperm (61). Other significant tissue-specific estimators include Hannum and colleagues’ (56) blood sample clock and various sparse clocks based on a few CpGs (65, 91, 103, 153). We group the many different epigenetic clocks by generation.

GENERATIONS OF EPIGENETIC CLOCKS

The classification into first-, second-, and third-generation epigenetic clocks generally refers to the progression in the development and utility of these clocks.

First-Generation Epigenetic Clocks

First-generation clocks are typically designed to predict chronological age accurately across a wide range of tissues and cell types. These include the original Hannum clock (56). Another example is the Horvath pan-tissue clock, which uses 353 CpG sites (regions of DNA where a cytosine nucleotide is followed by a guanine nucleotide) to predict chronological age (61). While these clocks can reflect biological age and are associated with a variety of age-related conditions, they were not initially designed to predict health outcomes. Even though research on first-generation clocks continues using novel approaches, including deep learning (31, 46, 124), the most precise clocks may not lead to significant biological insights (163). These clocks have high correlations (r) with chronological age, with the majority above 0.9 all the way up to 0.99 (150). However, even with high correlations, the median absolute error (MAE) is high at 3.8 years, which suggests that while these metrics are valuable for population-based studies, they may not be accurate enough for personalized medicine (34).

Second-Generation Epigenetic Clocks

Second-generation clocks were developed to predict not just chronological age but also health span and lifespan and thus are often associated with age-related morbidity and mortality. The two most familiar clocks in this category are the PhenoAge and GrimAge clocks (91, 103). GrimAge is especially well suited to predict time-to-death, time-to–coronary heart disease, and time-to-cancer, with hazard ratios for mortality between 1.1 and 1.8 among various studies (59, 101, 109, 113). The GrimAge clock was built using several DNA methylation–based surrogate measures for plasma proteins, including leptin and plasminogen activator inhibitor-1, and smoking pack-years (103). Lu et al. (101) recently developed an updated version of GrimAge that also includes DNA methylation–based estimates of c-reactive protein and hemoglobin A1C, which slightly enhances the association with morbidity. The updated version outperforms the earlier version across racial and ethnic groups, with a hazard ratio between 1.06 and 1.15 for mortality and between 1.06 and 1.21 for coronary heart disease (101).

Researchers developed two epigenetic DNA methylation–based biomarkers that do not estimate age per se but estimate a rate of aging instead: Dunedin POAM (pace of aging methylation) and PACE (6, 7). The outcome, pace of aging, was estimated based on health data, such as body mass index, HbA1c, grip strength, lung function, cognitive decline, and subjective health status following individuals over a 20-year period (6, 7). Individuals with a PACE rate faster than 1 standard deviation (SD) or more above the mean, compared with those with an average rate (within 1 SD of the mean), had a hazard ratio of 1.26 [95% confidence interval (CI): 1.14–1.40] for mortality and a hazard ratio of 1.16 (95% CI: 1.12–1.20) for chronic disease morbidity (7).

A recent longitudinal study among European Caucasians validated the use of epigenetic clocks with mortality risks, when reviewing both epigenetic age at baseline and change of epigenetic age over time (84). The most accurate prediction for mortality was for the PhenoAge and GrimAge clocks with a hazard ratio of 1.31 (95% CI: 1.17–1.46) per SD for baseline and 1.23 (95% CI: 1.10–1.37) for the slope for the PhenoAge as well as a hazard ratio of 1.46 (95% CI: 1.29–1.65) per SD for baseline and 1.13 (95% CI: 1.02–1.26) for the slope for the GrimAge clock (84). This study indicates that both baseline epigenetic age and changes in epigenetic age over time independently predict mortality, implying that interventions that lower epigenetic age may reflect changes in mortality risk.

Third-Generation Epigenetic Clocks

Third-generation clocks utilize cytosine methylation levels within highly conserved DNA segments to determine chronological age across multiple species. Thus, a single regression model based on the same covariates (cytosines in conserved stretches of DNA) is used to predict age across multiple species (63, 102, 145). The resulting multispecies clocks hold the potential to bridge the gap between animal model research and human applications, thus facilitating the translation of research findings. This work is especially important for ongoing animal intervention studies, as successful interventions may not translate in humans. Several pan-mammalian clocks have been developed, including a mammalian clock for estimating relative age, which is defined as the ratio between age and maximum lifespan (102). The pan-mammalian clock is also useful for predicting human mortality risk in epidemiological cohort studies. Furthermore, conditions such as obesity and inflammation, measured by c-reactive protein, appear to accelerate aging according to the pan-mammalian clock, whereas vegetable intake seems to slow it down (102). Currently, second-generation clocks, such as GrimAge, are significantly more accurate in predicting human mortality risk and associated factors compared with the pan-mammalian clock (102), whereas third-generation clocks are more useful for translating animal-to-human studies aimed at interventions that influence the aging process.

FUTURE OF EPIGENETIC CLOCKS

Over the last few years, many new types of epigenetic clocks have been developed. Not merely chronological age trackers, these clocks have been instrumental in revealing associations between accelerated aging and numerous health aspects through analysis of diverse human DNA methylation profiles (67).

The most recent epigenetic clocks can be divided into new epigenetic clocks using novel modeling techniques or more complicated methodology (39, 46, 108) developed for a specific subpopulation (e.g., for children or centenarians) (12, 33, 52, 78, 111) or for different tissue, including placental tissue (88, 129), and saliva/buccal cells (35, 111, 133).

Although the second-generation epigenetic clocks are correlated with each other and are all associated with life- and health span, each individual clock was developed on its own training dataset and may be associated with slightly different health-related outcomes. When an exposure or outcome is associated with multiple clocks, the association is robust. In contrast, when associations are identified among only a subgroup of clocks, this finding may indicate the presence of different underlying mechanisms or could be due to spurious findings (119, 130).

In the following discussion, we present a curated selection of significant observations concerning epigenetic aging and its implications for public health rather than offering an exhaustive compilation of all available epigenetic clocks. Throughout most of this review, our attention is directed toward the three second-generation epigenetic clocks that are predominantly utilized at the time of writing. We exclude studies that lie outside the realm of second-generation clocks because first-generation clocks were covered in several previous reviews (5, 40, 67).

UNDERLYING MECHANISMS PROPOSED WITH THE EPIGENETIC CLOCKS

Age-related gain of methylation can be observed at polycomb repressive complex 2 (PRC)-binding sites in all mammalian species, highlighting the intricate relationship between fundamental developmental pathways and aging mechanisms (102). Moqri et al. (115) recently developed clocks that focus exclusively on methylation levels at PRC2 target sites.

Many epigenetic clock studies provide substantial backing for Barker’s hypothesis concerning the developmental origins of chronic diseases (105). The epigenetic clock theory of aging posits that DNA methylation patterns provide a continuous readout of age throughout the entire lifetime, linking purposeful processes during development and cancer protection to later adverse effects causing organ dysfunction (67). Developmental aging theories emphasize how intentional and deterministic developmental processes can lead to organ dysfunction as time progresses (within the framework of antagonistic pleiotropy) (67). de Magalhães (32) recently proposed the analogy that aging could arise from a developmental software error. Overall, extensive research indicates that developmental programming plays a significant role in chronic diseases, promising an unexpected interpretation of aging and its related dysfunctions.

Each epigenetic clock is based on a distinct set of CpG sites and effect estimates. These associations are frequently agnostic to biology and derived through a mathematical optimization procedure (such as penalized regression). However, since each clock is constructed using a particular population and diverse aging groups (clocks for pediatric patients versus those for centenarians), the underlying mechanisms may differ among specific epigenetic clocks. Furthermore, while clocks serve as proxies for aging, it remains uncertain which underlying CpGs are causal or simply reflect aging processes. One way to identify the potential causal effects of clocks and aging is through Mendelian randomization (MR) analyses. MR has been used to select CpGs that have a causal effect on age-related conditions (159). This ambiguity underscores the complexity of interpreting epigenetic clock data and emphasizes the need for further research to elucidate their underlying mechanisms and implications for aging (67).

Various Tissues Used in Clocks

Most epigenetic clocks were developed using blood samples, since blood samples are often available from previously established cohort studies (83, 120, 144). However, blood is not always the tissue of interest, e.g., when reviewing whether certain diseases are associated with an increased epigenetic age. While this discrepancy in tissue is a caveat of the most commonly used clocks, investigators have described a modest to moderate correlation between blood and brain and between blood and muscle tissue (53, 98, 135, 141). Other studies have indicated that certain tissues tend to measure older or younger than anticipated using first-generation clocks; e.g., breast tissue tends to age faster, whereas the cerebellum appears to age more slowly (56, 61, 64). However, aging may vary across tissues within an individual, indicating heterogeneity of systems aging. To address this issue, one can use tissue-specific epigenetic clocks, e.g., the cortical clock (132) or the skeletal muscle clock (147).

While blood samples are commonly used in observational studies of adults, collecting blood can be challenging in pediatric cases or very elderly individuals. Consequently, several epigenetic clocks have been developed using saliva or buccal cells (70).

Specific Age Categories

Specialized epigenetic clocks are emerging to cater to specific subpopulations. Notably, clocks have been devised for pediatric and centenarian cohorts to address the extremes of the lifespan. Epigenetic age tends to be lower among the older population. Centenarians or supercentenarians can have 5–20 years lower age acceleration (1, 28, 30, 36, 54, 66, 80), possibly indicating either slower aging leading to their longevity or the influence of training data sourced primarily from the adult lifespan (typically between 18 and 80 years) (151).

Dec et al. (33) devised a clock tailored specifically for centenarians and conducted an epigenome-wide association study (EWAS) to explore the aging process, stratified by age category. This analysis revealed nuanced differences in aging processes between centenarians and younger individuals, suggesting that previously developed epigenetic clocks might not accurately capture epigenetic age in this older demographic. This centenarian clock estimates primarily chronological age and thus is categorized as a first-generation clock. Consequently, while it may not be as effective at predicting mortality and morbidity risk compared with second-generation clocks, it serves a valuable role in validating claims of exceptional old age.

In addition to the centenarian clocks for the end of the lifespan, several clocks have been developed for the early part of the lifecycle. There are currently four clocks to decipher gestational age: two based on blood tissue (e.g., cord blood) (12, 78) and two based on placenta tissue (88, 107). Placental age has been associated with early-onset preeclampsia and prenatal depression, suggesting that placental aging may be part of the underlying mechanism (107, 129). One clock, the NeoAge, was developed to determine neonatal aging in preterm infants; it uses buccal cells and can estimate the gestational (postmenstrual) age and the postnatal age (52). For children, the best epigenetic clock at this point is the pedBE clock, based on buccal cells, which has a strong correlation (r = 0.98; MAE 0.35 years) with chronological age (111). PedBE has been associated with maternal exposure during pregnancy, suggesting that even in-utero contact with harmful exposures or lifestyle could influence the aging process of the offspring (43, 140). Early exposures, such as early-life adversity, have been associated with increases in PedBE, which is subsequently associated with long-term neurodevelopmental effects in later stages of childhood (48, 134). One study has shown that pedBE can be reduced through positive parenting intervention in children with developmental delays (142), suggesting that PedBE may serve as both an indicator of success and an indicator of high-risk children for future health interventions.

Sex Differences

According to a study by Horvath et al. (62) in 2016, men typically exhibit an older epigenetic age compared with women when blood samples are evaluated. This finding was confirmed in a recent study using second-generation clocks that used samples worldwide and a diverse sample in the United States (27, 160). In addition to these findings, a recent study indicated sex differences with several epigenetic clocks. The study showed specific patterns that varied by sex; e.g., among females, GrimAge and GrimAge2 were associated with obesity and depression (increase of 0.9–1.9 years in age acceleration among females versus 0.4–0.7 years among males), whereas in males, stronger associations were identified with hypertension, diabetes, and chronic kidney disease (increase of 0.9–3.0 years among males versus 0.2–0.8 years among females) (123). Although cross-sectional studies cannot determine directionality, they may indicate differences in underlying mechanisms or risk profiles that could explain variations in prevalence. These differences could also be attributed to variations in hormone profiles between men and women.

Influence by Hormones

Gender differences underscore the potential link between hormone levels and epigenetic aging. Our understanding of the relationship between hormone levels and epigenetic clocks is still in its infancy, making it challenging to establish broad principles regarding their interaction. However, given the fundamental role of hormones in human development and bodily maintenance, they may significantly influence the pace and progression of the epigenetic clock.

One of the areas where this relationship becomes evident is during puberty. The rate of pubertal development in girls has been connected to their epigenetic age (10) and epigenetic age acceleration at about age 40 (55). This notion suggests that the hormonal changes occurring during this critical developmental stage may have a substantial impact on an individual’s biological aging process and could highlight a critical window of time for interventions.

Menopause, a significant hormonal transition in women, is another stage that appears to have a profound impact on the epigenetic clock. Early onset of menopause is associated with an increase in epigenetic age (90). This observation suggests that the decline in estrogen levels characteristic of menopause may accelerate biological aging, but further research is needed to confirm this hypothesis. Conversely, menopausal hormone therapy is associated with a lower epigenetic age in buccal cells but not in blood (90). A recent cohort study indicated that hormone therapy among postmenopausal women was associated with a decrease in phenotypic age, with larger beneficial effects among women with lower socioeconomic status (SES), though overall effects were very minimal (96). Although this study did not use epigenetic age, phenotypic age is strongly correlated with epigenetic PhenoAge (r = 0.74) (90), suggesting that similar associations would likely be observed for PhenoAge. These studies highlight that hormone therapy may have beneficial implications on mortality and morbidity and on how epigenetic age can be used as a proxy for the efficacy of an intervention.

In men, a correlation has been observed between testosterone levels and epigenetic clocks (86). This association implies that variations in testosterone, a hormone integral to male physiology, and testosterone–estradiol ratio, an indicator for hormonal balance (androgens/estrogens), might influence the pace of the epigenetic aging process. However, the nature and extent of this relationship require more extensive investigation.

In addition to sex hormones, hormones related to the somatotropic axis have been proposed to influence biological aging. The somatotropic axis is a regulatory system that includes hormones such as growth hormone and insulin-like growth factors. Several studies have found increased longevity in animal models with growth hormone receptor knockouts. Disruptions of the somatotropic axis lead to decreased epigenetic age in mouse models. Mice lacking growth hormone production (Snell mice) or growth hormone receptor knockout mice (GHRKO mice) exhibit lower epigenetic ages according to most mouse clocks, including the pan-mammalian clock (101).

Some recent studies have found that metformin, a well-known medication for diabetes that improves insulin sensitivity and lowers glucose levels, has been associated with a decrease in epigenetic age among monkeys (157). In addition, two small observational studies indicated that metformin, or metformin in combination with growth hormone and dehydroepiandrosterone, was associated with lower and decreasing epigenetic age by around two years, respectively (38, 93). As such, a phase II trial is currently recruiting to assess whether metformin does improve methylation age among males as well as other indicators for senescence. Overall, these novel studies indicate that there may be interventions related to hormones that could protect individuals from biological aging and improve mortality/morbidity risks. However, how much these hormones could protect, or whether specific subgroups, such as males only, would benefit more from treatment, still needs to be established.

VALIDATION AMONG EPIGENETIC CLOCKS AMONG ETHNIC POPULATIONS

Several studies have validated these clocks across ethnically diverse populations. For instance, both PhenoAge and GrimAge have been validated to be associated with mortality and age-related morbidity in US populations of different ancestries, including Hispanic, African, and European ancestry. Another example is a Taiwanese study that used second-generation clocks (95). Furthermore, these clocks were validated in most European populations such as an Irish longitudinal population, demonstrating associations with mortality and age-related outcomes (109). While Dunedin PACE was originally developed on white New Zealanders, subsequent studies have observed changes in Dunedin PACE across diverse ethnic backgrounds, including Han Chinese, Native Americans, African Americans, and Hispanics (114). In addition, GrimAge2 was developed on an admixed population and subsequently validated across various ethnic groups, including White non-Hispanics, African Americans, and Hispanics (101).

DIFFERENCES OF EPIGENETIC AGING AMONG VARIOUS POPULATIONS

Variations in epigenetic aging among different ethnic populations could stem from limitations inherent in the development of epigenetic clocks. Nevertheless, the main second-generation clocks are now applied across diverse populations and have undergone validation. Discrepancies have indeed been observed when comparing ethnic groups (49, 62, 97, 160). One large meta-analysis studied Pheno-, GrimAge, and Dunedin PACE across 22 countries and 31 ethnicities, drawing data from the Gene Expression Omnibus. Individuals were derived from all continents, with the largest populations from the United Kingdom and United States (160). This study indicated that Chinese and Malays exhibit accelerated aging compared with Indians, while White Americans age marginally slower than do African Americans, Latinos, and Asian Americans. Yakuts age at a quicker rate than do Russians, and among all the populations considered, Congolese, Malays, and Tsimanés demonstrate the most rapid aging (160). These distinctions are not attributed solely to ethnic background but also reflect geographical and regional disparities. In broad terms, White non-Hispanic individuals appear to age more slowly. The observed increases in epigenetic age among minorities are theorized to result from diverse environmental and psychosocial exposures. To assess whether certain populations may be at higher risk for accelerated biological aging, we need to examine the impact of environmental and psychosocial factors in relation to their genetic imprint in order to identify modifiable risk factors that could be targeted to reduce biological aging disparities.

ENVIRONMENTAL AND PSYCHOSOCIAL CHARACTERISTICS AND EPIGENETIC AGING

Research on some of these environmental and psychosocial factors is ongoing and rapidly expanding, highlighting how variables such as socioeconomic factors, psychosocial stress, trauma, environmental exposures, and lifestyle can influence epigenetic aging. This growing body of evidence underscores the importance of how these factors may influence aging and also who in the general population may be at an increased risk for developing aging-related morbidity or is at a higher risk for premature death.

Socioeconomic Factors

Several studies have identified that poverty is associated with an increased epigenetic age (2, 19, 24, 29, 110, 131, 143). Economic hardship during adolescence during the Great Recession has been associated with an average of a 1.4-year increase in Hannum and Horvath age among African Americans several years later (postrecession) (24). Several studies have indicated that this finding may be different for specific subgroups within their populations. However, results are not consistent and do not indicate a specific high-risk subpopulation. For example, in the Healthy Aging in Neighborhoods of Diversity across the Life Span (HANDLS) cohort, poverty was associated with an increased Dunedin PACE, with a 7% increase in aging rate. This effect was stronger among populations of European ancestry (12% increased rate versus 5% among those with and without childhood poverty), whereas in individuals who identified as African American, there was no increase in PACE related to childhood poverty status (10% versus 10%); however, Dunedin PACE was significantly higher among the African American population above the poverty line compared with those of European ancestry above the poverty line (131). This finding may suggest that other socioeconomic factors, but not poverty, are associated with an increased epigenetic age among African Americans (110). Assari & Zare (2) identified an effect of childhood poverty on PhenoAge at age 15 (−0.09 years per income-to-poverty ratio), with a pronounced effect among males (−0.13 years), but a reduced and not statistically significantly effect was observed among females (−0.06). A study discovered that the impact of childhood poverty on epigenetic age acceleration was partially mediated by lifetime smoking (48–80%). Although these results are promising, mediation was measured at the same time as the outcome. In addition, the underlying pathways may be more complex than a direct mediation pathway (110). Even after adjusting for age, race/ethnicity, sex, several poverty characteristics, marital status, education, and lifestyle (e.g., alcohol, smoking), food insecurity was associated with an increased GrimAge, as well as an increased second-generation Zhang clock (143).

Neighborhood deprivation or SES has also been associated with increased epigenetic age (average of a 0.3–0.8-year increase in epigenetic age acceleration) (71, 87, 89, 106, 128, 136, 137). One study found that this association (r = −0.09 among total population with Dunedin PACE) was slightly reversed among individuals who reported physical activity or a healthy eating index, suggesting that individual-level behavior may protect from biological aging. Unfortunately, the effects of this positive behavior were very minimal and appeared to differ by ethnic background (106).

Lower education is a well-known risk for higher epigenetic age, potentially mediated through unhealthy lifestyles (27, 42, 81, 127). One large, multicohort study indicated that educational attainment was associated with GrimAge acceleration (34 excess deaths per 10,000 person-years for those with education inequalities) and mortality (57 excess deaths per 10,000 person-years) and that an increase in GrimAge acceleration was mediating the increase in mortality by 60%. However, this effect was seen only among males, whereas the effect of educational attainment was minimal on both GrimAge and mortality (42). However, another study indicated that upward mobility in education across multiple generations and social mobility from childhood to later life was associated with lower Dunedin PACE within the Framingham Heart Study cohort (50, 51), though results were not consistent with findings in the Health and Retirement Study, which did not find any evidence of upward mobility and epigenetic aging (81).

As more advanced and better-powered studies emerge, we will gain extensive knowledge of how poverty, neighborhood exposure, and education contribute to accelerated aging and identify those at highest risk for negative consequences. However, the complexity of interactions between various socioeconomic variables, along with inconsistencies across studies in identifying who is most at risk and how these associations are mediated by modifiable factors, makes it challenging to determine the best public policy course of action. Nevertheless, continued research will provide valuable insight that could guide targeted interventions and the identification of high-risk populations.

Psychosocial Stress and Trauma

The relationship between psychosocial stress and accelerated epigenetic aging has been explored in several review papers, highlighting a consistent link between chronic stress and increased epigenetic age across multiple studies (45, 60, 104, 121, 125, 138, 161). For example, in the Women’s Health Initiative, stressful life events were linked to a 0.34-year increase in GrimAge acceleration (138). Similarly, perceived stress at age 11 was associated with a 0.14-year increase in GrimAge at age 40 (89). Cumulative stress was associated with epigenetic aging by 0.48 years, which was modulated by emotion regulation and self-control (57). In addition, several other studies identified that resilience may buffer the effect of stress on epigenetic age acceleration (8, 162).

Perceived (racial) discrimination has been associated with an increased GrimAge among various racial groups, even after adjusting for income, education, and neighborhood disadvantage, with an average epigenetic age acceleration between 0.4 and 1.8 years (99, 106–108). In the Health and Retirement Study, perceived discrimination was associated with increases in various epigenetic clocks, including GrimAge and DunedinPoAm. Depressive symptoms partially mediated this increase by 20–50%, highlighting the potential need to address depressive symptoms in order to mitigate the impact of discrimination on biological aging (9). While perceived racial discrimination has been associated with an increase in Hannum age [0.62 years in the Strong African American Healthy Adult Project (SHAPE) and 1.45 years in the Adults in the Making (AIM) study] among African American youth, supportive family appears to mediate this association (−0.54 years for SHAPE; −0.43 years for AIMS for the interaction between discrimination and family support). Although these findings are very promising and were consistent across two study populations, further studies need to validate these findings in a longitudinal study (18). This result aligns with the hypothesis that even in adverse environments, social support and resilience can partially mediate negative effects.

Traumatic events, such as adverse childhood experiences (ACEs), also have detrimental effects on epigenetic aging. Multiple studies have reported that ACEs are associated with an increase in second-generation clocks among young adults (6, 7, 14, 55, 58, 74). In addition, both lifetime trauma and post-traumatic stress disorder (PTSD) have been linked to accelerated epigenetic aging (14, 60, 73, 94, 156). Emerging evidence suggests that individual coping styles and resilience may partially mediate the relationship between trauma and epigenetic age acceleration (8). However, the relationship between resilience and trauma is complex, with some studies indicating that this relationship requires further in-depth investigation (112, 118).

In conclusion, there is consistent research supporting a link between psychosocial stress, discrimination, and trauma and a small increase in epigenetic age. Resilience and social support are emerging as potential moderators. However, future studies, particularly longitudinal studies and those with more complex modeling, are essential to further clarify these relationships and their implications for mitigating biological aging.

Environmental Toxins

Research on the impact of air pollution on epigenetic age is limited and challenging due to methodological constraints. While some studies suggest a link between higher air pollution levels and an increase in second-generation epigenetic age, inconsistencies exist, and pinpointing the exact causative exposures remains difficult (79, 139, 154). For instance, one study found that among non-Hispanic White women, certain clusters of PM_2.5, possibly from sources such as wood smoke and surface oil, were associated with a six-year increase in PhenoAge per interquartile range (154). In addition, several studies suggest that these effects are more pronounced among Black participants, possibly due to their higher exposure levels (79). Exposure to metals, including arsenic, was associated with an increased epigenetic age (0.4-year, 0.8-year, and 0.01-year increase in PhenoAge, GrimAge, and Dunedin PACE, respectively) among a Native American community (15, 69). Arsenic gestational exposure was also linked to epigenetic age at birth, while prenatal and early-life arsenic exposures were suggestively associated with PhenoAge among Chilean adults (16, 17).

Lifestyle Characteristics

While psychosocial stress and environmental exposures appear to increase epigenetic aging, this outcome may be moderated by a healthy lifestyle. Physical activity is associated with a decrease in aging-related morbidity and mortality, whereas the effect on epigenetic age seems more limited. For example, the effects of physical activity on GrimAge acceleration were null among the Coronary Artery Risk Development in Young Adults (CARDIA) study participants at age 20 (75), whereas other studies indicated that physical activity was associated with slower epigenetic aging, potentially due to changes in immune cell composition and a decrease in cardiovascular risks (3, 41, 44, 75, 117, 126).

Previous studies have indicated that a healthy diet with higher intake levels of fruits, vegetables, and fish or a Mediterranean diet is associated with a lower epigenetic age, indicating a slightly younger biological age (47, 76, 82, 92, 126, 146, 149). These effects may already be present in utero, as one study found that maternal intake of more fat was associated with increased epigenetic age acceleration by six weeks among newborns, whereas vitamin D supplementation may decrease epigenetic age by eight weeks (122).

In addition, several studies in humans, similar to animal studies, have reviewed calorie restriction as a potential means to increase lifespan using epigenetic clocks as a proxy for change in biological age and effect of intervention on the risk profile for decreased mortality. Caloric restriction decreased the Dunedin PACE in the Comprehensive Assessment of Long-Term Effects of Reducing Intake of Energy (CALERIE) study, with effects still visible after 2 years by ~0.02 years, though no effects were found for the PC GrimAge or PC PhenoAge (152).

Alcohol is well known for a J-shape association with mortality, where the use of alcohol in limited portions has the lowest mortality rate. Alcohol consumption has been associated with a slight increase in epigenetic age [0.45 year increase in GrimAge per ln(drinks/week) among African Americans and between 0.4 and 0.8 years in PhenoAge/GrimAge acceleration per drink in the Framingham Heart Study] (13, 148, 164). The effect may be stronger when using an epigenetic proxy instead of socially desirable answers from a questionnaire (4).

Smoking is a well-known risk for various comorbidities and mortality and, as such, is also very strongly associated with epigenetic aging among diverse populations (20, 23, 27, 77, 101, 103, 164). It is important to highlight that GrimAge and GrimAge2 both use an epigenetic proxy for smoking pack-years as a component within the clock because it is strongly associated with health span and lifespan (101, 103). While smoking itself is associated with an increase in epigenetic age, smoking cessation can reverse this effect. One study found that three months after smoking cessation, epigenetic age acceleration within the airway cells was similar to that among nonsmokers (155).

Healthy sleep has consistently been associated with lower epigenetic age (142–145). To date, only a few studies have examined the effects of insomnia symptoms and short sleep on epigenetic aging. These studies have indicated increased epigenetic aging among women postpartum and postmenopausal with insomnia symptoms and sleep deprivation and among an older population in the Health and Retirement Study in the United States (21, 22, 85). These studies could indicate that short sleep duration causes an accelerated aging process (3, 41, 44, 75, 117).

INTERVENTION AND HUMAN CLINICAL TRIALS

Given the effects of lifestyle on epigenetic age in cross-sectional studies, researchers have conducted several longitudinal studies on lifestyle interventions. Epigenetic clocks have been employed in several phase 1 clinical trials. While the primary objective of these trials was to demonstrate safety rather than effectiveness, there were indications of a possible epigenetic rejuvenation effect, as noted in studies by Fahy et al. (38) and Clement et al. (26). Lifestyle interventions to slow aging, resulting in a reduced epigenetic age, include caloric restriction and exercise. In addition to lifestyle interventions, use of medications such as metformin and rapamycin has been proposed (116).

One small study among females indicated a decrease in GrimAge after women were assigned a low-glycemic, high-antioxidant, and mainly plant-based diet but found no effects for physical activity (41). A healthy Mediterranean diet with high polyphenols has been associated with a decrease in epigenetic age after 18 months (158).

Caloric restriction has shown strong effects on longevity in animal studies. However, feasibility of restricting calories among humans is limited, and it would be important to recognize that long-lasting restriction could also have negative effects. The CALERIE study randomly assigned adults to a 25% caloric restriction with no effects on PhenoAge or GrimAge but a modest effect on PACE. Dunedin PACE was decreased even after two years of the intervention; however, no measurements were performed after the intervention ended (152). While calorie restriction is a very promising intervention, it may be difficult to implement in the real world.

In addition to lifestyle and medication, plasmapheresis has been proposed to have rejuvenation effects and has had promising results in animal studies (25, 37, 68). So far, one small study in humans indicated that plasma transfusion with human umbilical cord plasma injected weekly for 10 weeks decreased GrimAge by an average of 0.8 years (26).

While these intervention studies are promising, more research needs to be performed. The effect estimates appear limited, and some of these interventions are not feasible to implement for a longer duration or on a larger population. The studies that have been performed so far were done on a small sample size, with limited ethnic diversity, and the duration of these effects postintervention has yet to be determined.

THE FUTURE OF EPIGENETIC CLOCKS IN PUBLIC HEALTH RESEARCH

Hundreds of studies have demonstrated the relevance of epigenetic clocks to disease risk factors. These clocks have shown significant associations with numerous lifestyle factors such as diet, exercise, smoking, and stress, all of which play pivotal roles in determining the risk of various diseases. By providing a measure of biological age that reflects the cumulative impact of these factors, epigenetic clocks can offer a more nuanced understanding of disease risk. In future public health research, investigators must carefully consider the DNA source and recognize that a difference in epigenetic age acceleration, unrelated to the exposure of interest (e.g., confounding by genetic ancestry and sex), may exist within the study populations. In addition, while DNA derived from blood samples is commonly utilized, it may not fully capture tissue-specific effects. Therefore, it is important to exercise caution when interpreting epigenetic age acceleration as a proxy for diseases related to specific tissues.

Epigenetic clocks have shown predictive power for certain age-related conditions, including metabolic syndrome, cardiovascular disease, atrial fibrillation, kidney disease, and even certain types of cancer. In some instances, accelerated epigenetic aging precedes the onset of these diseases, highlighting their potential for early detection and prevention in large populations. However, further development is needed to determine if improved epigenetic clocks can achieve the accuracy required for individual health monitoring and personalized health care. Improved future versions of epigenetic clocks could transform our approach to disease prevention and public health promotion.

While current epigenetic clocks are not yet accurate enough for personalized medicine, they can help identify subpopulations at higher risk following a specific exposure, underscoring the potential for targeted interventions. In addition, these clocks are valuable for assessing whether an intervention or policy change may impact long-term health outcomes and which subgroup it affects most. However, to date, few studies have fully harnessed this potential in public health practice. This area of research remains underexplored and warrants further investigation and application.

CONCLUSION

Epigenetic clocks, which estimate biological age through patterns of DNA methylation, present an exciting prospect for public health. With the ability to provide a picture of an individual’s aging process, second-generation clocks predict mortality risk better than traditional demographic variables such as age, sex, and ethnicity do.

By identifying individuals who are aging faster biologically, interventions could be targeted more effectively, potentially slowing the aging process and leading to significant improvements in population health. Ongoing research in the area of DNA methylation–based biomarkers remains an exciting frontier in preventive medicine.