Epigenetic clocks are mathematical models that estimate biological age by measuring patterns of DNA methylation, the attachment of methyl groups to cytosine-guanine dinucleotide sites across the genome. These methylation patterns shift in predictable ways as tissues age, allowing researchers to compute an epigenetic age that can be compared with a person’s chronological age. When epigenetic age exceeds chronological age, a person is said to show epigenetic age acceleration, a state previously linked to elevated risks of cardiovascular disease, cognitive impairment, hypertension, and all-cause mortality. Because socioeconomic position is also tied to each of those outcomes, epigenetic aging has emerged as a plausible biological mechanism connecting social circumstances to long-term health. What has been missing, the authors argue, is evidence from a nationally representative cohort using a comprehensive measure of socioeconomic position rather than isolated indicators such as education or income alone.

To fill that gap, the team constructed a novel composite index of socioeconomic position grounded in the conceptual framework of Nancy Krieger and colleagues, which treats socioeconomic position as encompassing both resource-based measures and prestige-based markers of social standing. The index drew on four components: income, captured by the family poverty-to-income ratio, which compares self-reported family income with federal poverty guidelines adjusted for household size; wealth, proxied by home ownership; occupational status, classified as white collar, blue collar, or never worked based on the longest job a participant reported; and educational attainment, defined as the highest degree completed. Participants who scored favorably on all four dimensions were classified as higher socioeconomic position, those who scored unfavorably on all four as lower, and everyone else as middle. This approach deliberately required consistency across domains, distinguishing the study from prior work that examined each component in isolation.

The analytic sample comprised 2,532 adults drawn from the 1999–2000 and 2001–2002 NHANES survey waves, the periods for which DNA methylation profiling, leukocyte telomere length, and socioeconomic data were all available. Blood samples from adults aged 50 and older were processed for DNA methylation using Illumina’s Infinium MethylationEPIC BeadChip, with quality control steps including removal of outlier samples, mean imputation of missing probes, and BMIQ normalization across probe types. From these data the researchers computed thirteen different epigenetic clocks, spanning measures trained to predict chronological age, such as Horvath and Hannum; phenotypic age; mitotic cell division; mortality, via GrimAgeMort and GrimAge2Mort; telomere length, via HorvathTelo; and the pace of aging, via DunedinPoAm. Telomere length was also measured directly by quantitative PCR in adults aged 20 and older as the ratio of telomere to reference DNA.

The results showed a clear pattern. In models adjusted for chronological age, gender, race and ethnicity, and marital status, higher socioeconomic position was associated with lower epigenetic age acceleration on six of the thirteen clocks, including HannumAge, PhenoAge, GrimAgeMort, GrimAge2Mort, HorvathTelo, and DunedinPoAm. The associations were strongest for the mortality-predicting clocks: higher versus lower socioeconomic position corresponded to roughly 2.8 to 3.4 years lower epigenetic age on GrimAgeMort and GrimAge2Mort in the sociodemographically adjusted models. After additional adjustment for health behaviors and prevalent chronic conditions, including body mass index, smoking, physical activity, heavy alcohol use, hypertension, diabetes, coronary artery disease, and cancer, the associations for GrimAgeMort, GrimAge2Mort, and HorvathTelo remained statistically significant, though attenuated to approximately one year of slower aging.

The choice of clocks that survived full adjustment is telling. GrimAgeMort and GrimAge2Mort are unusual among epigenetic clocks because they were built not to predict chronological age but to predict mortality, using DNA methylation surrogates of key plasma proteins, including adrenomedullin, beta-2 microglobulin, cystatin C, growth differentiation factor 15, leptin, hemoglobin A1c, C-reactive protein, and tissue inhibitor metalloproteinase 1, alongside methylation-predicted smoking pack-years. Given that socioeconomic position is one of the most robust predictors of mortality in epidemiologic research, the authors note that strong associations with these mortality-oriented measures are unsurprising. Exploratory analyses of the clock components pointed to growth differentiation factor 15, beta-2 microglobulin, and tissue inhibitor metalloproteinase 1 as the strongest drivers of the observed differences, hinting at inflammatory and other pathways that may link social disadvantage to accelerated biological aging.

Income emerged as the single most salient component of the composite index. When the researchers mutually adjusted the four SEP components for one another, greater income was independently associated with lower epigenetic aging across GrimAgeMort, GrimAge2Mort, and HorvathTelo, while educational attainment was independently associated with the HorvathTelo clock, a DNA methylation-based estimator of telomere length. Notably, socioeconomic position was associated with the methylation-predicted telomere measure but not with directly measured telomere length by PCR, a discrepancy the authors interpret in light of evidence that the HorvathTelo clock outperforms measured telomere length at predicting mortality. The finding suggests that different molecular readouts of aging may capture different facets of the socioeconomic gradient in health.

The study also uncovered a striking gender difference. Statistical tests for interaction between socioeconomic position and gender were significant for both GrimAgeMort and GrimAge2Mort, and stratified analyses showed that the protective association was concentrated among men. Men in the higher socioeconomic position group had roughly two years lower epigenetic age on GrimAgeMort and 2.4 years lower on GrimAge2Mort, whereas the corresponding estimates in women were small and their confidence intervals included the null. The authors caution that the confidence intervals for men and women overlapped, so the difference should be treated as hypothesis-generating, but they argue that the social and biological contributors to this gender divergence deserve dedicated investigation. Analyses stratified by race and ethnicity showed a similar directional pattern, with non-Hispanic white participants showing the clearest associations, though the estimates were imprecise and require replication in larger, more diverse samples.

One of the most consequential findings concerns blood cell composition. When the researchers added the proportions of lymphocytes, monocytes, neutrophils, eosinophils, and basophils to their fully adjusted models, most confidence intervals widened to include the null, even though the magnitude and direction of the estimates remained largely stable. Because methylation patterns differ across immune cell types, shifts in the cellular makeup of blood can masquerade as epigenetic aging, and the results imply that cell proportions partially account for the socioeconomic gradient in clock readings. The authors suggest that cell-specific epigenetic clocks could help disentangle whether aging of particular immune cell subsets tracks socioeconomic circumstances. A sensitivity analysis restricted to participants without hypertension, diabetes, coronary artery disease, or cancer showed consistent directions but attenuated, statistically nonsignificant estimates, underscoring that chronic disease itself sits on the pathway between social position and molecular aging.

The authors are careful about what the study can and cannot show. As a cross-sectional analysis, it demonstrates correlation rather than causation, and reverse causation, in which poorer health limits economic attainment, cannot be excluded. The wealth component relied solely on home ownership, an admittedly incomplete proxy for a construct that properly includes assets and debts across multiple sources, and future work should replicate the findings with richer wealth measures. The NHANES sample also excludes vulnerable populations such as incarcerated and housing-insecure people, and oversampling of Asian Americans had not yet begun in these early survey waves, limiting generalizability. Even so, the study is the first in NHANES to operationalize socioeconomic position with a composite index spanning income, wealth, occupation, and education, and its nationally representative design gives the findings unusual breadth. The authors call for longitudinal studies and, provocatively, for trials testing whether income supplementation interventions can slow the epigenetic clock, a question that would directly test whether reducing poverty changes the molecular biology of aging.

Subject of Research: Associations between a composite socioeconomic position index and epigenetic aging measures in a nationally representative US cohort

Article Title: Associations between socioeconomic position index and epigenetic aging in NHANES

Article References: Associations between socioeconomic position index and epigenetic aging in NHANES. (n.d.). https://doi.org/10.1007/s11357-026-02565-5

Image Credits: AI Generated

DOI: 10.1007/s11357-026-02565-5

Keywords: epigenetic aging, socioeconomic position, DNA methylation, NHANES, epigenetic clocks, GrimAge, telomere length, health disparities, income, biological aging, GeroScience, cell composition