Epigenetic Age Acceleration as a Modifiable Public Health Target: A Systematic Review and Meta-Analysis of Environmental, Behavioral, and Social Determinants with Development of the MEAB-Index

This systematic review and meta-analysis was conducted and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement [20]. The study protocol was registered in PROSPERO (CRD420261371271) [21] following completion of the literature search and study selection, and prior to data extraction and quantitative synthesis.

Minor deviations from the registered protocol were introduced to refine the analytical framework. These included a modest refinement of the study title to better reflect the final scope and emphasis of the work, as well as the development and inclusion of the Modifiable Epigenetic Aging Burden Index (MEAB-Index) to quantify the cumulative preventable burden of epigenetic age acceleration. All deviations were defined and documented prior to the start of quantitative data synthesis and did not affect the pre-specified eligibility criteria or primary outcomes.

The primary research question was: What is the magnitude and direction of the association between modifiable behavioral, environmental, and social determinants and epigenetic age acceleration (EAA) measured using validated DNA methylation-based clocks in adult populations?

The study had three operational objectives: (i) to systematically identify and synthesize all relevant studies examining associations between modifiable determinants and epigenetic aging; (ii) to standardize heterogeneous effect estimates into a common, interpretable metric, primarily expressed as years of EAA; and (iii) to quantify the overall impact of modifiable determinants on biological aging and derive a composite measure of preventable public health burden, defined as the Modifiable Epigenetic Aging Burden Index (MEAB-Index).

Eligibility criteria were defined a priori according to the PICOS framework: ○ Population: Adults (≥18 years) from the general population or community-based samples. ○ Exposure: Modifiable behavioral, environmental, psychosocial, or socioeconomic determinants (e.g., lifestyle factors, diet, physical activity, smoking, alcohol consumption, sleep, stress, environmental exposures, socioeconomic conditions). ○ Comparator: Different levels of exposure or reference/unexposed groups. ○ Outcome: Epigenetic age acceleration (EAA), defined as the residual from regressing epigenetic age on chronological age, measured using validated DNA methylation-based clocks (i.e., Horvath, Hannum, PhenoAge, GrimAge/GrimAge2, DunedinPoAm, DunedinPACE). ○ Study design: Observational studies (cross-sectional, cohort, case–control) and interventional studies reporting quantitative associations.

Exclusion criteria included studies conducted in pediatric populations, animal models, reviews, editorials, protocols, methodological-only studies (e.g., clock development without exposure–outcome association), and studies lacking extractable quantitative data.

A comprehensive literature search was conducted in PubMed/MEDLINE and Scopus from database inception to 7 April 2026. Europe PMC was additionally searched to enhance retrieval of open-access publications.

The search strategy was structured around three conceptual domains: (i) epigenetic aging measures, (ii) modifiable determinants, and (iii) study design and association terms. Within each domain, terms were combined using the Boolean operator OR, and the three domains were combined using AND.

The epigenetic aging domain included terms such as “epigenetic clock”, “DNA methylation age”, “epigenetic age acceleration”, “biological age”, and specific clock names (Horvath, Hannum, PhenoAge, GrimAge, DunedinPoAm, DunedinPACE). The modifiable determinants domain included lifestyle, behavioral, environmental, and socioeconomic exposures (e.g., diet, physical activity, smoking, alcohol consumption, sleep, psychosocial stress, air pollution, and socioeconomic status). The study design domain included terms such as association, regression, observational, cohort, cross-sectional, case–control, and randomized controlled trial.

The full electronic search strategy for PubMed is reported below and was adapted for the other databases:

(“epigenetic clock” [Title/Abstract] OR “DNA methylation age” [Title/Abstract] OR “epigenetic age” [Title/Abstract] OR “biological age” [Title/Abstract] OR “epigenetic age acceleration” [Title/Abstract] OR “age acceleration” [Title/Abstract] OR Horvath [Title/Abstract] OR Hannum [Title/Abstract] OR PhenoAge [Title/Abstract] OR GrimAge [Title/Abstract] OR DunedinPoAm [Title/Abstract] OR DunedinPACE [Title/Abstract]) AND (diet [Title/Abstract] OR nutrition [Title/Abstract] OR “physical activity” [Title/Abstract] OR exercise [Title/Abstract] OR smoking [Title/Abstract] OR alcohol [Title/Abstract] OR sleep [Title/Abstract] OR stress [Title/Abstract] OR “air pollution” [Title/Abstract] OR “environmental exposure*” [Title/Abstract] OR “socioeconomic status” [Title/Abstract]) AND (association [Title/Abstract] OR regression [Title/Abstract] OR observational [Title/Abstract] OR cohort [Title/Abstract] OR “cross-sectional” [Title/Abstract] OR “case–control” [Title/Abstract] OR “randomized controlled trial” [Title/Abstract]).

No language or geographical restrictions were applied, except that only articles published in English were considered eligible. Additional studies were identified through manual screening of reference lists of included studies and relevant reviews, as well as forward citation tracking via Google Scholar. No search updates were performed after 7 April 2026.

Study selection was performed in two stages: (i) screening of titles and abstracts, and (ii) full-text assessment. To enhance efficiency and reproducibility while minimizing bias, a semi-automated Python-based workflow (Python 3.14.5, Python Software Foundation, Wilmington, DE, USA, 2023) was used exclusively for deduplication, data organization, and preliminary keyword-based flagging of records. This tool served only as decision support; no records were automatically excluded and it had no influence on final inclusion or exclusion decisions.

All records were independently screened by two reviewers (S.M.A. and P.M.). Each record was classified as “include”, “exclude”, or “unclear” based on the pre-defined eligibility criteria. Disagreements were resolved by discussion and consensus, with consultation of a third reviewer when necessary. Reasons for exclusion at the title and abstract stage were recorded using pre-specified categories: (i) wrong outcome (not epigenetic age or epigenetic age acceleration), (ii) wrong exposure (not a modifiable determinant), (iii) wrong population (e.g., pediatric-only populations, animal studies), (iv) wrong study design (e.g., review, editorial, protocol), (v) methodological-only studies (e.g., clock development without exposure–outcome analysis), and (vi) other reasons.

Potentially eligible studies underwent full-text assessment by the same two reviewers working independently. Reasons for exclusion at the full-text stage were documented in detail. The study selection process is presented in a PRISMA flow diagram (Figure 1). Of the records selected for full-text assessment, most articles were successfully retrieved through institutional subscriptions and open-access sources; however, a subset of records could not be accessed in full-text form despite repeated retrieval attempts via institutional library services, publisher platforms, and open-access repositories, and therefore did not provide extractable quantitative data.

Data extraction was performed using a pre-piloted, standardized data collection form developed specifically for this review. The following information was extracted from each included study: study identifiers (first author, year, journal), study design, population characteristics (sample size, mean age, sex distribution), detailed definition and measurement of the exposure(s), type of DNA methylation-based epigenetic clock used, definition and operationalization of epigenetic age acceleration (EAA), effect estimates (including beta coefficients, mean differences, standardized mean differences, odds ratios, hazard ratios), corresponding measures of uncertainty (95% confidence intervals, standard errors, and p-values), statistical modelling approaches, and covariates adjusted for in multivariable models. Geographical classification of studies was based on the location of the study population (i.e., where data were collected), rather than on author affiliations.

To facilitate data organization and initial harmonization, a semi-automated Python-based pipeline was used. However, all extracted data were independently verified and manually checked for accuracy by the reviewers. Data extraction was conducted independently by two reviewers (S.M.A. and P.M.). Any discrepancies were resolved through discussion and, when necessary, consultation with a third reviewer. To further ensure data quality and internal consistency, an additional duplicate data verification was performed on a random subset of 20% of the included studies, with consistency checks applied across key variables.

Studies were eligible for quantitative synthesis if they reported a quantitative association between modifiable determinants and epigenetic age acceleration (EAA) and provided sufficient information to calculate effect sizes and their measures of variance.

Due to expected heterogeneity in effect metrics across studies, a pre-specified harmonization framework was applied. Effect estimates were classified into four mutually exclusive analytical pools based on their statistical format: ○ Pool A (primary analysis): Unstandardized beta coefficients or mean differences expressed in years of epigenetic age acceleration (EAA). ○ Pool B (sensitivity analysis): Standardized beta coefficients or standardized mean differences (SMD). ○ Pool C: Odds ratios (OR). ○ Pool D: Hazard ratios (HR).

For Pool A, the effect size (y_i) was taken directly as the reported beta coefficient or mean difference. When standard errors were not reported, they were derived from 95% confidence intervals using the formula:SE=CIupper−CIlower3.92

For Pools C and D, effect estimates were log-transformed (y_i = In(OR) or In(HR)), and corresponding standard errors were calculated from the reported confidence intervals.

When exposures were reported on different scales (e.g., per 1-unit increase, per standard deviation, per interquartile range, or as categorical contrasts), effect sizes were harmonized to a common scale using pre-specified conversion rules to maximize comparability. Studies that did not provide adequate data for harmonization, as well as exposure categories not meeting the minimum threshold for quantitative synthesis (k ≥ 3), were included in the qualitative synthesis but excluded from quantitative meta-analysis.

Meta-analyses were performed using a random-effects model, which assumes that true effects may vary across studies and therefore accounts for expected between-study heterogeneity in populations, exposure definitions, and analytical approaches. Between-study variance (τ²), representing the amount of true heterogeneity, was estimated using the DerSimonian–Laird method [22].

Pooled effect estimates (θ) were calculated as inverse-variance weighted averages, giving greater weight to studies with more precise estimates (i.e., smaller standard errors):whereis the weight assigned to study i, andis the within-study variance. w i = 1 v i + τ 2 w i v i

Statistical heterogeneity was assessed using Cochran’s Q test (p < 0.10 considered statistically significant) and quantified with the I² statistic, which expresses the proportion of variability due to true heterogeneity rather than chance, and with τ².

Pre-specified subgroup analyses were conducted according to exposure categories of modifiable determinants (e.g., smoking, diet, physical activity, socioeconomic status), where applicable.

Meta-regression analyses, which evaluate whether study-level characteristics explain variability in effect sizes, were performed to explore potential sources of heterogeneity, including study design, population characteristics, type of epigenetic clock, and level of covariate adjustment.

The robustness of the pooled estimates was evaluated through sensitivity analyses, including: (1) restriction to studies with full multivariable adjustment, (2) exclusion of studies with high risk of bias, and (3) removal of influential studies identified through leave-one-out analysis and Cook’s distance. Publication bias was assessed visually using funnel plots and statistically using Egger’s test (when ≥10 studies were available).

To quantify the cumulative impact of modifiable determinants on biological aging, we developed the Modifiable Epigenetic Aging Burden Index (MEAB-Index) based on pooled estimates derived from the primary meta-analysis (Pool A).

Three complementary metrics were defined: MEAB-composite: The overall burden across all exposure categories, calculated as an inverse-variance weighted average of the category-level pooled effect estimates:MEAB-composite=∑(wk⋅βk)∑wk where wk=1SEk2, and βk and SEk represent the pooled effect size and its standard error for exposure category k, respectively. Category-level estimates were combined using a fixed-effect framework. The standard error of the MEAB-composite was computed as SE(MEAB-composite)=1/∑wkMEAB-positive: Calculated using the same inverse-variance weighted approach but restricted to exposure categories that showed statistically significant positive association with epigenetic age acceleration (EAA, p < 0.05).Cumulative Preventable Burden (CPB): The unweighted sum of all statistically significant positive pooled estimates: CPB=∑k∈Sβk where S is the set of exposure categories with significant positive associations. The CPB represents the theoretical maximum reduction in EAA achievable by eliminating all identified risk factors and should be interpreted as an upper-bound estimate, given potential correlations and overlaps between exposures.

Uncertainty for all three metrics was quantified using non-parametric bootstrap resampling (10,000 iterations). Category-level pooled estimates were sampled from normal distributionsand percentile-based 95% confidence intervals were derived. Sensitivity analyses were performed using a leave-one-category-out approach to evaluate the influence of individual exposures on index stability. N β k , SE k 2

Risk of bias was assessed separately according to study design. For observational studies, the Newcastle-Ottawa Scale (NOS) was used. For randomized or interventional studies (if any), the Cochrane Risk of Bias 2 (RoB 2) tool was applied.

Two reviewers (S.M.A. and P.M.) independently evaluated the risk of bias, with disagreements resolved by discussion and consensus.

Publication bias was examined through visual inspection of funnel plots and formally tested using Egger’s regression test and Begg’s rank correlation test. When at least 10 studies were available for a given exposure, the trim-and-fill method was used to explore the potential impact of small-study effects.

The overall certainty of the body of evidence for each exposure category was rated using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework. The five domains evaluated were: risk of bias, inconsistency, indirectness, imprecision, and publication bias. The certainty of evidence was graded as high, moderate, low, or very low.

All statistical analyses were conducted using R version 4.3.2 (R Core Team, R Foundation for Statistical Computing, Vienna, Austria, 2024) [23] primarily through the metafor package [24] and Python 3.14.5 (Python Software Foundation, Wilmington, DE, USA, 2023) [25].

The complete analytical workflow, including data cleaning, harmonization, meta-analysis, MEAB-Index calculation, and sensitivity analyses, was predefined within a scripted pipeline to ensure computational reproducibility.

To further enhance transparency and reproducibility, all R and Python scripts used for data cleaning, harmonization, meta-analytic modelling, MEAB-Index computation, and sensitivity analyses are openly available in the dedicated GitHub repository: https://github.com/smaliberti/eaa-meta-analysis↗ (accessed on 28 May 2026).

All R and Python scripts are available from the corresponding author upon reasonable request.

The study selection process is presented in Figure 1, following the PRISMA 2020 guidelines.

A total of 309 records were identified through database searching (PubMed/MEDLINE: n = 159; Scopus: n = 150). After removal of 6 duplicates, 303 unique records were screened at the title and abstract level. Of these, 203 records were considered potentially eligible and advanced to full-text assessment. Full texts were successfully retrieved for 171 articles, while the remaining records lacked an accessible full-text version despite retrieval attempts. Following full-text evaluation, 57 studies were excluded due to predefined criteria. Consequently, 114 studies initially met the inclusion criteria. After removal of 2 residual duplicates identified during data extraction, 112 unique studies were retained.

All studies underwent effect size harmonization. During this phase, 29 studies were excluded from quantitative synthesis due to incompatible analytical direction (EAA used as predictor rather than outcome), non-poolable effect measures, or insufficient data for variance estimation. Ultimately, 83 studies providing 118 distinct exposure–clock associations were included in the meta-analysis.

These studies were classified into four analytical pools according to the reported effect metrics: Pool A (unstandardized beta coefficients or mean differences, n = 60 studies), Pool B (standardized beta coefficients or SMD, n = 9), Pool C (odds ratios, n = 10), and Pool D (hazard ratios, n = 4). Detailed results for each analytical pool are presented in Supplementary Tables S1–S4. The main characteristics of all included studies are summarized in Table 1, while a summary of the pooled effect sizes across the four pools is presented in Table 2.

More than half of the studies were conducted in North America (57.8%), predominantly in the United States, followed by Europe (25.3%) and Asia/Other regions (16.9%). Cross-sectional designs were the most common (50.6%), followed by prospective cohort studies (34.9%). The median sample size was 8742 participants (range 158–343,723), with a mean participant age of 56.4 years. Fully adjusted models were used in 79.5% of the studies.

Lifestyle and behavioral factors represented the largest exposure category (42.2%), followed by environmental exposures (22.9%) and socioeconomic/psychosocial determinants (19.3%). Second- and third-generation clocks were predominant, with GrimAge/GrimAge2 (31.3%) and PhenoAge (25.3%) being the most frequently used.

This systematic review and meta-analysis provides a comprehensive and quantitatively integrated synthesis of the associations between modifiable environmental, behavioral, and social determinants and epigenetic age acceleration (EAA), positioning biological aging as a measurable and potentially modifiable outcome of the exposome. By integrating 83 studies and 118 distinct exposure–clock associations across four complementary analytical pools (A–D), the present work provides robust evidence that adverse exposures are consistently associated with accelerated biological aging, whereas protective factors show directional trends toward deceleration. Collectively, these findings support the conceptualization of epigenetic aging as a dynamic, biologically embedded, yet modifiable process.

Environmental exposures emerged as a major domain influencing EAA. Multiple lines of evidence converged on the detrimental impact of air pollution and chemical mixtures. Studies focusing on ambient fine particulate matter and nitrogen dioxide [39,41,42,44] consistently documented accelerated EAA, likely mediated by oxidative stress, systemic inflammation, and direct epigenetic modifications at loci regulating inflammation and DNA repair. These associations were reinforced by research on household air pollution from solid fuel combustion [40] and complex occupational or environmental chemical mixtures [38], suggesting that the cumulative burden of diverse pollutants may exert additive or synergistic effects on the epigenome. Temperature-related exposures further highlighted the sensitivity of biological aging to climatic factors. Both short- and long-term temperature variability [37,43] were associated with increased EAA, possibly through mechanisms involving heat shock proteins, inflammatory responses, and disruption of circadian rhythms. In contrast, protective environmental features showed clear beneficial effects. Residential greenness [45] was linked to slower epigenetic aging, an effect plausibly explained by reduced exposure to pollutants, lower psychological stress, increased physical activity, and enhanced microbial diversity, all factors known to modulate inflammatory pathways.

A particularly noteworthy aspect concerns the potential protective role of mild and stable climate conditions. Regions characterized by temperate, Mediterranean-type climates with limited extreme temperature fluctuations—such as the Cilento area in southern Italy, a candidate Blue Zone—may contribute to slower epigenetic aging trajectories. This hypothesis aligns with the present findings on temperature variability as a risk factor and is supported by previous work from the present research group (Aliberti et al., 2022a, 2023, 2025a) [17,109,110], where mild climate, mineral-rich water, and low exposure to thermal extremes coexist with exceptional healthspan. The convergence between these population-level observations and the molecular evidence from the current meta-analysis suggests that stable, non-extreme climatic conditions may represent an underappreciated protective environmental factor, possibly by minimizing chronic activation of stress-response systems and preserving epigenetic stability over the life course.

Metabolic and inflammatory markers exhibited some of the strongest and most consistent associations with accelerated EAA. This domain included complementary indicators of cardiometabolic dysfunction and systemic inflammation. Robust associations were observed for the cardiometabolic index [47], metabolic syndrome severity [49], and metabolic dysfunction-associated steatotic liver disease [46]. These metabolic findings were reinforced by studies focusing on direct inflammatory burden, particularly the systemic immune-inflammation index [50] and neutrophil-related inflammatory markers [48]. Collectively, this body of evidence strongly supports the inflammaging paradigm [5,86] and the exposome-driven pathobiological framework previously outlined by the present author (Aliberti 2025b) [111]. Within this model, metabolic and inflammatory disturbances represent critical nodes where environmental, behavioral, and social exposures converge to accelerate biological aging. The findings mirror the lower burden of cardiometabolic and inflammatory dysregulation documented in Cilento and Sicilian Mountain villages, where traditional Mediterranean dietary patterns and active lifestyles appear to mitigate these pathways.

Psychosocial stress demonstrated robust and largely independent associations with greater epigenetic age acceleration. This domain encompassed a broad spectrum of stressors operating at different life stages and ecological levels. Adverse childhood experiences [57] were strongly linked to faster EAA. These long-term effects were complemented by evidence on adult psychosocial burden, including general psychosocial stress [55,56], clinical depression [53], anxiety disorders [59], work-related stress and job strain [52], and neighborhood-level social stressors and deprivation [54,58]. A key strength of this body of evidence is the persistence of associations even after adjustment for behavioral mediators. This suggests that psychosocial stress influences the epigenome through direct biological pathways, including sustained activation of the hypothalamic–pituitary–adrenal (HPA) axis, sympathetic nervous system overdrive, and pro-inflammatory signaling. The consistency across different constructs—from early-life trauma to chronic adult stress and neighborhood disadvantage—supports a life-course model of psychosocial embedding. These mechanisms align with the Brain–Energy–Microbiome–Exposome framework previously proposed by the present author (Aliberti et al., 2025c) [112], whereby chronic psychosocial adversity contributes to accelerated biological aging through interconnected neuroendocrine, inflammatory, and epigenetic pathways, ultimately influencing healthspan trajectories observed in long-lived populations such as those in the Cilento region.

Dietary and nutritional factors and physical activity generally showed protective associations. Higher diet quality, Mediterranean-type patterns, and beneficial nutrients such as polyunsaturated fatty acids [67] and vitamin intake [70] tended to be associated with slower EAA. Similarly, leisure-time physical activity [76], objectively measured activity [74], and life-course physical activity trajectories [75,77] were linked to reduced epigenetic aging. These protective effects align closely with the lifestyle characteristics observed in Cilento and other Blue Zones, where traditional Mediterranean diets and sustained physical activity linked to daily life are considered central pillars of exceptional longevity (Aliberti et al., 2022, 2024) [17,19].

Although these domains consistently showed protective trends, no statistically significant associations were observed in the pooled estimates. This lack of statistical significance likely reflects the heterogeneity of exposure assessment methods, variability in measurement instruments (e.g., self-reported vs. objective physical activity), and the limited number of available studies in some categories. Importantly, the directionality of effects remained coherent across studies, suggesting that the absence of statistical significance may be attributable to methodological variability rather than to a true lack of biological relevance.

Socioeconomic and social determinants and sleep-related exposures showed more moderate or heterogeneous associations. Neighborhood deprivation [80], socioeconomic position [79], and composite social determinants of health [81] were positively linked to EAA, while sleep quality and patterns yielded mixed results [83,84,85]. These findings further highlight the importance of structural and social factors within the broader exposome.

The 11 studies retained only for narrative synthesis provide additional valuable insights. They address heterogeneous exposures, including mixed lifestyle profiles [26,27], smoking across life stages [28,29,30], cognitive function [31], reproductive aging [35], chronic disease states [33,34], and specific prenatal or environmental toxicants [32,36]. Although not suitable for quantitative pooling, these studies enrich the overall picture and point to emerging frontiers in life-course and disease-specific epigenetic research.

Findings from secondary analytical pools (B, C, and D) reinforce and substantially extend the primary results obtained in Pool A, providing multi-metric validation of the observed associations. This convergence across different statistical frameworks is a major strength of the present synthesis, as it mitigates concerns related to scale-dependent or metric-specific biases and enhances the robustness and generalizability of the conclusions.

In Pool B (standardized effect sizes), consistent associations were observed across a wide range of exposures. Psychosocial factors such as aging anxiety [88], behavioral traits including self-control [90], and social relationships [92] showed significant standardized effects, mirroring the patterns seen in the primary analysis. Lifestyle factors, including physical activity [89] and various dietary patterns [93], also demonstrated protective standardized associations. The inclusion of evidence from randomized intervention settings in this pool is particularly noteworthy. Combined supplementation and exercise interventions [91] produced measurable improvements in epigenetic aging trajectories, offering preliminary support for causal plausibility and highlighting the potential reversibility of accelerated epigenetic aging through targeted lifestyle modifications.

Pool C (odds ratios) further confirmed an increased likelihood of accelerated epigenetic aging in relation to several modifiable exposures. Tobacco use [98,101], environmental toxicants [97], adverse dietary patterns [100,104], metabolic dysfunction [102,105], and sleep traits evaluated through genetic instruments in Mendelian randomization designs [103] were all associated with higher odds of accelerated EAA. The use of Mendelian randomization approaches in some of these studies is especially valuable, as it helps address residual confounding and strengthens causal inference for selected exposures.

Pool D (hazard ratios) extended the findings to clinically meaningful longitudinal outcomes. Associations were observed between environmental characteristics such as green and blue space availability [106], favorable dietary patterns [107], cardiorespiratory fitness [108], and lower inflammatory burden [109] with reduced hazard of accelerated aging phenotypes over time. These results are particularly relevant because they demonstrate that the exposures associated with EAA in cross-sectional and short-term studies also translate into differences in long-term biological aging trajectories, reinforcing the prognostic value of epigenetic clocks.

The overall convergence of results across all four analytical pools—unstandardized betas (Pool A), standardized effects (Pool B), odds ratios (Pool C), and hazard ratios (Pool D)—substantially reduces concerns regarding metric-specific bias and provides a high degree of triangulation. This multi-metric consistency strengthens confidence in the main findings and supports their generalizability across different statistical representations and study designs. Moreover, the inclusion of both observational and interventional evidence, as well as genetic instrumental variable approaches, adds complementary layers of support for the modifiability of epigenetic aging and the biological plausibility of the observed associations.

The molecular findings of this meta-analysis show remarkable consistency with epidemiological observations from populations characterized by exceptional longevity. The Cilento region in southern Italy, identified as a candidate Blue Zone, is characterized by a unique combination of traditional Mediterranean dietary patterns, sustained physical activity integrated into daily life, favorable environmental conditions (including mild climate and high-quality air and water), and strong intergenerational social cohesion. Previous work conducted by part of the present research group (Aliberti et al., 2023; Aliberti et al., 2025a, 2025b, 2025c, 2025d) [109,110,111,112,116] has extensively characterized this exposome and documented exceptionally high rates of healthy longevity and successful aging in this area.

The present meta-analysis provides a critical translational bridge by demonstrating at the epigenetic level that the same modifiable factors associated with exceptional longevity at the population level—low exposure to air pollution and environmental toxicants, favorable metabolic and inflammatory profiles, reduced psychosocial stress, high diet quality, regular physical activity, and supportive social environments—are indeed linked to slower epigenetic aging. This convergence between macro-level epidemiological patterns observed in Cilento and other Italian longevity hotspots (including Sicilian Mountain villages) and micro-level epigenetic evidence strengthens the plausibility of a shared mechanistic framework. It suggests that the exposome of these long-lived populations operates, at least partly, by preserving epigenetic stability and slowing the biological aging process. Such integration moves the field forward from descriptive observations of longevity to a more mechanistic understanding of how specific modifiable determinants influence healthspan at the molecular level.

A major contribution of this work is the development of the Modifiable Epigenetic Aging Burden Index (MEAB-Index). By synthesizing category-specific pooled estimates through inverse-variance weighting and bootstrap validation, the MEAB-Index moves beyond isolated associations toward a cumulative, prevention-oriented quantification of the modifiable burden on biological aging. To our knowledge, this represents the first meta-analytic framework that systematically integrates multiple exposure domains, harmonizes heterogeneous effect measures across four analytical pools, and operationalizes the cumulative impact of modifiable determinants on EAA.

The multi-pool analytical strategy, combined with rigorous sensitivity analyses, leave-one-out procedures, and assessment of publication bias, addresses several limitations common in previous reviews in the geroscience field. Furthermore, the explicit separation of studies suitable for quantitative synthesis from those retained for narrative review provides transparency regarding the current state of evidence standardization. These methodological advancements offer a reproducible template for future updates of this rapidly evolving literature and for comparative analyses across different populations and epigenetic clocks.

As expected, given the breadth of exposures, populations, and epigenetic clocks included, substantial between-study heterogeneity was observed. Meta-regression analyses identified exposure category and type of epigenetic clock as significant moderators of this variability, while study design and level of covariate adjustment showed more limited influence. Heterogeneity was particularly pronounced in environmental and metabolic domains, likely reflecting differences in exposure assessment (e.g., specific pollutants versus composite indices), timing of exposure across the life course, and population susceptibility. In contrast, psychosocial stress and socioeconomic domains showed relatively lower dispersion, suggesting more consistent underlying biological pathways.

Importantly, the direction and statistical significance of associations remained highly consistent across subgroups and sensitivity analyses. This pattern indicates heterogeneity primarily reflects differences in effect magnitude rather than conflicting evidence. Such variability is inherent to the multidimensional and context-dependent nature of biological aging and should be viewed as informative rather than problematic, highlighting the need for greater standardization in future studies while simultaneously revealing the richness of exposome-biology interactions.

The robustness of the findings was thoroughly evaluated through multiple sensitivity analyses. Restriction to fully adjusted models, exclusion of studies with high risk of bias, and leave-one-out analyses consistently confirmed the stability of the main results. Although formal tests (Egger’s and Begg’s) and visual inspection of funnel plots suggested possible asymmetry in some domains, the trim-and-fill method identified no missing studies, and the adjusted pooled estimates remained virtually unchanged. These findings collectively support the internal validity and stability of the meta-analytic conclusions.

This study has several strengths, including the large number of included studies, rigorous harmonization of heterogeneous effect measures, a pre-specified multi-pool analytical strategy, comprehensive sensitivity analyses, and the innovative construction of the MEAB-Index with bootstrap validation. The integration of narrative synthesis for studies not suitable for quantitative pooling further enriches the interpretative depth of the review.

Limitations must also be acknowledged. The predominance of observational (mostly cross-sectional) designs limits causal inference, and residual confounding or reverse causation cannot be entirely excluded. Variability in exposure measurement, timing, and epigenetic clock selection contributed to heterogeneity, although this was partially addressed through subgroup and meta-regression analyses.

The deliberate inclusion of cross-sectional studies was a methodological choice aimed at maximizing coverage of modifiable determinants, given the current scarcity of longitudinal and interventional studies with EAA as the primary outcome. While this approach provides a broad map of associations, it underscores the urgent need for more high-quality longitudinal and intervention studies in this field.

While the consistency of associations across multiple analytical approaches and exposure domains supports the robustness of the findings, caution is warranted in their interpretation. The observational nature of most included studies precludes definitive causal claims, and bidirectional relationships between exposures and biological aging cannot be excluded. For example, metabolic dysfunction may both contribute to and result from accelerated epigenetic aging.

Nevertheless, the triangulation of evidence across observational studies, Mendelian randomization analyses, and emerging interventional trials, together with the strong biological plausibility of the identified pathways, substantially strengthens the interpretation that modifiable determinants are likely to exert directional influences on epigenetic aging trajectories.