Writing

Does volunteering reduce epigenetic age acceleration among retired and working older adults? Results from the Health and Retirement Study

Seoyoun Kim a,b,*, Cal Halvorsen c, Claire Potter d, Jessica Faul b

a Department of Sociology, Texas State University, USA b Institute for Social Research, University of Michigan, USA c Brown School of Social Work, Washington University St. Louis, USA d Irish Clinical Academic Training (ICAT), Queen’s University Belfast, UK

A B S T R A C T

Objectives: The current study aims to explore the relationship between the frequency of volunteering and biological aging, as measured by epigenetic age acceleration. It also investigates whether this relationship differs between retired and working older adults. Understanding this connection could inform interventions promoting healthy aging and reducing age-related chronic health conditions. Method: Data were derived from the Health and Retirement Study (HRS), including pre-treatment covariates (2012), volunteer frequency and work status (2014), and five DNA methylation measures (2016) (N → 2,605). Generalized linear models were estimated to examine the relationship between volunteering and epigenetic age acceleration, stratified by retirement status. The analyses adjusted for relevant covariates and utilized energy balancing weights to account for selection into volunteering. Results: Findings show that volunteering, especially for 1–49 h per year and 200↑ hours per year, was linked to less epigenetic age acceleration, with significant effects on DNA methylation measures PhenoAge, GrimAge, and DunedinPACE clocks. Among retired individuals, moderate volunteering was significantly associated with decelerated epigenetic age acceleration, indicating greater benefits for retirees compared to working individuals. Conclusions: The study found that frequent volunteering may lead to decelerated epigenetic aging, potentially offering a public health intervention to enhance health and quality of life among older adults. Further research is needed to confirm these findings and to understand how volunteering might differentially impact retired and working individuals. Such insights could guide the development of targeted strategies to promote healthy aging and address age-related health disparities.

1. Introduction

Biological aging represents cumulative changes in biological and physiological functioning over time, including telomere lengths, cellular aging, and epigenetic modification (Belsky et al., 2015). Among the numerous biomarkers of aging, epigenetic age acceleration has garnered significant attention for its potential to predict morbidity and mortality beyond chronological age (Lu et al., 2019). Though the prevalence of age-related diseases increases in later life, chronological age is neither a practical nor accurate predictor of health and functioning. Unlike chronological age, biological aging differs considerably among older individuals; slower agers experience delayed onset of age-related dis- eases, greater cognitive and physical functioning, and better integrity of metabolic, cardiovascular, and immune systems, compared to their more rapidly aging counterparts (Fraga and Esteller, 2007; Hannum et al., 2013; Teschendorff et al., 2010). Given that epigenetic age ac- celeration is influenced by genetic, environmental, and behavioral fac- tors, understanding the predictors of epigenetic age acceleration could

serve as a crucial step in studying the environmental etiology of age-related conditions and developing public health interventions that may potentially decelerate biological aging and extend healthy lifespan.

Social participation, particularly in the form of structured or formal volunteering, has been associated with various health benefits in older adults. Volunteering is associated with various physical health outcomes including hypertension, chronic inflammation, stress regulation, func- tional ability, and cognitive function (Brown and Okun, 2014; Burr et al., 2015; S. Kim and Ferraro, 2014; Proulx et al., 2018; Sneed and Cohen, 2013). A series of meta-analyses and review papers have concluded that the benefits of volunteering for physical health are evident, and that volunteering can become a public health intervention to increase health and quality of life among older adults (Jenkinson et al., 2013; Morrow-Howell, 2010; Okun et al., 2013). Volunteer ac- tivities can provide opportunities for meaningful social engagement, which can buffer against stress, and they promote mental and physical health. In recent years, a small yet growing body of studies has begun to suggest that social factors such as social engagement, receipt of social

* Corresponding author. 601 University Dr., San Marcos, TX, 78666, USA. E-mail address: [email protected] (S. Kim).

Contents lists available at ScienceDirect

Social Science & Medicine

journal homepage: www.elsevier.com/locate/socscimed

https://doi.org/10.1016/j.socscimed.2024.117501 Received 2 August 2024; Received in revised form 21 October 2024; Accepted 11 November 2024

Social Science & Medicine 364 117501

Available online 13 November 2024 0277-9536/© 2024 Elsevier Ltd. All rights are , , including those for text and data ) ) ( AI , ) ) ( and similar technologies.

support, and experiences of parental death may affect epigenetic aging processes (Farina et al., 2024; Nakamura et al., 2023; Rentscher et al., 2023). One paper specifically focuses on volunteering frequency and epigenetic age acceleration using cross-sectional data (Nakamura et al., 2023). However, the relationship between frequent volunteer activity and epigenetic age acceleration remains underexplored, particularly in the context of working and retired older adults. Extending the extant literature, the current study aims to explore the potential link between frequent volunteering and biological aging, as measured by epigenetic age acceleration, and to investigate whether the relationship differs by retirement status. Understanding this relationship is crucial as it could inform interventions aimed at promoting healthy aging and reducing age-related health disparities among retired and working individuals in later life.

1.1. Volunteering and epigenetic age acceleration

Epigenetic aging is recognized as an important factor in under- standing the complex mechanisms underlying aging and age-related diseases. Epigenetic clocks were developed to estimate the changes in the epigenome—most commonly through DNA methylation (DNAm)— to approximate how fast or slow someone is aging biologically compared to their chronological age (Oblak et al., 2021). Understanding the connection between volunteering and epigenetic aging, measured via epigenetic clocks, holds promise for both scientific research and the promotion of healthy aging in communities. Assuming volunteering may be linked with slower epigenetic aging, the resulting public health im- plications would include new calls to volunteer not only for civic and community health, but also age-associated conditions or diseases.

The current research regarding the impact of formal volunteering on epigenetic age acceleration is limited in scope; however, this line of investigation gains credence based on the studies that have shown re- lationships between social support, regular physical activity, purpose in life and epigenetic age acceleration (Fox et al., 2023; E. Kim et al., 2023; Rentscher et al., 2023). Studies have shown that lower social support is associated with accelerated epigenetic aging in older adults (Rentscher et al., 2023). Regular physical activity, measured by step count, light intensity workout, and moderate intensity workout, is associated with decelerated epigenetic aging (Fox et al., 2023). Additionally, higher psychosocial resilience, such as having a stronger sense of purpose in life, is correlated with decelerated epigenetic age acceleration (E. Kim et al., 2023). Specifically for volunteering, frequent volunteering was associated with lower epigenetic age acceleration in PhenoAge, GrimAge, DunedinPACE, Zhang mortality, and Yang miotic, all of which predict health and longevity (Nakamura et al., 2023). Overall, while the exact relationship between formal volunteering and epigenetic age ac- celeration remains underexplored, the broader implications of social engagement and psychological well-being on biological aging are evident. The current study contributes to this literature by incorporating longitudinal data to assess how retirement status serves as an effect modifier in the relationship between volunteering and biological aging.

1.2. Retirement as a potential effect modifier

Notwithstanding the robust research findings on social engagement in general and epigenetic age acceleration, the association between volunteering and epigenetic aging may be attributable to multiple confounders such as retirement (Eibich et al., 2022; Grünwald et al., 2021). Although social engagement and productive activities are considered a key element of successful aging for all older adults (Rowe and Kahn, 1997), retiring from work may be an important variable to consider during investigations of the health effects of later-life volun- teering, as retirement opens up new opportunities for other meaningful activities, including volunteering (Eibich et al., 2022; Grünwald et al., 2021; Tang, 2016).

Indeed, volunteering has been shown to increase after retirement

(Bj!alkebring et al., 2021), with some volunteer programs consisting of nearly all retirees (Pillemer et al., 2017). In terms of the health benefits, some randomized controlled studies show that engaging in formal vol- unteering for at least 60 minutes per week leads to meaningful im- provements in psychological outcomes among retired older adults (Jongenelis et al., 2022). This is partly due to the need to replace re- tirees’ social and professional connections and roles they once gained through the workplace. For retired individuals, formal volunteering can provide structure, social interaction, and a sense of purpose, which are beneficial for health and may be linked with slower epigenetic aging. Maintaining or replacing these connections after retirement may have important biological implications, given that older adults with fewer relationships experience accelerated epigenetic aging (Rentscher et al., 2023). For working individuals, the effect of volunteering on epigenetic age might be mixed. On one hand, volunteering can be linked to more social bonds and greater social engagement, which are beneficial. However, those who are actively working might not derive additional benefits from volunteering. Further, role overload from balancing de- mands of work and volunteering could be associated with no significant changes in epigenetic age acceleration or potentially accelerated epigenetic aging.

Given the sizable shifts in lifestyle and health that retirement brings, encompassing changes in both social interactions and roles (Szinovacz and DeViney, 1999), it is crucial to study whether retirement operates as a sensitive period during which older adults reap more health benefits from the same volunteering activities. The current project hypothesizes that the relationships between formal volunteering and epigenetic clocks would be more pronounced among retired older adults.

1.3. Selection into volunteering

When investigating epigenetic age acceleration in the context of formal volunteering, an important consideration is the selection into volunteering. Volunteers come from diverse backgrounds; however, research shows that healthier and more socioeconomically advantaged older adults are more likely to volunteer (Kail and Carr, 2020; S. Kim, 2020; S. Kim and Halvorsen, 2021). These pre-existing advantages can bias results, making it appear that volunteering itself is the primary predictor of better health outcomes. Without accounting for these fac- tors, studies might overestimate the positive effects of volunteering, failing to distinguish between the benefits of the activity and the ad- vantages already possessed by the participants. As such, research that attempts to estimate the effect of volunteering on epigenetic aging must reduce the bias caused by these selection effects. Using the Health and Retirement Study, we examined the net effects of volunteering on epigenetic age among workers and retirees aged 51 and older using quasi-experimental methods to reduce the bias due to self-selection into volunteering. For the current study, we examined two first-generation clocks, 2 second-generation clocks, and one third-generation clock. The different epigenetic age measures show varying characteristics and associations with chronological age, morbidity, and mortality.

2. Methods

Sample. We used 2012, 2014, and 2016 waves of data from the Health and Retirement Study (HRS), a nationally representative and biannual panel study of adults 51 and older. HRS is sponsored by the National Institute on Aging (NIA U01AG009740) with a representative sample of approximately 20,000 community-dwelling Americans 51 and older. We included data from three points in time: pre-treatment cova- riates in 2012 (i.e., predictors of volunteering), volunteer frequency and work status in 2014, and DNA methylation data from 2016. DNA methylation assays were collected from a subsample of respondents who consented to the HRS 2016 Venous Blood Study (VBS). The detailed data description of the VBS sample is published elsewhere (Crimmins et al., 2020). A total of 4,018 individuals with DNAm data passed Quality

S. Kim et al. Social�Science�&�Medicine�364��������117501�

2�

Control (QC). The current analytic sample included participants who were 62 and older in 2012 and had valid data on the outcomes of interest in 2016 (n → 2,605). Age 62 was chosen for the current project, particularly with respect to retirement status, because it is the earliest age at which individuals can start receiving Social Security retirement benefits. Sample weights were used to adjust for sample selection into the 2016 Venous Blood Study.

Volunteering. Volunteering was measured by asking HRS partici- pants: “Have you spent any time in the past 12 months doing volunteer work for religious, educational, health-related, or other charitable or- ganizations?” If they answered affirmatively, they were then asked on a cascading scale how many hours they volunteered a year: 1–49, about 50, 51–99, about 100, 101–200, about 200, and more than 200 hours. The adjacent categories were combined which resulted in five cate- gories: non-volunteers (0), 1–49 (1), 50–99 (2), 100–199 (3), and 200↑ hours (4). Volunteering behaviors remained relatively consistent across multiple waves, which is why 2012 and 2014 waves were used to create a cumulative volunteering frequency measure. For the current analyses, we calculated this cumulative volunteering frequency by combining two volunteering variables from 2012 and 2014, ensuring a robust repre- sentation of participants’ overall volunteering activities during this period. The final variable ranged from not volunteering at either wave (0) to volunteering more than 200 h both waves (8).

Retirement. Retirement was evaluated by asking HRS respondents’ self-reported retirement status in 2014. The original categories included working full time, working part time, partly retired, retired, disabled, unemployed, and not in labor force. In the main analysis, we included everyone regardless of their labor force status. For the analyses stratified by retirement status, we compared full-time and part-time workers to full-time and part-time retirees. In a supplementary analysis, we used work hours as a more objective measure of retirement. Following the Internal Revenue Service’s (IRS) definition of full-time employment, we used 30 hours a week as a threshold for full-time employment. The re- sults were very similar to the findings from main analysis.

DNA methylation age. This study employs five commonly used epigenetic clocks- Horvath, Hannum, PhenoAge, GrimAge, and Dun- edinPACE- due to their unique strengths in capturing different di- mensions of biological aging (Faul et al., 2023; Rentscher et al., 2023). DNAm measures (epigenetic clocks) were derived from the 2016 HRS Venous Blood Study (Crimmins et al., 2020). We included 2 first-generation, 2 second-generation clocks, and 1 third-generation clock. First-generation epigenetic aging measures were trained on chronological age. Horvath’s clock was derived from DNAm measured at 353 CpG sites (Horvath, 2013), and Hannum’s measure was designed using a linear sum of methylation from 71 CpG sites (Hannum et al., 2013). Horvath and Hannum clocks are well-established for estimating chronological age and predicting age-related diseases from diverse tis- sue types. Among the second-generation clocks, PhenoAge was devel- oped using 9 blood-based phenotypic aging measures, specifically on immune functioning (Levine et al., 2018). GrimAge was trained on 7 DNAm based protein surrogates to predict all-cause mortality (Lu et al., 2019). PhenoAge and GrimAge are dubbed ‘second’ generation clocks because they extend beyond age to include morbidity and cardiovas- cular risks. Finally, a third-generation clock, Dunedin Pace of Aging (DunedinPACE) was created based on changes in 18 biomarkers of organ system integrity and health outcomes to approximate pace of aging based on data from the Dunedin cohort over 12 years and is expressed in years of epigenetic aging compared to one chronological year. Compared to earlier generation epigenetic clocks, DunedinPACE has consistently demonstrated accuracy in reflecting the biological aging process associated with specific diseases (Belsky et al., 2022). Epigenetic aging measures are available as restricted health data. More detailed information about epigenetic aging measures is available in Crimmins et al. (2020).

The five epigenetic clocks were trained on different samples who lived in different countries with various age groups represented. We

used the principal components version of the GrimAge and PhenoAge clocks for improved reliability. Further, all clocks except DunedinPACE are scaled in years and estimate epigenetic age at time of measurement, whereas DunedinPACE (i.e., whether one is aging faster or slower than average for a given age), epigenetic age was regressed on chronological age and six cell types (NK, B cells, CD 4↑, CD 8↑, CD 8↑ Naïve, Monocytes) to create intrinsic epigenetic clocks. From these regression models, we took the residual differences between the regression esti- mates and chronological age, which had positive and negative values. Positive values indicated faster epigenetic aging and negative values indicated slower epigenetic aging compared to chronological age. Conversion of DunedinPACE was not necessary because acceleration or deceleration is inherent in its unit metric.

2.1. Covariates

As potential confounders, we adjusted for relevant covariates in 2012 in the analyses, including the 2012 volunteering frequency (S. Kim et al., 2022; Nakamura et al., 2023). Demographic factors included age, gender (female, male), race/ethnicity (non-Hispanic white, non-Hispanic Black, Hispanic, other races), and marital status (part- nered, not partnered). Socioeconomic status was assessed by educa- tional attainment (less than high school, GED or high school diploma, college degree or higher), household income (inverse hyperbolic sine transformed), and household wealth (assets minus debts, inverse hy- perbolic sine transformed). Inverse hyperbolic sine (IHS) transformation is useful for normalizing highly skewed data like income and wealth, which often include zero or negative values that log transformation cannot handle (Friedline et al., 2015). Health behaviors included fre- quency of physical activity (every day, more than once a week, once a week, 3 times a month or less), smoking status (currently smoking, currently not smoking), binge drinking (more than 4 drinks per day for men or 3 drinks per day for women), and obesity (body mass index lower than 30kg/m2, body mass index 30kg/m2 or higher). Finally, health-related factors were assessed by self-reported physical health (excellent, very good, good, fair, poor), activities of daily living (having any ADL limitations, not having ADL limitation) and number of chronic conditions (heart disease, diabetes, cancer, stroke, hypertension).

2.2. Analytic plans

Our analyses consisted of three major steps. First, we generated en- ergy balancing weights using the R package WeightIt to balance the “treatment” (volunteers in 2014) and “control” (non-volunteers in 2014) groups (Greifer, 2023). Energy balancing weights aim to balance the distribution of covariates, resulting in pre-treatment covariates that have similar, and often nearly identical, means and distributions in both the treatment and control groups (Huling and Mak, 2024). Energy balancing is part of a larger family of data pre-processing methods that are often called inverse probability of treatment weighting or propensity score weighting that aim to replicate a desirable aspect of randomized controlled trials: similar treatment and control groups (Austin, 2011; Guo and Fraser, 2014). For a variety of reasons, including the relative ease of specifying the models that create the weights, a lower likelihood of creating large weights, and a higher likelihood of nearly perfectly balancing treatment groups across their means and distributions, energy balancing has been shown to be a robust method (Huling and Mak, 2024). For quality control, the means and distributions of the treatment and control groups were tested before and after weighting to confirm that, as examples, gender and education breakdowns across the treat- ment and control groups became more equal after weighting than before.

Our study leverages the longitudinal nature of the HRS. For correct temporal order, pre-treatment covariates that are associated with vol- unteering were drawn from 2012 and were included with specifications described earlier (age, gender, race/ethnicity, marital status, household

S. Kim et al. Social�Science�&�Medicine�364��������117501�

3�

income, household wealth, self-reported health); a continuous version of education (number of years); and additional variables for the number of depressive symptoms (0–8) and providing informal help (0–4). We then multiplied the energy balancing weights by the sample weights to create a single “grand” weight to use in the outcome models, which has been shown to reduce bias in the treatment effect estimates (DuGoff et al., 2014).

Second, in order to estimate age acceleration using the DNAm measures, we regressed each epigenetic age on chronological age and cell types to calculate the residuals, with the exception of DunedinPACE, which is already indicative of the rate of epigenetic aging. The calcu- lated measures represent the average pace of biological aging per year. For ease of comparison between the clocks, we standardized all of the clocks.

Third, we ran generalized linear models to examine the relationships between (a) volunteering frequency in 2014 and the five epigenetic clock measures in 2016 controlling for 2012 volunteering and cova- riates, and (b) cumulative volunteering between 2012 and 2014 and five epigenetic clocks. Both models adjusted for covariates measured in 2012 and included the grand weights. To assess effect modification by retirement status, we followed the recommendations by Knol and Van- derWeele (Knol and VanderWeele, 2012), stratifying the models by retirement status. Non-volunteers served as a reference group. We handled missing data on volunteering and covariates using multiple imputation by chained equations (m → 5) using the R package, mice. Outcome variables were not imputed since the sample respondents included only those with epigenetic data.

Several additional analyses were performed to test the robustness of the analytical model. First, we tested several specifications of volun- teering variables, including the volunteering status (yes or no) and four categories (0, ω100, 100–199, 200↑ hours per year). We also tested multiple specifications of the long-term volunteering indicator, such as

the six-category variable for volunteering initiation, cessation, low, decreased effort, increased effort, and sustained high engagement. These variables yielded very similar results and the cumulative variable was chosen for a more intuitive interpretation. Further, several additional covariates were considered such as being underweight, frequency of informal helping, retirement status, and network size. All analyses were performed in early 2024 using R, version 4.2.1. The data for the current analyses are available from the Health and Retirement Study website with restricted data user agreement. The R codes for the statistical an- alyses are available upon request.

3. Results

Our first step included the creation of energy balancing weights. As shown in Fig. 1, the pre-treatment covariates were almost perfectly balanced after weighting. The left-hand panel shows the absolute stan- dardized mean differences between the treatment (2014 volunteers) and control (2014 non-volunteers) groups by 2012 covariates. The dashed vertical line at 0.1 is a commonly used maximum for determining bal- ance (Stuart et al., 2013); however, after balancing, all pre-treatment covariates had differences near zero. The right-hand panel shows the Kolmogorov-Smirnov statistics, which are a measure of variance (Austin and Stuart, 2015), also improved after weighting. This provides evi- dence that the energy balancing weights properly balanced the treat- ment groups, increasing our confidence in our study’s results.

The analytic sample included 2,605 respondents. Table 1 shows descriptive statistics for the study variables. In 2012, respondents were 73 years old on average, and the majority of respondents were female (57%), non-Hispanic White (73%), partnered (67%), and with high school diploma (35%). The average wealth among respondents was about $422k. Typical respondents were physically active every day (49%), non-smokers (88%), non-binge drinkers (93%), not obese (63%),

Fig. 1. Love plot for energy-balance-weighted covariate distributions. Note: The left-hand panel shows the absolute standardized mean differences between the “treatment” (2014 volunteers) and “control” (2014 non-volunteers) groups for each of the 2012 variables listed, with and without energy balance weights. The right-hand panel shows the Kolmogorov-Smirvov statistics, a measure of variance, with and without energy balance weights (Austin and Stuart, 2015). Love plot created using the WeightIt program in R (Greifer, 2023). Missing data within variables were also modeled and are indicated by “ωNAε”.

S. Kim et al. Social�Science�&�Medicine�364��������117501�

4�

and without any functional limitations (87%). In 2014, most of the sample respondents were currently retired (64%) and non-volunteers (62%). In 2016, respondents’ average DNAm measures were 70 (Hor- vath), 59 (Hannum), 62 (PhenoAge), 72 (GrimAge), all of which are younger than their chronological age, and respondents were epigeneti- cally aging 1.08 years per every chronological year (DunedinPACE), data consistent with other studies on epigenetic age acceleration (Rentscher et al., 2023). Supplementary Table 1 shows the correlations between the five epigenetic clocks and chronological age.

Table 2 presents the overall associations between volunteering fre- quency and DNAm clocks, adjusting for all the relevant covariates. Volunteering is generally associated with less epigenetic age accelera- tion, particularly at the lowest (1–49 hours per year) and highest (200↑ hours per year) frequencies. 1–49 hours of volunteering is related to decelerated epigenetic aging in Hannum (β →↓0.28, p ω .05), PhenoAge (β →↓0.67, p ω .001), GrimAge (β →↓0.21, p ω .01), and DunedinPACE (β → ↓0.01, p ω .001). On the other hand, volunteering 200↑ hours per year is associated with decelerated epigenetic aging but mostly for second- and third-generation clocks such as PhenoAge (β → ↓0.97, p ω .001), GrimAge (β →↓0.34, p ω .01), and DunedinPACE (β →↓0.02, p ω .001). Importantly, nearly half (9 of 20) of these relationships were statistically significant.

These trends were largely in the same direction when examining cumulative volunteering between 2012 and 2014 (Supplemental Table 2). Low to medium level engagement (1–200 cumulative hours between 2012 and 2014) was associated with less rapid epigenetic age acceleration for Horvath and Hannum. Sustained 200↑ hours of vol- unteering between 2012 and 2014 was a significant predictor of second and third generation epigenetic clocks (PhenoAge β → ↓1.10, p ω .001, GrimAge β → ↓1.38, p ω .001, DunedinPACE β → ↓0.03, p ω .001). Again, 21 out of 40 relationships were statistically significant.

The next set of analyses examined the relationship between volun- teering frequency and epigenetic age acceleration stratified by retire- ment status. Fig. 2 shows that the effects of medium level volunteering on second and third-gen epigenetic clocks are only significant for retired volunteers. As seen for the PhenoAge, GrimAge, and DunedinPACE, 50–199 hours of volunteering (i.e., moderate amount) were associated with decelerated age acceleration for the retired (PhenoAge 100–199 hours β → ↓1.92, p ω 0.05; GrimAge 50–99 hours β → ↓0.98, p ω .05; DunedinPACE 50–99 hours β → ↓0.04, p ω 0.05). The graph shows that moderate amounts of volunteering might be more beneficial for retired older adults compared to their working counterparts.

3.1. Supplementary analyses

We conducted several supplementary analyses to ensure the robustness of the final models. First, we tested multiple specifications for measuring volunteering. For 2014, we considered both a binary mea- surement (volunteers vs. non-volunteers) and four frequency categories (0, 1–99, 100–199, 200↑ hours). Consistent with the main analyses, volunteering—regardless of frequency—was associated with deceler- ated epigenetic aging in PhenoAge, GrimAge, and DunedinPACE (Supplementary Table 3). The analysis using the four categories of volunteering revealed that volunteering for 1–99 hours and 200↑ hours significantly predicted decelerated epigenetic aging in PhenoAge, GrimAge, and DunedinPACE (Supplementary Table 4). For 2012 and 2014, we explored six different types of changes in volunteering status (initiation, cessation, consistently low effort, decreased effort, increased effort, and high sustained effort), using consistent non-volunteering as the reference category. The results indicated that any change in volun- teering status—even cessation or reduced effort in 2014—was linked to decelerated epigenetic aging in 2016. Furthermore, volunteering was associated with deceleration in second- and third-generation epigenetic clocks.

Second, we reran the main analysis, focusing only on working and retired individuals. The results similarly demonstrated that volunteering was associated with decelerated aging in second- and third-generation epigenetic clocks, particularly in PhenoAge and DunedinPACE.

4. Discussion

To our knowledge, the current study is the first to investigate the longitudinal association between volunteering and epigenetic age ac- celeration among workers and retirees 62 and older while bolstering the case that volunteering is often good for older adults. In addressing these

Table 1 Descriptive statistics for key variables (n → 2,605).

Characteristics Mean (SD)a

N (percentage)b

Age 74.87 (7.38) Gender

Male 1,108 (43%) Female 1,497 (57%)

Race Non-Hispanic White 1,976 (73%) Non-Hispanic Black 351 (13%) Hispanic 200 (8%) Other race 136 (5%)

Partnered 1,571 (60%) Education

Less than high school 455 (17%) High school 900 (35%) Some college 614 (24%) College graduate 635 (24%)

Wealth (raw, $) 527,507.7 (1181763) Physical activity

1–3 per month 613 (24%) 1 per month 261 (10%) ε1 per week 216 (8%) Every day 1,504 (58%)

Current smoker 221 (9%) Binge drinker 104 (4%) Obese (BMI ↔30 kg/m2) 826 (32%) Any ADL limitation 158 (6%) Have heart condition 859 (33%) Currently retired 2,264 (87%) Volunteering in 2014

None 1,670 (64%) 1–49 h 349 (13%) 50–99 h 241 (9%) 100–199 h 185 (7%) 200↑ hours 154 (6%)

DNAm measures in 2016 Horvath 69.75 (8.6) Hannum 58.84 (7.9) PhenoAge 61.58 (8.9) GrimAge 72.15 (7.2) DunedinPACE 1.08 (0.1)

a Mean and SD are presented for continuous variables. b Number of individuals (N) and percentages are presented for cate-

gorical variables.

Table 2 Generalized linear model predicting five epigenetic measures.

Volunteering Frequency

1–49 50–99 100–199 200↑

Horvath ↓0.15 ↓0.34 0.13 ↓0.05 Hannum ↓0.28* ↓0.32 0.19 ↓0.01 PhenoAge ↓0.67*** ↓0.99*** 0.04 ↓0.97*** GrimAge ↓0.21** ↓0.15 ↓0.22* ↓0.34** DunedinPACE ↓0.01*** ↓0.01 0.01 ↓0.02***

p ω .05, p ω .01, p ω .001. Note: The analysis accounts for all the relevant covariates (age, gender, race/ ethnicity, marital status, education, income, wealth, physical activity, smoking status, binge drinking, obesity, self-reported physical health, ADL limitation, chronic conditions), 2012 volunteering, and the grand weights (energy balancing weights * survey weights).

S. Kim et al. Social�Science�&�Medicine�364��������117501�

5�

topics, we capitalized on a large, nationally representative dataset of older adults that includes a rich set of biological data. The results suggest that volunteering, overall, leads to reduced epigenetic age acceleration (measured via DNA methylation) among older adults. We suggest this may be due to at least two broad factors. First, volunteering itself can increase physical activity, perhaps due to walking to or at the volun- teering assignment and less sedentary behavior at home; more physical activity, in turn, may be associated with less rapid age acceleration (Parisi et al., 2015; Tan et al., 2009). Second, volunteering may provide a sense of purpose in life through intentional efforts to aid causes and issues of interest while buffering the loss of important roles (e.g., spousal, parental, labor) (Greenfield and Marks, 2004). More purpose, in turn, is associated with biological aging (E. Kim et al., 2023). Other mechanisms like increased social interactions, lower stress, cognitive engagement, and knowledge acquisition may also play significant roles (Gonzales et al., 2019; Han et al., 2020; Tan et al., 2006). These in- teractions also promote mental stimulation and emotional support, which can mitigate aging at the cellular level.

The findings also reveal distinct effects of contemporaneous (i.e., frequency-based) and sustained volunteering on epigenetic aging. 1–49 h then 200↑ hours of volunteering are beneficial, particularly for second and third generational clocks, aligning with insights from Rentscher et al. (2023). In contrast, any level of cumulative volunteering is broadly beneficial for epigenetic aging, indicating that longer-term engagement might have a more profound influence on these epigenetic clocks. Thus, while both moderate and sustained volunteering benefit epigenetic aging, the continuous commitment appears to play a role in supporting biological resilience by fostering social re-engagement and stability.

Interestingly, the benefits of volunteering appear to be influenced by both the amount of time dedicated and the individual’s employment status. Specifically, at 200↑ hours of volunteering, the health benefits are significant for both retirees and working individuals, suggesting that higher engagement in volunteering can promote well-being regardless of one’s work status. However, at 50–199 hours of volunteering, the benefits are more pronounced for retirees compared to their working counterparts. For working older adults, the demands of professional life may limit the additional gains from volunteering, as they already have structured social interactions and a sense of purpose through their ca- reers (Webster et al., 2021). In contrast, retiring could reduce social and professional relationships, the absence of which has been linked to

accelerated epigenetic aging (Rentscher et al., 2023). As such, the epigenetic benefits of volunteering could benefit retirees the most through the strengthening or reemergence of these social bonds. Our stratified results by retirement found that moderate levels of volun- teering (50–99 hours per year) are particularly beneficial for retired individuals in terms of decelerated epigenetic age acceleration. This association remains significant even after adjusting for age, relevant health covariates, and selection into volunteering, indicating a robust relationship between volunteering and slower biological aging.

One possible explanation for this is that retirement marks a signifi- cant life transition that often brings about notable social and relational changes. There tends to be a decrease in work-related connections and daily social interactions and a potential increase in family and community-based ties (Kauppi et al., 2021; R!ozer et al., 2020; Wang et al., 2011). This shift can have both positive and negative implications. On one hand, stronger family bonds and more time spent with close network partners can enhance support and well-being (Grünwald et al., 2021). On the other hand, the loss of work can reduce the diversity of social networks, potentially limiting opportunities for new experiences and perspectives (Wang et al., 2011; Webster et al., 2021). Thus, engaging in volunteer work or community activities may provide structured social interactions, a sense of purpose, and opportunities to form new relationships, contributing to slower epigenetic age acceler- ation. Additionally, the consistent association of moderate levels of volunteering with decelerated aging across second and third-generation DNAm clocks further supports the notion that sustained engagement in meaningful activities is beneficial for biological health. Perhaps main- taining and cultivating social relationships and activities pre-retirement has important biological implications post-retirement, given that older adults with fewer relationships generally experience accelerated epige- netic aging (Rentscher et al., 2023). Future studies must address the mechanisms through which different levels of volunteer work influence epigenetic aging process that are relevant to all-cause mortality, im- mune functioning, and organ integrity particularly in the context of retirement.

5. Limitations

The study findings should be interpreted with caution due to the following study limitations. First, DNAm measures as candidate markers

Fig. 2. Volunteering and PhenoAge, GrimAge, and DunedinPACE, stratified by retirement status. Note: The analysis accounts for all the relevant covariates (age, gender, race/ethnicity, marital status, education, income, wealth, physical activity, smoking status, binge drinking, obesity, self-reported physical health, ADL limitation, chronic conditions) and includes the grand weight (energy balancing weights * sur- vey weights).

S. Kim et al. Social�Science�&�Medicine�364��������117501�

6�

for biological aging and harbingers of morbidity and mortality are trained on different biological markers and provide information about different underlying epigenetic changes (Crimmins et al., 2020; Hannum et al., 2013; Horvath, 2013). For example, there are some CpG sites that are captured by PhenoAge that might not be represented in GrimAge. Thus, the mixed findings should be treated as complementary, rather than contradicting results. Second, the volunteering measures do not capture the modality, type, commitment, and longevity longer than two years. Although we created a cumulative volunteering frequency mea- sure using data from 2012 to 2014, it would be more advantageous to assess long-term volunteering across a broader time frame (e.g., 10, 20, or 30 years). Cumulative measures of volunteering that capture decades of engagement could be helpful in examining the epigenetic effects of lifelong volunteering. Further, while our use of energy balancing weights ensures that volunteers and non-volunteers have nearly iden- tical means and distributions across several variables (Huling and Mak, 2024), this quasi-experimental method only controls for pre-specified variables and not, theoretically, all variables that a randomized controlled trial would control for.

6. Conclusions

This study found that in US older adults, even low levels of volun- teering is associated with less rapid epigenetic age acceleration. High levels of volunteering and sustained volunteering are significantly associated with second and third generation clocks, trained on morbidity and biomarkers of aging. Further, the relationships between moderate levels of volunteering and epigenetic age acceleration were stronger for retirees compared to their working counterparts. These findings extend the current literature on volunteering and health and might provide policy recommendations of later life volunteerism in the context of biological aging.

CRediT authorship contribution statement

Seoyoun Kim: Writing – review & editing, Writing – original draft, Visualization, Software, Methodology, Formal analysis, Conceptualiza- tion. Cal Halvorsen: Writing – review & editing, Writing – original draft, Methodology, Conceptualization. Claire Potter: Writing – review & editing, Writing – original draft, Conceptualization. Jessica Faul: Writing – review & editing, Project administration, Funding acquisition.

Ethical statement

This study utilizes data from the Health and Retirement Study (HRS), a publicly available dataset managed by the University of Michigan. The HRS data is made available to researchers under conditions that ensure the confidentiality and privacy of the participants. As this dataset is publicly accessible and fully anonymized, this study is exempt from institutional ethics board approval under the guidelines for research using existing, de-identified data. The analysis and use of this data comply with the terms of service set forth by the HRS. No new data collection involving human subjects was conducted as part of this research.

Declaration of competing interest

The authors declare no conflict of interest.

Acknowledgement

This study was supported by the National Heart, Lung, and Blood Institute (NHLBI, 1R01 HL171806), National Institutes of Health-funded training from the National Institute on Aging (R25-AG053227), and the National Center for Complementary and Integrated Health (R25- AT010664). Claire Potter is supported by the Wellcome Trust/Health

Research Board Ireland’s Irish Clinical Academic Training (ICAT) Fellowship (203930/B/16/Z).

Appendix ASupplementary data

Supplementary data to this article can be found online at https://doi. org/10.1016/j.socscimed.2024.117501.

Data availability

Both social and epigenetic datasets are publicly available through the Health and Retirement Study website (https://hrs.isr.umich. edu/data-products).

References

Austin, P.C., 2011. An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behav. Res. 46, 399–424.

Austin, P.C., Stuart, E.A., 2015. Moving towards best practice when using inverse probability of treatment weighting (IPTW) using the propensity score to estimate causal treatment effects in observational studies. Stat. Med. 34, 3661–3679.

Belsky, D.W., Caspi, A., Corcoran, D.L., Sugden, K., Poulton, R., Arseneault, L., et al., 2022. DunedinPACE, a DNA methylation biomarker of the pace of aging. Elife 11, e73420.

Belsky, D.W., Caspi, A., Houts, R., Cohen, H.J., Corcoran, D.L., Danese, A., et al., 2015. Quantification of Biological Aging in Young Adults, vol. 112. Proceedings of the National Academy of Sciences, pp. E4104–E4110.

Bj!alkebring, P., Henning, G., V!astfj!all, D., Dickert, S., Brehmer, Y., Buratti, S., et al., 2021. Helping out or helping yourself? Volunteering and life satisfaction across the retirement transition. Psychol. Aging 36, 119–130.

Brown, S.L., Okun, M.A., 2014. Using the Caregiver System Model to Explain the Resilience-Related Benefits Older Adults Derive from Volunteering. The Resilience Handbook: Approaches to Stress and Trauma. Routledge/Taylor & Francis Group, pp. 169–182.

Burr, J.A., Han, S.H., Tavares, J.L., 2015. Volunteering and cardiovascular disease risk: does helping others get “under the skin?”. Gerontol. 56, 937–947.

Crimmins, E., Kim, J., Fisher, J., Faul, J., 2020. HRS Epigenetic Clocks. Survey Research Center. University of Michigan.

DuGoff, E.H., Schuler, M., Stuart, E.A., 2014. Generalizing observational study results: applying propensity score methods to complex surveys. Health Serv. Res. 49, 284–303.

Eibich, P., Lorenti, A., Mosca, I., 2022. Does retirement affect voluntary work provision? Evidence from Europe and the U.S. Lab. Econ. 76, 102185.

Farina, M.P., Klopack, E.T., Umberson, D., Crimmins, E.M., 2024. The embodiment of parental death in early life through accelerated epigenetic aging: implications for understanding how parental death before 18 shapes age-related health risk among older adults. SSM Popul. Health 26, 101648.

Faul, J.D., Kim, J.K., Levine, M.E., Thyagarajan, B., Weir, D.R., Crimmins, E.M., 2023. Epigenetic-based Age Acceleration in a Representative Sample of Older Americans: Associations with Aging-Related Morbidity and Mortality, vol. 120. Proceedings of the National Academy of Sciences, e2215840120.

Fox, F.A.U., Liu, D., Breteler, M.M.B., Aziz, N.A., 2023. Physical activity is associated with slower epigenetic ageing—findings from the Rhineland study. Aging Cell 22, e13828.

Fraga, M.F., Esteller, M., 2007. Epigenetics and aging: the targets and the marks. Trends Genet. 23, 413–418.

Friedline, T., Masa, R.D., Chowa, G.A.N., 2015. Transforming wealth: using the inverse hyperbolic sine (IHS) and splines to predict youth’s math achievement. Soc. Sci. Res. 49, 264–287.

Gonzales, E., Shen, H.W., Perry, T.E., Wang, Y., 2019. Intersections of home, health, and social engagement in old age: formal volunteering as a protective factor to health after relocation. Res. Aging 41, 31–53.

Greenfield, E.A., Marks, N.F., 2004. Formal volunteering as a protective factor for older adults’ psychological well-being. J. Gerontol. B Psychol. Sci. Soc. Sci. 59, S258–S264.

Greifer, N., 2023. WeightIt: weighting for covariate balance in observational studies. R package version 1.0.0.

Grünwald, O., Damman, M., Henkens, K., 2021. The differential impact of retirement on informal caregiving, volunteering, and grandparenting: results of a 3-year panel study. J. Gerontol.: Ser. B 76, 607–619.

Guo, S., Fraser, M.W., 2014. Propensity Score Analysis: Statistical Methods and Applications. SAGE publications.

Han, S.H., Roberts, J.S., Mutchler, J.E., Burr, J.A., 2020. Volunteering, polygenic risk for Alzheimer’s disease, and cognitive functioning among older adults. Soc. Sci. Med. 253, 112970.

Hannum, G., Guinney, J., Zhao, L., Zhang, L., Hughes, G., Sadda, S., et al., 2013. Genome-wide methylation profiles reveal quantitative views of human aging rates. Mol. Cell 49, 359–367.

Horvath, S., 2013. DNA methylation age of human tissues and cell types. Genome Biol. 14, 3156.

S. Kim et al. Social�Science�&�Medicine�364��������117501�

7�

Huling, J.D., Mak, S., 2024. Energy balancing of covariate distributions. J. Causal Inference 12.

Jenkinson, C.E., Dickens, A.P., Jones, K., Thompson-Coon, J., Taylor, R.S., Rogers, M., et al., 2013. Is volunteering a public health intervention? A systematic review and meta-analysis of the health and survival of volunteers. BMC Publ. Health 13, 1–10.

Jongenelis, M.I., Jackson, B., Newton, R.U., Pettigrew, S., 2022. Longitudinal associations between formal volunteering and well-being among retired older people: follow-up results from a randomized controlled trial. Aging Ment. Health 26, 368–375.

Kail, B.L., Carr, D.C., 2020. More than selection effects: volunteering is associated with benefits in cognitive functioning. J. Gerontol.: Ser. B 75, 1741–1746.

Kauppi, M., Virtanen, M., Pentti, J., Aalto, V., Kivim!aki, M., Vahtera, J., et al., 2021. Social network ties before and after retirement: a cohort study. Eur. J. Ageing 18, 503–512.

Kim, E., Nakamura, J.S., Strecher, V.J., Cole, S.W., 2023. Reduced epigenetic age in older adults with high sense of purpose in life. J. Gerontol. Ser. A: Biol. Sci. Med. Sci. 78, 1092–1099.

Kim, S., 2020. Human, social and cultural predictors of productive activities in later life. Ageing Soc. 40, 328–347.

Kim, S., Ferraro, K.F., 2014. Do productive activities reduce inflammation in later life? Multiple roles, frequency of activities, and C-reactive protein. Gerontol. 54, 830–839.

Kim, S., Halvorsen, C.J., 2021. Volunteering as an equalizer: a quasi-experimental study using propensity score analysis. Am. J. Prev. Med. 61, 628–635.

Kim, S., Halvorsen, C.J., Han, S., 2022. Volunteering and Changes in Cardiovascular Biomarkers: Longitudinal Evidence from the Health and Retirement Study.

Knol, M.J., VanderWeele, T.J., 2012. Recommendations for presenting analyses of effect modification and interaction. Int. J. Epidemiol. 41, 514–520.

Levine, M.E., Lu, A.T., Quach, A., Chen, B.H., Assimes, T.L., Bandinelli, S., et al., 2018. An epigenetic biomarker of aging for lifespan and healthspan. Aging (Albany NY) 10, 573–591.

Lu, A.T., Quach, A., Wilson, J.G., Reiner, A.P., Aviv, A., Raj, K., et al., 2019. DNA methylation GrimAge strongly predicts lifespan and healthspan. Aging (Albany NY) 11, 303–327.

Morrow-Howell, N., 2010. Volunteering in later life: research frontiers. J. Gerontol.: Ser. B 65B, 461–469.

Nakamura, J.S., Kwok, C., Huang, A., Strecher, V.J., Kim, E., Cole, S.W., 2023. Reduced epigenetic age in older adults who volunteer. Psychoneuroendocrinology 148, 106000.

Oblak, L., van der Zaag, J., Higgins-Chen, A.T., Levine, M.E., Boks, M.P., 2021. A systematic review of biological, social and environmental factors associated with epigenetic clock acceleration. Ageing Res. Rev. 69, 101348.

Okun, M.A., Yeung, E.W., Brown, S., 2013. Volunteering by older adults and risk of mortality: a meta-analysis. Psychol. Aging 28, 564–577.

Parisi, J.M., Kuo, J., Rebok, G.W., Xue, Q.L., Fried, L.P., Gruenewald, T.L., et al., 2015. Increases in lifestyle activities as a result of experience Corps® participation. J. Urban Health 92, 55–66.

Pillemer, K., Wells, N.M., Meador, R.H., Schultz, L., Henderson Jr., C.R., Cope, M.T., 2017. Engaging older adults in environmental volunteerism: the retirees in service to the environment program. Gerontol. 57, 367–375.

Proulx, C.M., Curl, A.L., Ermer, A.E., 2018. Longitudinal Associations between Formal Volunteering and Cognitive Functioning, vol. 73. Journals of Gerontology Series B- Psychological Sciences and Social Sciences, pp. 522–531.

Rentscher, K.E., Klopack, E.T., Crimmins, E.M., Seeman, T.E., Cole, S.W., Carroll, J.E., 2023. Social relationships and epigenetic aging in older adulthood: results from the Health and Retirement Study. Brain Behav. Immun. 114, 349–359.

Rowe, J.W., Kahn, R.L., 1997. Successful aging. Gerontol. 37, 433–440. R!ozer, J.J., Hofstra, B., Brashears, M.E., Volker, B., 2020. Does unemployment lead to

isolation? The consequences of unemployment for social networks. Soc. Network. 63, 100–111.

Sneed, R.S., Cohen, S., 2013. A prospective study of volunteerism and hypertension risk in older adults. Psychol. Aging 28, 578–586.

Stuart, E.A., Lee, B.K., Leacy, F.P., 2013. Prognostic score–based balance measures can be a useful diagnostic for propensity score methods in comparative effectiveness research. J. Clin. Epidemiol. 66, S84–S90.e81.

Szinovacz, M.E., DeViney, S., 1999. The retiree identity: gender and race differences. J. Gerontol.: Ser. B 54B, S207–S218.

Tan, E.J., Rebok, G.W., Yu, Q., Frangakis, C.E., Carlson, M.C., Wang, T., et al., 2009. The long-term relationship between high-intensity volunteering and physical activity in older African American women. J. Gerontol.: Ser. B 64B, 304–311.

Tan, E.J., Xue, Q.L., Li, T., Carlson, M.C., Fried, L.P., 2006. Volunteering: a physical activity intervention for older adults–The Experience Corps program in Baltimore. J. Urban Health 83, 954–969.

Tang, F., 2016. Retirement patterns and their relationship to volunteering. Nonprofit Voluntary Sect. Q. 45, 910–930.

Teschendorff, A.E., Menon, U., Gentry-Maharaj, A., Ramus, S.J., Weisenberger, D.J., Shen, H., et al., 2010. Age-dependent DNA methylation of genes that are suppressed in stem cells is a hallmark of cancer. Genome Res. 20, 440–446.

Wang, M., Henkens, K., van Solinge, H., 2011. Retirement adjustment: a review of theoretical and empirical advancements. Am. Psychol. 66, 204–213.

Webster, N.J., Ajrouch, K.J., Antonucci, T.C., 2021. Volunteering and health: the role of social network change. Soc. Sci. Med. 285, 114274.

S. Kim et al. Social�Science�&�Medicine�364��������117501�

8�