Childhood socioeconomic status and changes in biological aging across mid-to-later life

Abstract

Lower childhood socioeconomic status (SES) has been associated with markers of advanced biological aging. Much of this work, however, is cross-sectional or focuses on single markers of biological aging. The current study examined whether childhood SES is associated with change in a multisystem composite of blood-based biomarkers of aging across a 14-year period from mid-to-later life. Participants in the longitudinal Adult Health and Behavior (AHAB) project (N=674, 45.4% female, mean age at Wave 1=45.5 years) retrospectively reported at Wave 1 parental education and occupation, which were combined to calculate childhood SES using the Hollingshead Index. They also provided blood at Wave 1 and Wave 2, which was assayed for eight markers of biological aging and combined into a single composite, informed by the work of Justice and colleagues (Justice et al., 2018): insulin, IGF-1, GDF15, NT-proBNP, cystatin C, IL-6, TNF-α, and CRP. Regression models tested the association between childhood SES and residualized change in the biological aging composite, adjusting for Wave 1 age, sex, and time between waves. Lower childhood SES was associated with a larger increase in the biological aging composite across a 14-year period (β= −.13, p=.001), and in sensitivity analyses, this association remained when further controlling for adult SES (β= −.11, p=.006) and childhood trauma (β= −.11, p=.008). These findings suggest that relative disadvantage in childhood may accelerate the rate of biological aging in mid- to later life, providing a possible pathway to increased risk for poorer late life health.

Introduction

Physical health tracks a socioeconomic gradient that is influenced not only by circumstances in adulthood, but also by those in childhood when trajectories of pathophysiology are thought to become established. Converging evidence from epidemiology, health psychology, neuroscience, geroscience, epigenetics, and fetal programming points to childhood as a period when experiences are programmed into the biology of the developing organism to shape lifelong trajectories of physical health (Shonkoff, Boyce, & McEwen, 2009). In this regard, relative socioeconomic disadvantage in childhood has been associated with an earlier onset of age-related chronic health conditions, including cardiovascular, metabolic, digestive, and respiratory diseases and some cancers, as well as all-cause and cause-specific mortality (Cohen, Janicki-Deverts, Chen, & Matthews, 2010; Galobardes, Lynch, & Smith, 2008). Life course models suggest that childhood and adult socioeconomic status (SES) contribute independently to trajectories of aging and lifelong health (Finkel, Nilsen, Sindi, & Kareholt, 2024). Furthermore, observed relationships between childhood exposures and adult health are largely independent of access to healthcare (Cohen et al., 2010; Galobardes et al., 2008; Miller & Chen, 2013). Thus, the lifelong health impact of socioeconomic disadvantage in childhood is a pressing health concern, particularly in light of the over one billion children worldwide and 11.5 million in the United Sates (16% of children under 18 years) who live in poverty (www.unicef.org; www.census.gov).

Pathways linking childhood exposures to lifelong health remain unclear. Theoretically, it has been proposed that early life experiences during sensitive biological periods program a phenotype that is contextually adaptive in the short-term but shapes a longer-term trajectory of accelerated aging pathophysiology and health risk (Ellis & Del Giudice, 2019; Miller & Chen, 2013; Perez-Stable & Webb Hooper, 2023). This conforms to a broader life-history framework, where uncertain or adverse early environmental conditions are thought to be associated with ‘faster’ life-history strategies that favor early maturation and reproductive fitness over long-term health (D. W. Belsky et al., 2017; Ellis & Del Giudice, 2019). Within this framework, socioeconomic disadvantage in childhood is conceptualized to program allocation of metabolic energy to maximize reproductive success by accelerating the pace of aging, at the expense of health and longevity in later life (D. W. Belsky et al., 2017; Ellis & Del Giudice, 2019). Thus, socially-patterned exposures in childhood may plausibly influence the rate of biological aging independently of chronological age and adult SES to shape lifelong trajectories of health and disparities in population health (Epel, 2020; N. L. Jones et al., 2019; Nielsen, Marsland, Hamlat, & Epel, 2024; Shapiro, 2017).

In the latter conceptual context, empirical attention has focused on identifying biomarkers that quantify aging and are impacted by childhood exposures, contributing to accelerated physiological decline and age-related health outcomes (Crimmins, 2020; Moffitt, 2020). Inter-related cellular and molecular mechanisms are proposed to characterize trajectories of physiological aging (Lopez-Otin, Blasco, Partridge, Serrano, & Kroemer, 2013). These mechanisms include cellular (leukocyte telomere attrition, cellular senescence), genomic (epigenetic changes and genomic alterations), metabolic, and organ system (immune, cardiovascular, etc.) aging (Franceschi et al., 2018; Lopez-Otin et al., 2013). To date, studies that examine the association of childhood SES with markers of aging have typically focused on one of these biological “hallmarks” of aging. For example, lower childhood SES has been associated with increased circulating and stimulated levels of inflammatory mediators and an up-regulation of inflammatory gene expression (Milaniak & Jaffee, 2019; Miller, Chen, & Cole, 2009), older epigenetic age, measured by genome-wide DNA methylation clocks (Austin et al., 2018; Brown et al., 2024; Hughes et al., 2018; McCrory et al., 2022), shorter telomere length (Mitchell et al., 2014), and biomarkers of poorer cardiometabolic health (Miller, Chen, & Parker, 2011). Much of this work, however, has been cross-sectional, thus limiting the interpretation of findings due to cohort and period effects, with differential childhood exposures across generations. An exception to this is prospective evidence from the Dunedin Study, a representative cohort of over 1000 individuals born in Dunedin, New Zealand in 1972–73 and closely followed since birth (D. W. Belsky et al., 2017; Moffitt, Belsky, Danese, Poulton, & Caspi, 2017). Findings from this cohort show that lower childhood SES predicts faster pace of aging as measured by combined measures of changes in biomarkers of multiple organ systems and by a DNA methylation-based biomarker of aging (DunedinPACE) across a 19-year period from age 26 – 45 years (D. W. Belsky et al., 2017; D. W. Belsky et al., 2022; Elliott et al., 2021). The current study aims to extend this work to an older cohort and assess whether childhood SES is associated with the multiyear change in a multisystem composite of blood-based biomarkers of aging from mid-to-later life.

Multisystem phenotypes of aging are proposed to reflect diverse biological processes across organ systems, providing a more comprehensive measure of aging than single system measures. However, it remains unclear whether there is a single metric that characterizes aging across systems and marks trajectories of health risk within individuals. Indeed, evidence suggests that in early adulthood metrics of biological aging are often unrelated to one another. For example, in the Dunedin study, telomere length was not associated with epigenetic age (D. W. Belsky et al., 2018). This led Belsky and colleagues to conclude that different metrics may not assess the same aspects of the aging process. However, it is also possible that markers of aging become more homogeneous in later life, when multisystem aging is prevalent and can be more reliably detected. It is also likely that composite multisystem measures provide a more reliable estimate of biological aging than single measures (Moqri, Herzog, Poganik, et al., 2023). In the current study, biological aging is quantified using a composite of changes in biomarkers that (a) index the function of cardiovascular, immune, renal, and metabolic systems in the second half of the lifespan and (b) account for their covariance.

The biomarkers of aging selected for the current study were informed by the work of Justice and colleagues (Justice et al., 2018) who conducted a comprehensive literature review and systematically evaluated 258 candidate biomarkers of aging to identify a subset of blood-based biomarkers that reliably assess biological age and predict healthspan. Each biomarker was evaluated by an expert panel based on evidence for its measurement reliability and feasibility, relevance to aging, ability to predict all-cause mortality and clinical and functional outcomes, and responsiveness to intervention. This process led to the identification of nine blood-based biomarkers reflecting multiple physiological systems that were recommended for use in aging research. The panel includes markers of systemic inflammation (interleukin (IL)-6, tumor necrosis factor-α (TNF-α)-receptor I or II, and C-reactive protein(CRP)), mitochondrial dysfunction (growth differentiation factor 15 (GDF15)), nutrient signaling (fasting insulin, insulin-like growth factor 1 (IGF-1)), and kidney (cystatin C), cardiovascular (N-terminal B-type Natriuretic peptides (NT-proBNP)), and metabolic (hemoglobin A1c) aging (Justice et al., 2018).

The current study examined the association of retrospectively assessed childhood SES with multiyear prospective changes in a composite of blood biomarkers that included eight of the nine multisystem markers of biological aging identified by Justice et al (Justice et al., 2018). We extend prior work showing an association between lower childhood SES and faster pace of biological aging prior to age 45 years in the Dunedin cohort (D. W. Belsky et al., 2017; D. W. Belsky et al., 2022; Elliott et al., 2021) to examine change in a multisystem composite of biological aging among a community sample of adults across a 14-year period starting at a mean age of 45.5 years. We hypothesized that childhood SES would predict biological aging as assessed by this composite, with individuals who report lower childhood SES showing accelerated aging (larger increases in the composite) compared with those who report higher childhood SES. We also examined whether any observed associations of childhood SES with biological aging were independent of adult SES. Additionally, in light of prior evidence that exposure to childhood trauma is associated with acceleration of biological aging (Colich et al., 2020), we explored the contribution of recalled childhood trauma. Finally, two theories have been proposed to explain how patterns of health inequalities may change with age. The cumulative disadvantage theory suggests that the adverse health effects of low childhood SES will strengthen over time, such that lower- versus higher-SES individuals experience increasing divergence in health as disadvantages accumulate over the life course (Cohen et al., 2010; Dupre, 2007; Willson, Shuey, & Elder, 2007). Conversely, the age-as-leveler hypothesis suggests that inequalities in health weaken with age due to selective mortality or age-related health deterioration that reduces the influence of socioeconomic differences (Crimmins, Kim, Seeman, 2009; House, Lantz, Herd, 2005). We evaluated support for these hypotheses by examining in primary analyses whether childhood SES predicted change in biological aging over time (cumulative disadvantage), and testing in exploratory analyses whether this association varied by baseline age (age-as-leveler).

Methods

Participants

Data for this study were derived from the Adult Health and Behavior (AHAB) project, a registry of sociodemographic, behavioral, and biological correlates of biological aging and cardiometabolic risk (Manuck, Phillips, Gianaros, Flory, & Muldoon, 2010; Marsland, McCaffery, Muldoon, & Manuck, 2010). A total of 1,773 adults (30–54 years) were recruited via mass-mail solicitation from communities in southwestern Pennsylvania (principally Allegheny County) to participate in AHAB during two phases at baseline/Wave 1 (AHAB-I: 2001–2005; N=1,291 and AHAB-II: 2008–2011; N=482). At enrollment, study exclusions included a reported history of atherosclerotic cardiovascular disease, chronic kidney or liver disease, cancer treatment in the preceding year, major neurological disorders, psychotic illness, systolic/diastolic blood pressure ≥180/110 mmHg (AHAB-I) or ≥160/100 mmHg (AHAB-II), alcohol consumption > 21 drinks/week (AHAB-I) or > 5 drinks 3–4 times/week (AHAB-II), pregnancy, and use of insulin, psychotropic, antiarrhythmic, or prescription weight-loss medications (Manuck et al., 2010). Further exclusions specific to AHAB-II included all other diabetic medications, antihypertensives and lipid-lowering drugs, shift-work schedules and use of fish-oil supplements, and AHAB-II participants were all employed ≥25 hr/week. Participants in both AHAB-I and AHAB-II were recruited to take part in a second wave (W2) of data collection that took place between 2018 and 2024 (12–22 years following initial participation; W1). A total of n=706 (40–71 years) participated in W2. The current analytic sample included AHAB participants with data for all covariates and blood samples available to assess biological markers of aging at both waves (n=674; see Supplemental Figure S1 for CONSORT diagram depicting how the final analytical sample was derived). All participants provided informed consent, and all procedures were approved by the University of Pittsburgh Institutional Review Board.

Procedure

Data collected at the first study visit (demographic information, medical history and medication use, and a fasting blood draw) of both AHAB waves were used in the current analyses. Participants were instructed to fast overnight and abstain from exercise for 12 hr, alcohol for 24 hr, and nicotine for 1 hr prior to this morning lab visit. At the visit, a nurse completed a medical history and medication use interview, determined body mass index (BMI; kg/m2), and drew a 40 ml blood sample. At W1 of AHAB-II and W2 of both phases of the study, the blood draw was rescheduled if participants reported symptoms of acute illness or infection, or having received a vaccination or taken antibiotics in the prior two weeks. Rescheduling was not conducted at W1 of AHAB-I; however, we performed sensitivity analyses excluding participants who were taking antibiotics or cold medicines at the time of the W1 blood draw and/or at the W2 blood draw who were identified after an audit to confirm W2 procedure adherence.

Measures

Childhood Socioeconomic Status

Participants at W1 retrospectively reported childhood SES, which was estimated from parental education and parental occupational grade recalled when the participant was 5, 10, and 15 years old using the two-factor Hollingshead Index (Freedman & Hollingshead, 1957). As reported in our prior work (Natale et al., 2023; Sweitzer et al., 2013), level of parental education was assessed on a seven-item scale from eighth grade or less (score of 1) to completion of graduate or professional training (score of 7). Parental occupation was coded using the nine Hollingshead grades, ranging from manual labor or unskilled service work (score of 1) to large-business proprietorship and high-level professions (score of 9). For both education and occupation, if reported for two parents, the highest score reported was entered into the Hollingshead algorithm to derive a weighted index score (3*education score + 5*occupation score). The Hollingshead index was calculated at ages 5, 10, and 15 years and the median score across assessments was used to reflect the representative family SES across this period of childhood.

Adult Socioeconomic Status

We were unable to estimate adult SES using the Hollingshead algorithm for all participants due to missing spousal occupational data. Instead, we employed two conventional indicators of SES as assessed at W1: years of education and net annual household income. Household income in AHAB-I was scored over seven bracketed ranges, extending from <$15,000 to ≥$80,000. Although AHAB-II participants reported over a wider span of income brackets, the highest cutoff was set at ≥$80,000 to harmonize data across the two cohorts. The midpoint dollar value for each bracket was then calculated (e.g., $7,500 for the <$15,000 bracket, $20,000 for the $15,000–24,999 bracket), and for the highest tier of income scaled to 25% above $80,000 (i.e., $100,000). These values were then adjusted to 2022 dollars to account for inflation across years of enrollment.

Markers of Biological Aging

Blood samples were processed and serum stored in an ultra-cold freezer until analysis. To derive an index of biological aging, a panel of eight blood-based biomarkers was assessed following the recommendations of a multidisciplinary Biomarker Working group (Justice et al., 2018). This group considered 258 candidate aging biomarkers and recommended a short-list of nine blood-based biomarkers (insulin, IGF-1, GDF15, NT-proBNP, cystatin C, IL-6, TNF-α, CRP, and hemoglobin A1c) for use in studies of aging. In the current study, we measured all the recommended markers except hemoglobin A1c at both study waves.

Simple Plex assays using the antibody-based Ella^™ system (ProteinSimple, Biotechne) were utilized to assess serum levels of Cystatin C, GDF15, and TNF-α. Insulin concentration was measured with a radioimmunoassay with iodine 125 (Code-a-Count; Diagnostic Products, Inc, Los Angeles, Calif) by the Heinz Nutrition Laboratory, University of Pittsburgh Graduate School of Public Health. NT-proBNP, IGF-1, and IL-6 were measured in serum using high-sensitivity quantitative sandwich enzyme immunoassay kits (NT-proBNP: Abcam, Waltham, MA, ab263877; IGF-1: R&D Systems, Minneapolis, MN, DG100B; IL-6: R&D Systems, Minneapolis, MN, HS600C) following the manufacturer’s directions. Serum CRP levels were measured using a particle-enhanced immunonephelometric assay with the BNII nephelometer from Dade Behring (Newark, DE). All samples were processed in duplicate, and intra- and inter-plate coefficients of variation (CVs) were < 10%. All W1 and W2 analytes except IL-6 and CRP were assayed together at the time of W2. To examine possible degradation of stored samples over time, an IL-6 kit was run at W2 including samples that were analyzed at W1. Results showed a high level of concordance with those obtained at the time of W1 (ICC = 0.96).

All biomarkers except IGF-1 were transformed prior to analyses to correct for non-normal distribution (insulin, NT-proBNP, GDF15, IL-6, and CRP were log transformed; cystatin C was square root transformed; and TNF-α was inverse transformed). Generally, individual biomarkers were correlated with each other in expected ways, ranging from small- to moderate-size correlations (|r|=.01 - .53); for example, inflammatory markers were positively correlated with each other and with insulin, GDF15, cystatin C, and negatively correlated with IGF-1 (see Supplemental Figure S2 and Table S1). As recommended by Justice et al (2018) a composite measure of biological aging at W1 and W2 was computed as the mean of the z-scored distributions. W2 biomarker values were standardized using W1 means and standard deviations to preserve change over time. On average, participants had data for 7.41 out of 8 biomarkers at W1 (range 1–8) and 7.89 out of 8 biomarkers at W2 (3–8). Given that the majority of participants had biomarker data for 6 or more biomarkers (89.02% at W1 and 98.66 % at W2), we created W1 and W2 biological aging composites across any available biomarker data. The reliability of this composite was adequate (W1 α =.46 and W2 α =.55).

Covariates

A priori standard covariates included participant age at W1, self-reported sex at birth (coded as female = 0, male = 1), and time between W1 and W2. Race was not included as a covariate because doing so risks conditioning on a social process that precedes and shapes socioeconomic conditions from early life onward (Shapiro, 2017), which could obscure and attenuate the association between childhood SES and biological aging (Kaufman et al., 1997; VanderWeele & Robinson, 2014).

Childhood Trauma Questionnaire

In sensitivity analyses, we included total scores from the Childhood Trauma Questionnaire (CTQ; (D.P. Bernstein & Fink, 1998; D.P. Bernstein, Fink, Handelsman, & Foote, 1994; D. P. Bernstein et al., 1994) assessed at W2 as an additional covariate. This 28-item measure assesses recalled experiences of emotional, physical, and sexual abuse and emotional and physical neglect prior to age 18 years. Responses were summed to create a total score ranging from 25 to 125, with higher scores indicating more childhood trauma. Cronbach’s α in current sample = .92. A subset of participants completed the CTQ (n=647), so the sample size for this sensitivity analysis is slightly smaller than the analytic sample.

Data Analysis

Statistical analyses were performed in R (Version 4.4.1 (2024-06-14 ucrt)). First, descriptive statistics and bivariate correlations among study variables were examined. Next, a series of regression models tested 1) the main effect of childhood SES on residualized change in biological aging from baseline to follow-up, and 2) the independent effect of childhood SES, controlling for adult SES, on residualized change in biological aging. To obtain residualized biological aging change scores, the W2 composite was regressed on the W1 composite and the resulting residuals were extracted per person and used as the dependent variable. Because individual aging biomarkers may be of particular interest to readers, results of primary regression analyses predicting residualized change in each biomarker are presented in supplemental material. All models controlled for W1 age, sex, and time between W1 and W2. In sensitivity analyses, we further added CTQ scores to the model to explore whether the association of childhood SES with changes in biological aging was independent of childhood trauma. In a second set of sensitivity analyses, we excluded participants (n=157) who were taking antibiotics or cold medicines at the time of W1 and/or W2 blood draw to exclude any possible influences on inflammatory markers. In subgroup analyses, we tested associations in participants identifying as White (n=587) because childhood SES is partly confounded by race/ethnicity. Last, in exploratory analyses we tested whether baseline age moderated the association between childhood SES and residualized change in biological aging. Variance inflation factors (VIF) for all models were < 1.3 and therefore multicollinearity was not a concern. Results in tables are reported in unstandardized (b) and standardized (β) units; corresponding 95% confidence intervals (CI) and p-values are also reported.

Results

Descriptive Statistics and Bivariate Correlations

Descriptive characteristics of the sample at both waves are shown in Table 1. Participants in the analytic sample were 45.40% female (n = 306), with mean ages of 45.5 years at W1 and 59.6 years at W2. On average, the sample was well educated with 16.4 (range 10–24) years of schooling and had a mean family income of $95,009 (range $10,230-$165,279 adjusted for inflation to year 2022) at W1. Childhood SES spanned the range of possible Hollingshead scores (8–66), with a mean value of 43.2. As expected, the biological aging composite significantly increased on average from W1 to W2 (t(df)= −11.92 (673), p<.001). Figure 1 displays standardized changes in the eight biomarkers of the biological aging composite. As expected, GDF15, Cystatin-C, NT-proBNP, IL-6, and CRP significantly increased from W1 to W2 and IGF-1 significantly decreased (ps <.001). Insulin significantly decreased from W1 to W2 (p<.001) and TNF-α (inverse transformed) significantly increased (p=.002), but for practical purposes, values remained similar at W1 and W2. Participants included in the analytic sample (i.e., those who returned for W2) differed from those who did not return for W2 (see Supplemental Table S2): participants included in the analytic sample were older and had more years of education, higher income, higher childhood SES, and lower biological aging at W1.

Table 1.

Descriptive statistics for Wave 1 and Wave 2 study variables

	Wave 1 (N=674)	Wave 2 (N=674)	P-Value
Age (Y)	45.49 (6.55)	59.64 (7.32)	<0.001
Sex at Birth (%F)	306 (45.40%)	306 (45.40%)
Race (% non-White § )	87 (12.91%)	87 (12.91%)
Education (Y)	16.44 (2.71)	16.72 (2.93)	<0.001
Missing	0	2
Family Income (2022 USD ¥ )	95009.08 (44043.50)	84988.04 (32795.21)	<0.001
Missing	0	6
Time to Follow-up (Days)	--	5176.18 (1083.97)
BMI (kg/m 2 )	26.69 (5.34)	28.25 (5.96)	<0.001
Missing	1	5
Childhood SES	43.21 (13.82)	--
CTQ Total	--	35.59 (11.97)
Missing	--	27
Residualized Change	--	0.00 (0.41)
Bio. Age Composite	−0.05 (0.42)	0.15 (0.52)	<0.001
Insulin (ln(μIU/ml))	2.40 (0.44)	1.88 (0.84)	<0.001
Missing	10	19
GDF15 (ln(pg/ml))	6.39 (0.28)	6.72 (0.35)	<0.001
Missing	72	14
Cystatin-C (sqrt(mg/L))	0.85 (0.15)	0.88 (0.16)	<0.001
Missing	86	22
NT-proBNP (ln(pg/ml))	3.44 (0.96)	3.89 (1.08)	<0.001
Missing	65	9
TNF-α (inv(pg/ml))	0.118 (0.04)	0.124 (0.03)	0.00246
Missing	78	5
IGF-1 (ng/ml)	85.96 (26.84)	69.74 (23.04)	<0.001
Missing	66	17
IL-6 (ln(pg/ml))	0.20 (0.81)	0.53 (0.67)	<0.001
Missing	34	13
CRP (ln(μg/ml))	−0.10 (1.11)	0.32 (1.08)	<0.001
Missing	26	40

Note. P-value indicates differences between variables at Wave 1 vs. Wave 2. Biomarker values have been transformed as indicated (sqrt=square root, ln=log, inv=inverse).

^§Participants self-identified as 87% White, 11% Black, <1% Asian, <1% bi-/multi-racial, and <1% other; participants identifying with racial groups other than White were aggregated for descriptive purposes.

^¥Family income is in 2022 dollars to adjust for inflation. W1 to W2 decline is likely due to income not keeping up with inflation; unadjusted family income did not significantly decline from W1 to W2.

BMI = body mass index; CRP = C-reactive protein; CTQ = Childhood Trauma Questionnaire; GDF15 = growth differentiating factor 15; IL-6 = interleukin-6; IGF-1= insulin-like growth factor 1; NT-proBNP = N-terminal B-type Natriuretic peptides; SES = socioeconomic status; TNF-α = tumor necrosis factor-alpha.

Figure 1.

Note. Figure displays standardized changes over a 14-year period from mid-to-later adulthood in the eight biomarkers (seven of eight transformed) that comprise the biological aging composite.

For descriptive purposes, Supplemental Table S3 presents participant characteristics by childhood SES tertile. Participants in the lowest childhood SES tertile were more likely to be male, identify with racial groups other than White, report fewer years of education, lower family income, and have higher biological aging composites at W1 and W2 than participants in the highest childhood SES tertile. Pearson bivariate correlations are presented in Supplemental Table S1. Lower childhood SES (Hollingshead Index) was associated with a greater residualized increase in the W1-to-W2 biological aging composite (r= −.17, p<.001), older age at W1 (r= −.15, p=.017), and more recalled childhood trauma on the CTQ (r= −.11, p=.007). Higher residualized change in the biological aging composite was significantly associated with fewer years of W1 education (r= −.10, p=.008) and with older W1 age (r=.20, p<.001), suggesting that biological aging accelerates with advancing age. Residualized change in the biological aging composite was not significantly associated with W1 income (r= −.04, p=.29) or with childhood trauma (r=.02, p=.63). Childhood SES was positively associated with adult SES (W1 income: r=.33, p<.001; W1 years of education: r=.16, p<.001).

Main Effect of Childhood SES on Changes in Biological Aging

Lower childhood SES was associated with a larger residualized increase in the biological aging composite – or put another way, higher childhood SES was associated with deceleration in the biological aging composite –beyond the effects of age, sex at birth, and time to follow-up (β= −.13, p=.001). Figure 2 displays this association; full model results are shown in Table 2. In secondary analyses that tested individual aging biomarkers (see Supplemental Tables S4–11), lower childhood SES was associated with larger residualized increases in IL-6 (p<.001) and TNF-α (p=.011) and with smaller residualized decreases in insulin (p=.044), but no other biomarkers.

Figure 2.

Note. Figure displays association between childhood SES (Hollingshead Index) and residualized change in biomarker composite over a 14-year period from mid to later adulthood.

Table 2.

Main effect of childhood SES on residualized change in biomarker composite

Predictors	b	β	CI	standardized CI	p
(Intercept)	−0.59	0.00	−0.89 – −0.28	−0.07 – 0.07	<0.001
W1 Age	0.01	0.18	0.01 – 0.02	0.11 – 0.25	<0.001
Sex	−0.05	−0.06	−0.11 – 0.01	−0.14 – 0.01	0.085
Time to Follow-up	0.00	0.16	0.00 – 0.00	0.09 – 0.23	<0.001
Childhood SES	−0.00	−0.13	−0.01 – −0.00	−0.20 – −0.05	0.001
Observations	674
R2 / R2 adjusted	0.090 / 0.085

Note. Unstandardized (b) and standardized (β) estimates are displayed with corresponding 95% confidence intervals (CI). SES = socioeconomic status; W1= Wave 1.

Independent Effect of Childhood SES on Changes in Biological Aging

We next examined the effect of childhood SES on residualized changes in biological aging controlling for adult SES, assessed by years of education and family income at W1. Lower childhood SES was similarly associated with a larger residualized increase in the biological aging composite – or put another way, higher childhood SES was associated with deceleration in the biological aging composite (β= −.11, p=.006), independently of age, sex at birth, time to follow-up, and adult SES (see Table 3). Neither adulthood years of education nor family income were independently associated with changes in biological age (β= −.03, p=.53 and β= −.04, p=.33, respectively). In secondary analyses that tested individual aging biomarkers (see Supplemental Tables S12–S19), lower childhood SES was independently associated with larger residualized increases in IL-6 (p=.009) and TNF-α (p=.017), but no other biomarkers.

Table 3.

Independent effect of childhood SES on residualized change in biomarker composite controlling for adult SES

Predictors	b	β	CI	standardized CI	p
(Intercept)	−0.51	0.00	−0.87 – −0.15	−0.07 – 0.07	0.006
W1 Age	0.01	0.18	0.01 – 0.02	0.11 – 0.26	<0.001
Sex	−0.06	−0.07	−0.12 – 0.01	−0.14 – 0.01	0.070
Time to Follow-up	0.00	0.16	0.00 – 0.00	0.08 – 0.23	<0.001
Childhood SES	−0.00	−0.11	−0.01 – −0.00	−0.19 – −0.03	0.006
Education	−0.00	−0.03	−0.02 – 0.01	−0.11 – 0.05	0.527
Family Income	−0.00	−0.04	−0.00 – 0.00	−0.12 – 0.04	0.327
Observations	674
R2 / R2 adjusted	0.093 / 0.084

Note. Unstandardized (b) and standardized (β) estimates are displayed with corresponding 95% confidence intervals (CI). SES = socioeconomic status; W1= Wave 1.

Sensitivity, Subgroup, and Exploratory Analyses

In a sensitivity analysis, we further adjusted for childhood trauma as assessed by the CTQ in the model. The observed association of childhood SES with increases in the biological aging composite remained the same (β= −.11, p=.008; see Supplemental Table S20), with no independent association of childhood trauma with biological aging (β= 0.02, p=.65).

We also conducted a conservative set of sensitivity analyses excluding participants who were taking antibiotics or cold medicines at the time of W1 or W2 blood draw. When excluding these participants, the main effect of lower childhood SES on increases in residualized change in biological aging remained (β=−.10, p=.019). The independent effect of childhood SES on change in biological aging when further controlling for adult SES (education and income) was no longer statistically significant (β=−.08, p=.103); however, the standardized beta for this association is contained within the original confidence interval of the larger sample (standardized CI = −0.19 to −0.03). Therefore, the lack of statistical significance for this association is likely due to loss of power.

Because childhood SES is partly confounded by race/ethnicity, we conducted subgroup analyses limited to participants identifying as White. All results remained the same: Lower childhood SES was associated with a larger residualized increase in the biological aging composite, independent of age, sex at birth, and time to follow-up (β= −.12, p=.003), and independent of adult SES (β= −.12, p=.004) and childhood trauma (β= −.13, p=.004). Owing to inadequate statistical power (i.e., 13% identified with racial groups other than White: 11% Black, <1% Asian, <1% bi-/multi-racial, <1% other), parallel analyses were not conducted on data of participants who identified with racial-ethnic groups other than White.

Last, in exploratory analyses, we tested whether the strength of the association between childhood SES and residualized change in biological aging differed by baseline age (range: 30 – 54 years). Baseline age did not significantly moderate these associations in models that tested childhood SES alone, in combination with adult SES, or with childhood trauma (ps > .11).

Discussion

The present longitudinal study provides novel evidence for an association between lower childhood SES and mid-to-later life change in a multisystem composite of blood-based biomarkers selected to provide a reliable index of biological aging that predicts health span (Justice et al., 2018). Among a relatively healthy sample of 674 volunteers, findings show an association between lower recalled childhood SES and a larger increase in this biological aging composite across a 14-year period from mean ages of 46 to 60 years from baseline to follow-up. This association was independent of age, sex at birth, time to follow-up, adult SES as indexed by education and income, and recalled childhood trauma. These findings suggest that growing up in more socioeconomically disadvantaged circumstances may accelerate the rate of biological aging in mid-to later life, a trajectory likely to increase risk for poorer late-life health and forecast mortality.

To date, most studies examining the association of childhood SES with markers of biological aging in adulthood have been cross-sectional or assessed single markers of aging and at a single point in time. Findings support an association of childhood socioeconomic disadvantage with cellular metrics of accelerated biological aging, including shorter telomere length (Rentscher, Carroll, & Mitchell, 2020), older epigenetic age (Austin et al., 2018; Brown et al., 2024; Hughes et al., 2018; McCrory et al., 2022), and faster pace of aging based on a DNA methylation measure trained to track declines in organ system function (Raffington et al., 2021). More closely aligned with the current findings, lower childhood SES has also been associated with higher blood-based biomarkers of aging in adulthood, including markers of inflammation (Milaniak & Jaffee, 2019) and cardiovascular and metabolic risk (Tamayo, Christian, & Rathmann, 2010), and combined measures of allostatic load (Misiak et al., 2022). The current study extends this work by examining change across mid-later life in a composite of blood-based markers that captures multiple physiological processes, including markers of inflammation and mitochondrial health, nutrient signaling, kidney aging, and cardiovascular health (Justice et al., 2018). Such composite multisystem biomarkers have been demonstrated to show a stronger correlation with other features of aging than single system biomarkers of aging (D. W. Belsky et al., 2017; Moqri, Herzog, Poganik, et al., 2023). The current findings extend results from the Dunedin cohort, which show an association of lower childhood SES with faster pace of aging as measured by a multisystem composite across a 19-year period from age 26 – 45 years (D. W. Belsky et al., 2017; Elliott et al., 2021). Here, we show the continued association of lower childhood SES with accelerated change in a multisystem composite over a 14-year period from mean ages 46 to 60 years. Findings are also consistent with a longitudinal study of over 3,000 individuals aged 6–18 years who were followed for over 30 years with repeated measurement of risk factors including measures of HDL cholesterol, fasting triglycerides, fasting plasma glucose, fasting serum insulin (Kivimaki et al., 2018). By early adulthood, individuals who were raised in lower SES neighborhoods had increased blood lipid levels, decreased insulin sensitivity, and increased compensatory insulin secretion. By midlife, the low SES group were more likely to be obese, have hypertension, a fatty liver, and diabetes compared to those who consistently lived in higher SES circumstances (Kivimaki et al., 2018). In sum, the current findings add to growing evidence that lower SES circumstances in childhood are associated with an accelerated trajectory of biological aging, possibly contributing to the increased risk for morbidity and mortality observed among adults who experience socioeconomic disadvantage in childhood (Cohen et al., 2010; Galobardes et al., 2008; Shonkoff et al., 2009). Our findings support the potential public health benefit of policies and programs that target socioeconomic disadvantage among families.

Conceptually, SES itself is a complex, multilevel, and multi-dimensional construct that is taken to reflect the stratification of individuals within social, employment, and economic structures and systems. This stratification is thought to shape access to resources, social prestige, assets, and opportunities across the lifespan. We note that SES intersects with other social constructs, including gender, race, and ethnicity. Socioeconomic disadvantage can accrue over the lifespan, or it may have pronounced (later-life) health effects when experienced at certain times of earlier life development. Although not all findings are consistent, growing evidence suggests that childhood SES plays a unique role in accelerated biological aging and the related etiology of comorbid diseases of aging that is not fully explained by adult SES (Cohen et al., 2010). The current findings support childhood as a sensitive period for establishing lifelong health, with no evidence that indices of SES in adulthood (education or income) attenuated or contributed independently to the association of childhood SES with biomarkers of aging (Austin et al., 2018; D. W. Belsky et al., 2017; Hughes et al., 2018). Fewer years of education were related to our composite measure of biological age at both waves (r= −.21), possibly reflecting continuity of socioeconomic circumstances across the lifespan (r= .33 between our measures of childhood SES and educational attainment in adulthood). However, we observed no additional effect of adult SES on changes in our biological aging composite beyond childhood SES. Consistent with theoretical models that propose that early life circumstances program biological systems to adapt to future environmental conditions and to favor reproductive fitness over longer term health (e.g., the developmental origins of health and disease (Barker, 2007); the biological embedding of adversity (Miller & Chen, 2013); and life history models (D. W. Belsky et al., 2017; J. Belsky, 2019; Ellis & Del Giudice, 2019)), the current pattern of results supports accelerated biological aging as a result of unique environmental exposures in childhood.

In secondary analyses, we examined the association of childhood SES with the eight separate aging biomarkers included in the composite. Most biomarkers showed expected age-related changes in mean levels over time, with significant increases in GDF15, Cystatin-C, NT-proBNP, IL-6, CRP, and decreases in IGF-1 and insulin levels (Bryhni, Arnesen, & Jenssen, 2010; Justice et al., 2018). Lower childhood SES was associated with significant increases in circulating levels of two inflammatory markers (IL-6 and TNF-α), but not CRP or the other biomarkers of aging. Although mean aging across the eight biomarkers captures aging of multiple physiological systems and likely provides a more reliable estimate of the unitary construct thought to reflect biological age (Moqri, Herzog, Poganik, et al., 2023), the current results suggest that inflammatory cytokines may be particularly impacted by childhood circumstances. These results are consistent with predictions based on the biological embedding model (Miller et al., 2011) and the stress sensitization model (Daskalakis, Bagot, Parker, Vinkers, & de Kloet, 2013), which suggest that early life exposures during sensitive periods of development program an immune phenotype that may drive systemic inflammation across the lifespan. Meta-analytic evidence supports this possibility and shows an inverse association of childhood SES with markers of inflammation in adulthood (Milaniak & Jaffee, 2019). It is unclear why associations of childhood SES with changes in IL-6 and TNF-α in the current study did not extend to CRP. This is inconsistent with prior findings, including our own cross-sectional findings from the first wave of the current cohort, showing an inverse relationship of childhood SES with CRP that contributed to the association of childhood SES with adult cardiometabolic risk (Natale et al., 2023). The stronger effects for IL-6 and TNF-α than CRP in the longitudinal results reported here may relate to pathways of risk associated with childhood SES. In contrast to CRP, an acute phase protein produced by the liver in response to immune activation, IL-6 and TNF-α are produced by a wide variety of cell subtypes, including adipocytes and myocytes, and levels are known to be impacted by multiple contemporaneous factors, including adiposity, exercise, psychological stress and other environmental circumstances. For example, psychological stress is associated with reliable increases in circulating levels of IL-6, but not CRP (Marsland, Walsh, Lockwood, & John-Henderson, 2017). Further work is warranted to consider the contribution of psychological stress, exercise and other health behaviors to the association of childhood SES with age-related changes in inflammatory cytokines.

The relationship of individual biomarkers with one another at each wave was relatively low and quite variable (r’s .01–.53), raising question about whether there is a single metric that characterizes aging across systems. This is consistent with findings of others showing that metrics of biological aging are often unrelated to one another (D. W. Belsky et al., 2018) and does not support the possibility that markers of aging coalesce in later life and reflect a single aging process. Although evidence supports gradual age-related declines across multiple physiological systems, rate of aging has been shown to vary by cell subtypes and organ systems within individuals (Moqri, Herzog, Justice, et al., 2023). Evidence shows that the individual markers included in the composite used in the current study predict declining function and increased risk for morbidity within the physiological systems they represent (Justice et al., 2018). It remains to be determined whether in combination they provide a marker of multisystem pathophysiology that improves prediction of comorbid health conditions that accompany aging (Crimmins et al., 2021). Given the association of childhood SES with increased risk for a wide range of health outcomes that are driven by shared pathophysiological mechanisms, this possibility warrants further investigation.

Contrary to expectations based on existing literature (Colich, Rosen, Williams, & McLaughlin, 2020), we did not observe a significant association of recalled childhood trauma with our composite marker of biological aging. This finding is inconsistent with results of a recent meta-analysis that showed an association of early life threat, but not deprivation or SES with accelerated cellular aging as measured by telomere length and DNA methylation clocks; however, it should be noted that the authors recommend caution in interpreting their findings due to the small number of studies included in the review and the significant potential for publication bias (Colich et al., 2020). Furthermore, telomere length and epigenetic measures of aging do not correlate well with clinical biomarker composites, raising question about whether the different measures reflect a single aging process (D. W. Belsky et al., 2018). In contrast to telomere length and epigenetic markers, which assess biological aging of blood cells, the measure used in the current study reflects aging of multiple systems throughout the body and thus may provide a different system-wide metric of aging.

We evaluated support for the cumulative disadvantage and age-as-leveler hypotheses. The results from our primary analysis – that lower childhood SES predicts a larger increase in the biological aging composite over a 14-year follow-up – is consistent with the cumulative disadvantage hypothesis (Cohen et al., 2010; Dupre, 2007). In exploratory analyses, baseline age did not significantly moderate this association; that is, the divergence in biological aging for lower- vs. higher-SES was similar across baseline age in midlife. Thus, there was little evidence of the age-as-leveler hypothesis. An important limitation is that the restricted baseline age range of our sample (30–54 years), along with potential health selection at older ages, likely constrained our ability to observe age-leveling processes. Prior research suggests that socioeconomic disparities in health widen through midlife and only begin to attenuate at older ages (e.g., Chen & Zhao, 2023). Future research drawing on samples with broader age ranges will be important to further clarify how cumulative disadvantage and age-related leveling may operate at different stages of the life course.

Multiple psychosocial and behavioral risk factors that are associated with childhood socioeconomic disadvantage likely contribute to poorer lifelong health and may conceivably mediate the association between childhood SES and biological aging. For example, exposure to stressful life events, lack of social support, and unhealthier lifestyle and coping behaviors contribute to the association of childhood SES with health outcomes (Brown et al., 2024; McEwen & Gianaros, 2010; Misiak et al., 2022) and are associated with markers of aging, such as heightened systemic inflammation (e.g. (Hostinar, Lachman, Mroczek, Seeman, & Miller, 2015). Regardless of the specific pathways, which are likely complex and variable, prospective evidence, including recent results from our own work (Cundiff, Boylan, Pardini, & Matthews, 2017) (E. J. Jones et al., 2025), shows that upward socioeconomic mobility across childhood associates with better health outcomes in adulthood. Future work should examine the possibility that upward social mobility across childhood impacts trajectories of biological aging and thus improves health span. If this is the case, programs and policies that provide families with socioeconomic resources may be of long-term benefit to public health.

Strengths and Limitations:

There are many strengths of the current study; to our knowledge it is one of the first longitudinal studies to examine the association between recalled childhood SES and change in a composite measure of biological aging across a 14-year period of mid-to-later life among a relatively large (N = 673) sample of community volunteers. The composite measure used reflects biological processes across organ systems, providing a more comprehensive measure than single system measures. The biomarkers that were included were selected from a pool of 258 blood-based candidate measures identified in a comprehensive literature review (Justice et al., 2018). A multidisciplinary group of experts applied stringent selection criteria to identify a short list of nine biomarkers that were reliable, relevant to aging, and predicted clinical and functional outcomes and all-cause mortality (Justice et al., 2018). We assessed eight of these nine candidate markers from plasma samples collected on two occasions separated by an ~14-year period. This addressed limitations of prior work, which is largely cross-sectional or based on measures assessed at one timepoint.

Several limitations should be considered when interpreting the current findings. First, the study is limited by the retrospective assessment of childhood SES when participants were 5–15 years old. This assessment lacks developmental specificity and may be subject to recall bias (Hardt & Rutter, 2004; Reuben et al., 2016). Evidence suggests that retrospective and prospective measures of childhood circumstances are at best only moderately correlated within person (Baldwin, Reuben, Newbury, & Danese, 2019). It is hoped that participants were better able to recall their parents’ occupation and educational levels than other aspects of childhood SES, such as parental income, which may be unknown or subject to greater bias. Second, SES as recalled in childhood may serve as a proxy that accesses a range of early life experiences, including not only resource deprivation, but also neighborhood disadvantage, poor health behaviors, family stressors, and exposure to chronic social and psychological stressors, all of which can impact rate of aging. A limitation to the present study is that we did not assess the childhood home environment (e.g., housing quality, crowding, toxic exposures), neighborhood environment (e.g., air and water quality), or school environment (e.g., educational quality, nutritional quality, built environment, etc.). Moreover, our measure of childhood SES is specific to this sample, born between 1947 and 1979 (~80% born in 1950s-1960s) and who spent their childhoods primarily during the post-war economic expansion in the United States. This time period was marked by sustained economic growth, occupational stability, and expanding educational opportunities, compared to later decades with less structured labor markets and more income volatility (French, 1997). Thus, childhood SES in this sample, indexed by parental education and occupation, may reflect stable employment, income security, and enhanced access to institutional resources. Despite these cohort-specific features, our findings align with results from other cohorts (e.g., Dunedin Study born in 1972–1973 in New Zealand), suggesting that early life SES may exert long-lasting effects on biological aging processes, even as the specific mechanisms and exposures linked to SES may vary across time and place. Third, although longitudinal, the current study is correlational, which precludes causal interpretation. It is possible that third factors relate to both reported childhood SES and markers of biological aging, such as inter-generational genetic influences or current health. Prospective studies following individuals throughout life are needed to more accurately identify predictors of trajectories of aging. Fourth, all participants were largely from one county in Pennsylvania, and because of exclusion criteria, were relatively healthy, so our results might not generalize to the broader population or higher-risk populations. Future work should also include individuals from a wider range of childhood SES, including a more representative sample of individuals from lower-SES populations, and consider factors that contribute to the association of childhood SES with biological aging, including access to healthcare and health literacy, and intersectionality (with race and exposure to unfair treatment). We did not test race as an effect modifier in this sample due to limited power (13% identified with racial groups other than White); future work could consider testing race moderation because structural racism shapes socioeconomic conditions across the life course (Shapiro, 2017), leading to lasting effects on health. Finally, the impact of retirement on adult SES was not assessed and warrants consideration in future work.

Conclusion

The current longitudinal findings support low childhood socioeconomic conditions as a social hallmark of accelerated biological aging that may contribute to adult health disparities and shorter healthspan. In the current study, lower childhood SES associates with a larger increase in a multisystem composite of blood-based biomarkers of aging across an ~ 14-year period starting at a mean age of 45.5 years. This association is independent of SES in adulthood and recalled exposure to childhood trauma. Future work should consider the possibility that programs and policies aimed at improving the socioeconomic circumstances of families may promote the life-long health of offspring and thus public health (MacGuire, 2020).

Data Availability:

Data are available upon reasonable request via data use agreement to reproduce these results.