Association between the oxidative balance score and biological aging measured by epigenetic clock: insight from the NHANES 1999-2002 Authors: Mingming Lv, Zhihui Jiang, Changjiang Deng, Zhiyan Du, Bingxin Bai, Zhilong Wang, Adilai Adilijiang, Ying-Ying Zheng, Xiang Xie Published: 2026 ABSTRACT BackgroundOxidative stress is a key factor leading to oxidative damage in cells and tissues, and is a major driving force behind aging and various age-related diseases. Epigenetic clock and oxidative balance score (OBS) serve as reliable indicators for assessing an individual’s aging process and oxidative stress levels, respectively. However, no study has comprehensively assessed the association between epigenetic clock and OBS.MethodsThis study analysed 1,797 subjects from the 1999–2002 National Health and Nutrition Examination Survey (NHANES). Multivariate regression models were used to assess the relationship between OBS and epigenetic age as well as epigenetic age acceleration (EAA). Restricted cubic spline (RCS) analyses were used to visualize the dose-response relationship between OBS and EAA. Subgroup analysis and interaction tests were used to investigate whether this association was stable across populations.ResultsMultiple linear regression analyses showed that higher OBS levels were significantly associated with lower epigenetic age. Further analyses showed that both the dietary OBS and the lifestyle OBS were significantly negatively correlated with epigenetic age. RCS analysis revealed a significant negative linear dose–response relationship between OBS and EAA. The association between OBS and EAA was homogeneous across subgroups.ConclusionsThis study suggests higher OBS levels are associated with reduced epigenetic age and decelerated biological aging, and antioxidant-rich diets and lifestyles may delay biological aging.Supplementary InformationThe online version contains supplementary material available at 10.1186/s13148-026-02141-w. Introduction As of 2019, 88 out of 188 countries have been defined as aging societies and population aging has increased the global burden of non-communicable diseases (NCDs), including death and disability from disease [1]. Aging is characterized by a progressive loss of physiological integrity, leading to impaired function and increased susceptibility to disorders [2]. Geroscience research has indicated that aging is the predominant risk factor for most age-related diseases, and that slowing the aging process can delay or prevent many chronic diseases [3]. Aging is a multifaceted process, and in order to slow down or even reverse it, there is a need for effective and reliable biomarkers of aging. One hallmark of aging is epigenetic changes or age-related DNA methylation changes [4]. In recent years, epigenetic clocks based on DNA methylation of selected sites have emerged as a precise tool for assessing aging [2]. Its age prediction, known as DNA methylation age or epigenetic age, is compared with chronological age to indicate accelerated or decelerated aging. Studies have shown that epigenetic age acceleration (EAA) is significantly associated with the incidence and mortality of various age-related diseases [5–8]. Individuals with EAA may face a higher risk of adverse health outcomes, indicating that epigenetic clocks are valuable aging biomarkers for predicting biological age and mortality risk, as well as for evaluating interventions that could potentially extend human lifespan. Oxidative stress is a key factor leading to the oxidative damage of cells and tissues and plays a crucial role in the development and progression of aging [9]. Oxidative stress induces cellular senescence through various pathways, such as mitochondrial dysfunction, DNA damage and telomere shortening, which in turn lead to a range of age-related diseases [9, 10]. However, the level of oxidative stress in the body is influenced by various factors, and an indicator that comprehensively reflects the redox state of an organism is crucial for stratifying the risk of aging. The oxidation balance score (OBS), calculated based on 20 antioxidant and pro-oxidative components in dietary and lifestyle factors, serves as a comprehensive indicator for assessing an individual’s oxidative/antioxidative status [11]. Generally, higher OBS indicates stronger antioxidant capacity and lower levels of oxidative stress in the body. Studies have shown that higher OBS are associated with a reduced mortality rate from various age-related diseases, including hypertension [12], diabetes and prediabetes [13], chronic kidney disease [14], and cancer [15]. In conclusion, epigenetic clock and OBS serve as reliable indicators for assessing an individual’s aging process and oxidative stress levels, respectively. However, no study has comprehensively assessed the association between OBS and biological aging measured by epigenetic clock. Therefore, the present study aims to investigate this association using data from the National Health and Nutrition Examination Survey (NHANES). Methods Study population This study included subjects from the NHANES 1999–2000 and 2001–2002, because DNA methylation was measured from biospecimens collected from participants during these cycles. The NHANES is a nationally representative cross-sectional study conducted in the United States by the National Center for Health Statistics (NCHS), with the primary objective of assessing the health and nutritional status of individuals. Data collection followed a complex, stratified, multistage probability sampling design, and all participants provided written informed consent. We merged data from two NHANES cycles, which included a total of 21,004 participants. The following individuals were excluded : (1) participants without OBS component data (n = 9,149); (2) participants without DNA methylation data (n = 9,694); (3) participants with extreme energy intake (n = 35); (4) participants without demographic data (n = 329). Ultimately, the study included 1,797 participants (Fig. 1).Fig. 1Flowchart for selecting participants. Flowchart for selecting participants. DNA methylation and epigenetic biomarker production DNA was extracted from whole blood of NHANES participants aged 50 and above from the 1999–2000 and 2001–2002 cycles and stored at −80°C. DNA methylation was measured using Illumina EPIC BeadChip arrays in the laboratory of Dr. Yongmei Liu at Duke University. The IDAT files generated by the Illumina EPIC BeadChip arrays were used to convert the red and green intensity signals into methylated and unmethylated signals, followed by quality control through color correction, background subtraction, and outlier removal. After imputation and normalization of the methylation data matrix, the coefficients provided by the authors for each epigenetic biomarker were applied to the corresponding subset of probes to generate various epigenetic biomarkers, including HorvathAge [16], HannumAge [17], SkinBloodAge [18], PhenoAge [19], GrimAge [20], GrimAge2 [21], ZhangAge [22], LinAge [23], WeidnerAge [24], VidalBraloAge [25], HorvathTelo [26], DunedinPoAm [27]. All pre-processing and DNA methylation epigenetic biomarker production were completed in the RStudio environment using the R language. In July 2024, NHANES publicly released these data files, from which we directly extracted the precomputed epigenetic biomarker variables for analysis. This dataset has been used in prior published studies [28]. Definition of OBS According to previous literature, the calculation of the OBS includes 16 nutrients and 4 lifestyle factors, comprising 15 antioxidants and 5 pro-oxidants [29]. The 16 nutrients were obtained through a 24-hour dietary recall survey and include dietary fiber, carotene, vitamin B2, niacin, vitamin B6, total folate, vitamin B12, vitamin C, vitamin E, calcium, magnesium, zinc, copper, selenium, total fat, and iron. The four lifestyle factors include physical activity, body mass index (BMI), alcohol consumption, and smoking. Smoking is estimated by cotinine levels, and physical activity data include detailed information on moderate and vigorous leisure activities, with intensity quantified using metabolic equivalent (MET) scores. The 5 pro-oxidant components include total fat, iron, BMI, alcohol consumption, and cotinine, while the remaining factors are classified as antioxidants. For alcohol consumption, non-drinkers, non-heavy drinkers (female 0–15g/day, male 0–30g/day), and heavy drinkers (female ≥ 15g/day, male ≥ 30g/day) are assigned scores of 2, 1, and 0 points, respectively. Physical activity was classified as low (< 600 MET-minute per week), moderate (600–3000 MET-minute per week), and high physical activity (≥ 3000 MET-minute per week), and these were assigned 0, 1, and 2 points, respectively [30]. We determined scores for the other components based on gender and the tertiles of each component. Antioxidants were assigned points from 0 to 2 for the lowest to the highest tertile group, respectively. The point assignment for pro-oxidants was inverse, with 0 points for the highest tertile group and 2 points for the lowest tertile group. Table 1 shows the assignment scheme of the OBS components.Table 1Oxidative balance score assignment scheme.OBS componentsPropertyMaleFemale012012Dietary OBS componentsDietary fiber (g/d)A≤11.8011.80-20.30>20.30≤10.0010.00-15.86>15.86Carotene (RE/d)A≤161.00161.00-434.00>434.00≤165.33165.33-382.67>382.67Vitamin B2 (mg/d)A≤1.531.53-2.33>2.33≤1.211.21-1.81>1.81Niacin (mg/d)A≤17.3617.36-25.21>25.21≤12.8212.82-18.83>18.83Vitamin B6 (mg/d)A≤1.401.40-2.26>2.26≤1.081.08-1.63>1.63Total folate (mcg/d)A≤287.00287.00-453.00>453.00≤238.00238.00-364.00>364.00Vitamin B12 (mcg/d)A≤2.912.91-5.61>5.61≤2.012.01-3.85>3.85Vitamin C (mg/d)A≤43.6043.60-113.75>113.75≤44.4844.48-108.62>108.62Vitamin E (ATE) (mg/d)A≤4.974.97-8.44>8.44≤4.194.19-6.95>6.95Calcium (mg/d)A≤476.00476.00-858.00>858.00≤429.18429.18-721.67>721.67Magnesium (mg/d)A≤219.00219.00-332.90>332.90≤176.37176.37-258.03>258.03Zinc (mg/d)A≤8.148.14-13.17>13.17≤6.016.01-9.42>9.42Copper (mg/d)A≤0.980.98-1.46>1.46≤0.760.76-1.14>1.14Selenium (mcg/d)A≤81.9181.91-122.80>122.80≤63.3563.35-90.87>90.87Total fat (g/d)P>88.2455.97-88.24≤55.97>66.1742.86-66.17≤42.86Iron (mg/d)P>17.5211.20-17.52≤11.20>13.619.06-13.61≤9.06Lifestyle OBS componentsPhysical activity (MET-min/wk)A<600600-3000≥3000<600600-3000≥3000Alcohol (g/d)P300-30None150–15NoneBody mass index (kg/m2)P>29.4826.05-29.48≤26.05>30.6525.86-30.65≤25.86Cotinine (ng/mL)P>0.240.04-0.24≤0.04>0.110.04-0.11≤0.04OBS: oxidative balance score; A: antioxidant; P: pro-oxidant; RE: retinol equivalent; ATE: alpha-tocopherol equivalent; MET: metabolic equivalent. Table 1Oxidative balance score assignment scheme.OBS componentsPropertyMaleFemale012012Dietary OBS componentsDietary fiber (g/d)A≤11.8011.80-20.30>20.30≤10.0010.00-15.86>15.86Carotene (RE/d)A≤161.00161.00-434.00>434.00≤165.33165.33-382.67>382.67Vitamin B2 (mg/d)A≤1.531.53-2.33>2.33≤1.211.21-1.81>1.81Niacin (mg/d)A≤17.3617.36-25.21>25.21≤12.8212.82-18.83>18.83Vitamin B6 (mg/d)A≤1.401.40-2.26>2.26≤1.081.08-1.63>1.63Total folate (mcg/d)A≤287.00287.00-453.00>453.00≤238.00238.00-364.00>364.00Vitamin B12 (mcg/d)A≤2.912.91-5.61>5.61≤2.012.01-3.85>3.85Vitamin C (mg/d)A≤43.6043.60-113.75>113.75≤44.4844.48-108.62>108.62Vitamin E (ATE) (mg/d)A≤4.974.97-8.44>8.44≤4.194.19-6.95>6.95Calcium (mg/d)A≤476.00476.00-858.00>858.00≤429.18429.18-721.67>721.67Magnesium (mg/d)A≤219.00219.00-332.90>332.90≤176.37176.37-258.03>258.03Zinc (mg/d)A≤8.148.14-13.17>13.17≤6.016.01-9.42>9.42Copper (mg/d)A≤0.980.98-1.46>1.46≤0.760.76-1.14>1.14Selenium (mcg/d)A≤81.9181.91-122.80>122.80≤63.3563.35-90.87>90.87Total fat (g/d)P>88.2455.97-88.24≤55.97>66.1742.86-66.17≤42.86Iron (mg/d)P>17.5211.20-17.52≤11.20>13.619.06-13.61≤9.06Lifestyle OBS componentsPhysical activity (MET-min/wk)A<600600-3000≥3000<600600-3000≥3000Alcohol (g/d)P300-30None150–15NoneBody mass index (kg/m2)P>29.4826.05-29.48≤26.05>30.6525.86-30.65≤25.86Cotinine (ng/mL)P>0.240.04-0.24≤0.04>0.110.04-0.11≤0.04OBS: oxidative balance score; A: antioxidant; P: pro-oxidant; RE: retinol equivalent; ATE: alpha-tocopherol equivalent; MET: metabolic equivalent. Oxidative balance score assignment scheme. OBS: oxidative balance score; A: antioxidant; P: pro-oxidant; RE: retinol equivalent; ATE: alpha-tocopherol equivalent; MET: metabolic equivalent. Covariate assessment In this study, the covariates comprised age, gender (male/female), race/ethnicity (Mexican American, non-Hispanic white, non-Hispanic black, other race), education (less than high school, high school diploma, some college or higher), marital status (divorced/separated/widowed, married/living with a partner, never married), poverty income ratio (PIR) and energy intake. Statistical analysis Baseline characteristics were expressed by quartiles of the OBS. Continuous variables were presented as mean ± standard deviation (SD), and categorical variables as frequencies and percentages. Group differences across OBS quartiles were assessed using chi-square tests for categorical variables and ANOVA for continuous variables. Pearson correlation coefficients were used to examine the association between chronological age and epigenetic age. For epigenetic clocks other than HorvathTelo and DunedinPoAm, the residual values obtained from the linear regression model comparing epigenetic age to chronological age represent aging. Positive values constitutes EAA, indicating a higher epigenetic age relative to chronological age and thus faster biological aging. Given the complex sampling design of NHANES, we developed three weighted multivariate linear regression models to evaluate the relationship between OBS and epigenetic age. Model 1 was unadjusted, Model 2 adjusted for age, gender, and race, and Model 3 further adjusted for education, marital status, PIR, and energy intake. To thoroughly investigate the relationship between OBS and epigenetic age, OBS was analyzed both as a continuous and a categorical variable. To further examine the relationship between OBS and EAA, we also constructed three weighted multivariable logistic regression models using the same adjustment sets. We constructed restricted cubic spline (RCS) analyses based on the fully adjusted logistic regression model to visualize the dose-response relationship between OBS and EAA. To determine the optimal number of knots, we calculated the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) for different knot numbers (3-6) to guide model selection. Subgroup analyses and interaction tests were performed using the fully adjusted model to explore potential effect modification. Additionally, sensitivity analyses were conducted to separately assess the associations of dietary OBS and lifestyle OBS with epigenetic age. Statistical analysis followed the CDC analysis guidelines. All statistical analyses were performed using R software (version 4.4.3). A two-tailed p-value < 0.05 was considered statistically significant. Results Baseline characteristics Table 2 shows the weighted baseline characteristics of the study population categorised by quartiles of OBS. A total of 1,797 eligible participants were included in this analysis with a mean age of 63.15 ± 9.95 years; 940 (46.76%) were male. Compared to the lowest OBS quartile (Q1), participants in the highest OBS quartile (Q4) had higher proportions of non-Hispanic white, higher education levels, higher PIR and higher energy intakes (all P < 0.001). There were no statistically significant differences in age, gender, and marital status among the groups (all P > 0.05).Table 2Baseline characteristics of participants categorised by quartiles of OBS.VariablesTotal (n = 1,797)Q1 (n = 461)Q2 (n = 523)Q3 (n = 420)Q4 (n = 393)PAge, years63.15 ± 9.9563.66 ± 9.9763.13 ± 10.1263.47 ± 9.8362.41 ± 9.820.278Age group, n(%)0.904< 65892 (58.38%)231 (57.25%)259 (58.12%)209 (57.85%)193 (60.16%)≥ s65905 (41.62%)230 (42.75%)264 (41.88%)211 (42.15%)200 (39.84%)Gender, n(%)0.961Male940 (46.76%)238 (45.07%)279 (46.79%)219 (47.41%)204 (47.52%)Famale857 (53.24%)223 (54.93%)244 (53.21%)201 (52.59%)189 (52.48%)Race, n(%)<0.001Mexican American504 (3.37%)132 (4.04%)149 (3.50%)117 (3.02%)106 (3.02%)Non-Hispanic White771 (80.72%)144 (73.10%)221 (79.61%)193 (81.35%)213 (87.75%)Non-Hispanic Black357 (7.64%)147 (14.33%)104 (7.96%)63 (5.65%)43 (3.61%)Other Race165 (8.26%)38 (8.53%)49 (8.93%)47 (9.98%)31 (5.62%)Education, n(%)<0.001Less Than High School421 (8.60%)141 (13.03%)131 (10.08%)83 (7.22%)66 (4.58%)High School Diploma339 (16.09%)116 (21.34%)107 (20.15%)78 (17.03%)38 (6.18%)Some college or higher1,037 (75.31%)204 (65.63%)285 (69.78%)259 (75.75%)289 (89.24%)Marital status, n(%)0.055Married or living together1,200 (70.64%)272 (64.92%)377 (75.87%)280 (68.63%)271 (71.49%)Alone597 (29.36%)189 (35.08%)146 (24.13%)140 (31.37%)122 (28.51%)PIR, n(%)<0.001< 1.3478 (17.26%)165 (26.54%)138 (17.20%)106 (18.15%)69 (8.70%)1.3–3.5716 (35.49%)191 (42.29%)220 (37.29%)163 (33.36%)142 (29.85%)> 3.5603 (47.24%)105 (31.16%)165 (45.51%)151 (48.49%)182 (61.45%)Energy, kcal1,933 ± 7781,275 ± 5071,750 ± 5612,165 ± 6142,464 ± 845<0.001Q1: OBS ≤ 14; Q2: 14 < OBS ≤ 21; Q3: 21 < OBS ≤ 26; Q4: OBS > 26. “Alone” included widowed, divorced, separated, and never married participants.Frequencies (n) are unweighted; percentages are weighted. P value was calculated by ANOVA and chi-square test. Table 2Baseline characteristics of participants categorised by quartiles of OBS.VariablesTotal (n = 1,797)Q1 (n = 461)Q2 (n = 523)Q3 (n = 420)Q4 (n = 393)PAge, years63.15 ± 9.9563.66 ± 9.9763.13 ± 10.1263.47 ± 9.8362.41 ± 9.820.278Age group, n(%)0.904< 65892 (58.38%)231 (57.25%)259 (58.12%)209 (57.85%)193 (60.16%)≥ s65905 (41.62%)230 (42.75%)264 (41.88%)211 (42.15%)200 (39.84%)Gender, n(%)0.961Male940 (46.76%)238 (45.07%)279 (46.79%)219 (47.41%)204 (47.52%)Famale857 (53.24%)223 (54.93%)244 (53.21%)201 (52.59%)189 (52.48%)Race, n(%)<0.001Mexican American504 (3.37%)132 (4.04%)149 (3.50%)117 (3.02%)106 (3.02%)Non-Hispanic White771 (80.72%)144 (73.10%)221 (79.61%)193 (81.35%)213 (87.75%)Non-Hispanic Black357 (7.64%)147 (14.33%)104 (7.96%)63 (5.65%)43 (3.61%)Other Race165 (8.26%)38 (8.53%)49 (8.93%)47 (9.98%)31 (5.62%)Education, n(%)<0.001Less Than High School421 (8.60%)141 (13.03%)131 (10.08%)83 (7.22%)66 (4.58%)High School Diploma339 (16.09%)116 (21.34%)107 (20.15%)78 (17.03%)38 (6.18%)Some college or higher1,037 (75.31%)204 (65.63%)285 (69.78%)259 (75.75%)289 (89.24%)Marital status, n(%)0.055Married or living together1,200 (70.64%)272 (64.92%)377 (75.87%)280 (68.63%)271 (71.49%)Alone597 (29.36%)189 (35.08%)146 (24.13%)140 (31.37%)122 (28.51%)PIR, n(%)<0.001< 1.3478 (17.26%)165 (26.54%)138 (17.20%)106 (18.15%)69 (8.70%)1.3–3.5716 (35.49%)191 (42.29%)220 (37.29%)163 (33.36%)142 (29.85%)> 3.5603 (47.24%)105 (31.16%)165 (45.51%)151 (48.49%)182 (61.45%)Energy, kcal1,933 ± 7781,275 ± 5071,750 ± 5612,165 ± 6142,464 ± 845<0.001Q1: OBS ≤ 14; Q2: 14 < OBS ≤ 21; Q3: 21 < OBS ≤ 26; Q4: OBS > 26. “Alone” included widowed, divorced, separated, and never married participants.Frequencies (n) are unweighted; percentages are weighted. P value was calculated by ANOVA and chi-square test. Baseline characteristics of participants categorised by quartiles of OBS. Q1: OBS ≤ 14; Q2: 14 < OBS ≤ 21; Q3: 21 < OBS ≤ 26; Q4: OBS > 26. “Alone” included widowed, divorced, separated, and never married participants. Frequencies (n) are unweighted; percentages are weighted. P value was calculated by ANOVA and chi-square test. Correlations between epigenetic age and chronological age Figure 2 presents the correlation coefficients (r) between chronological age and 12 epigenetic clocks. Except for WeidnerAge and VidalBraloAge, we observed a strong positive correlation between chronological age and epigenetic age, ranging from r = 0.77 (LinAge or PhenoAge) to r = 0.89 (ZhangAge). HorvathTelo was negatively correlated with chronological age (r = −0.60), as telomere length shortened with the increase of age. As expected, DunedinPoAm, an algorithm for quantifying the pace of biological aging in humans, was independent of chronological age (r = 0.05).Fig. 2Scatter plots of epigenetic age against chronological age for each epigenetic clock, with Pearson correlation coefficients (r). MAE: median absolute error. Scatter plots of epigenetic age against chronological age for each epigenetic clock, with Pearson correlation coefficients (r). MAE: median absolute error. Associations between epigenetic age and OBS Table 3 shows the results of the multivariable linear regression analyses of the associations between epigenetic age and OBS. We found that higher OBS levels was negatively associated with GrimAge and GrimAge2 in all three models. In the fully adjusted model (Model 3), each unit increase in OBS as a continuous variable was significantly associated with a decrease in epigenetic age (GrimAge: β = −0.12, 95% CI: ( −0.18, −0.05), P = 0.001, GrimAge2: β = −0.14, 95% CI: ( −0.20, −0.07), P < 0.001). when OBS was treated as a categorical variable and compared with Q1, participants in Q4 showed a significant reduction in epigenetic age (GrimAge: β = −2.32, 95% CI: ( −3.66, −0.99), P = 0.002, GrimAge2: β = −2.71, 95% CI: ( −4.16, −1.26), P = 0.001). Additionally, HorvathTelo and DunedinPoAm each exhibited a significant association with OBS. When OBS was treated as a categorical variable and compared with Q1 in the fully adjusted model (Model 3), participants in Q4 showed a significant increase in HorvathTelo (β = 0.08, 95% CI: (0.00, 0.16), P = 0.043) and a significant decrease in DunedinPoAm (β = −0.04, 95% CI: ( −0.07, −0.02), P = 0.002).Table 3Multivariate linear regression analysis of the association between epigenetic age and OBS.Epigenetic ageOBSModel 1PModel 2PModel 3Pβ (95%CI)β (95%CI)β (95%CI)HorvathAgeContinuous −0.06 ( −0.13, 0.00)0.065 −0.06 ( −0.11, −0.02)0.0090.01 ( −0.06, 0.08)0.688Q 1RefRefRefQ 2 −0.42 ( −2.41, 1.57)0.669 −0.37 ( −1.61, 0.86)0.5350.24 ( −1.05, 1.54)0.696Q 3 −0.85 ( −2.50, 0.80)0.297 −0.86 ( −1.80, 0.07)0.0680.17 ( −0.99, 1.33)0.759Q 4 −1.30 ( −2.86, 0.27)0.100 −1.06 ( −2.02, −0.10)0.0320.39 ( −0.92, 1.71)0.533P for trend0.0530.0170.595HannumAgeContinuous −0.07 ( −0.15, 0.00)0.060 −0.08 ( −0.12, −0.03)0.0040.02 ( −0.06, 0.09)0.627Q 1RefRefRefQ 2 −0.78 ( −3.19, 1.63)0.511 −0.81 ( −2.23, 0.62)0.252 −0.02 ( −1.50, 1.46)0.979Q 3 −0.48 ( −2.53, 1.56)0.630 −0.60 ( −1.83, 0.63)0.3210.66 ( −0.71, 2.03)0.321Q 4 −1.60 ( −3.27, 0.07)0.060 −1.42 ( −2.53, −0.32)0.0140.40 ( −1.07, 1.87)0.571P for trend0.0760.0220.377SkinBloodAgecategories −0.06 ( −0.12, −0.00)0.039 −0.06 ( −0.10, −0.02)0.0030.03 ( −0.04, 0.10)0.407Q 1RefRefRefQ 2 −0.16 ( −2.39, 2.08)0.885 −0.08 ( −1.35, 1.18)0.8910.65 ( −0.74, 2.03)0.334Q 3 −0.31 ( −2.11, 1.49)0.724 −0.30 ( −1.22, 0.63)0.5150.91 ( −0.32, 2.15)0.136Q 4 −1.22 ( −2.74, 0.29)0.110 −0.89 ( −1.80, 0.03)0.0580.87 ( −0.61, 2.35)0.229P for trend0.0860.0430.201PhenoAgeContinuous −0.13 ( −0.21, −0.04)0.006 −0.13 ( −0.19, −0.07)<0.001 −0.05 ( −0.14, 0.04)0.293Q 1RefRefRefQ 2 −1.19 ( −3.62, 1.23)0.321 −1.13 ( −2.62, 0.36)0.130 −0.48 ( −2.14, 1.18)0.548Q 3 −0.77 ( −2.89, 1.35)0.462 −0.76 ( −2.29, 0.76)0.3100.26 ( −1.60, 2.11)0.774Q 4 −2.88 ( −4.82, −0.95)0.005 −2.59 ( −3.81, −1.37)<0.001 −1.02 ( −2.60, 0.56)0.189P for trend0.006<0.0010.382GrimAgeContinuous −0.18 ( −0.25, −0.11)<0.001 −0.18 ( −0.23, −0.13)<0.001 −0.12 ( −0.18, −0.05)0.001Q 1RefRefRefQ 2 −1.17 ( −3.17, 0.83)0.240 −1.08 ( −2.17, 0.01)0.052 −0.51 ( −1.71, 0.69)0.382Q 3 −1.49 ( −3.16, 0.18)0.078 −1.44 ( −2.43, −0.45)0.007 −0.70 ( −1.78, 0.38)0.186Q 4 −3.80 ( −5.32, −2.28)<0.001 −3.50 ( −4.55, −2.45)<0.001 −2.32 ( −3.66, −0.99)0.002P for trend<0.001<0.0010.002GrimAge2Continuous −0.20 ( −0.27, −0.13)<0.001 −0.19 ( −0.24, −0.14)<0.001 −0.14 ( −0.20, −0.07)<0.001Q 1RefRefRefQ 2 −1.33 ( −3.37, 0.70)0.190 −1.21 ( −2.42, 0.01)0.051 −0.68 ( −2.02, 0.66)0.296Q 3 −1.57 ( −3.15, 0.01)0.051 −1.46 ( −2.46, −0.45)0.007 −0.81 ( −1.99, 0.38)0.167Q 4 −4.19 ( −5.72, −2.65)<0.001 −3.81 ( −4.92, −2.70)<0.001 −2.71 ( −4.16, −1.26)0.001P for trend<0.001<0.0010.001ZhangAgeContinuous −0.02 ( −0.05, 0.00)0.061 −0.02 ( −0.04, −0.01)0.0020.01 ( −0.02, 0.03)0.467Q 1RefRefRefQ 2 −0.22 ( −1.14, 0.69)0.616 −0.21 ( −0.69, 0.27)0.3700.05 ( −0.48, 0.59)0.837Q 3 −0.15 ( −0.86, 0.56)0.673 −0.16 ( −0.51, 0.18)0.3370.28 ( −0.18, 0.75)0.216Q 4 −0.58 ( −1.17, 0.02)0.057 −0.48 ( −0.83, −0.12)0.0100.16 ( −0.35, 0.67)0.512P for trend0.0560.0130.323LinAgeContinuous −0.03 ( −0.11, 0.05)0.387 −0.04 ( −0.11, 0.04)0.3160.08 ( −0.02, 0.18)0.100Q 1RefRefRefQ 2 −0.07 ( −2.57, 2.43)0.954 −0.00 ( −1.37, 1.36)0.9940.94 ( −0.67, 2.55)0.231Q 30.30 ( −1.63, 2.23)0.7530.31 ( −1.18, 1.79)0.6711.93 (0.07, 3.79)0.043Q 4 −1.05 ( −2.87, 0.76)0.244 −0.77 ( −2.21, 0.67)0.2791.40 ( −0.59, 3.39)0.155P for trend0.2920.4110.104WeidnerAgeContinuous −0.01 ( −0.11, 0.09)0.864 −0.01 ( −0.11, 0.08)0.8000.05 ( −0.07, 0.16)0.390Q 1RefRefRefQ 2 −0.90 ( −2.34, 0.54)0.209 −0.86 ( −2.45, 0.73)0.274 −0.36 ( −2.15, 1.44)0.679Q 3 −1.02 ( −2.91, 0.87)0.278 −1.01 ( −2.72, 0.71)0.238 −0.26 ( −2.14, 1.61)0.768Q 4 −0.54 ( −2.44, 1.36)0.564 −0.39 ( −2.24, 1.45)0.6630.57 ( −1.71, 2.86)0.600P for trend0.5890.6670.608VidalBraloAgeContinuous −0.01 ( −0.06, 0.05)0.806 −0.02 ( −0.06, 0.02)0.2930.02 ( −0.04, 0.08)0.578Q 1RefRefRefQ 2 −0.64 ( −1.92, 0.64)0.317 −0.75 ( −1.66, 0.17)0.104 −0.43 ( −1.49, 0.62)0.396Q 30.02 ( −1.16, 1.20)0.973 −0.16 ( −1.08, 0.77)0.7270.35 ( −0.71, 1.41)0.494Q 4 −0.58 ( −1.66, 0.49)0.274 −0.74 ( −1.51, 0.03)0.059 −0.06 ( −1.21, 1.09)0.911P for trend0.5520.2530.682HorvathTeloContinuous0.00 (0.00, 0.01)0.0040.01 (0.00, 0.01)<0.0010.00 (0.00, 0.01)0.023Q 1RefRefRefQ 20.05 ( −0.02, 0.12)0.1590.06 (0.00, 0.12)0.0440.05 ( −0.02, 0.11)0.138Q 30.07 ( −0.00, 0.14)0.0520.08 (0.03, 0.14)0.0030.07 (0.00, 0.13)0.042Q 40.10 (0.03, 0.17)0.0070.11 (0.06, 0.17)<0.0010.08 (0.00, 0.16)0.043P for trend0.004<0.0010.031DunedinPoAmContinuous −0.00 ( −0.00, −0.00)<0.001 −0.00 ( −0.00, −0.00)<0.001 −0.00 ( −0.00, −0.00)0.001Q 1RefRefRefQ 2 −0.01 ( −0.03, 0.00)0.083 −0.01 ( −0.03, 0.00)0.106 −0.01 ( −0.03, 0.01)0.264Q 3 −0.01 ( −0.03, 0.01)0.153 −0.01 ( −0.03, 0.01)0.182 −0.01 ( −0.04, 0.01)0.354Q 4 −0.05 ( −0.07, −0.03)<0.001 −0.05 ( −0.07, −0.03)<0.001 −0.04 ( −0.07, −0.02)0.002P for trend<0.001<0.0010.003β: beta coefficient; CI: confidence interval.Model 1 was unadjusted,Model 2 was adjusted for age, gender, and race,Model 3 was adjusted for age, gender, race, education, marital status, PIR, and energy intake.Bold indicates significant at P < 0.05 Table 3Multivariate linear regression analysis of the association between epigenetic age and OBS.Epigenetic ageOBSModel 1PModel 2PModel 3Pβ (95%CI)β (95%CI)β (95%CI)HorvathAgeContinuous −0.06 ( −0.13, 0.00)0.065 −0.06 ( −0.11, −0.02)0.0090.01 ( −0.06, 0.08)0.688Q 1RefRefRefQ 2 −0.42 ( −2.41, 1.57)0.669 −0.37 ( −1.61, 0.86)0.5350.24 ( −1.05, 1.54)0.696Q 3 −0.85 ( −2.50, 0.80)0.297 −0.86 ( −1.80, 0.07)0.0680.17 ( −0.99, 1.33)0.759Q 4 −1.30 ( −2.86, 0.27)0.100 −1.06 ( −2.02, −0.10)0.0320.39 ( −0.92, 1.71)0.533P for trend0.0530.0170.595HannumAgeContinuous −0.07 ( −0.15, 0.00)0.060 −0.08 ( −0.12, −0.03)0.0040.02 ( −0.06, 0.09)0.627Q 1RefRefRefQ 2 −0.78 ( −3.19, 1.63)0.511 −0.81 ( −2.23, 0.62)0.252 −0.02 ( −1.50, 1.46)0.979Q 3 −0.48 ( −2.53, 1.56)0.630 −0.60 ( −1.83, 0.63)0.3210.66 ( −0.71, 2.03)0.321Q 4 −1.60 ( −3.27, 0.07)0.060 −1.42 ( −2.53, −0.32)0.0140.40 ( −1.07, 1.87)0.571P for trend0.0760.0220.377SkinBloodAgecategories −0.06 ( −0.12, −0.00)0.039 −0.06 ( −0.10, −0.02)0.0030.03 ( −0.04, 0.10)0.407Q 1RefRefRefQ 2 −0.16 ( −2.39, 2.08)0.885 −0.08 ( −1.35, 1.18)0.8910.65 ( −0.74, 2.03)0.334Q 3 −0.31 ( −2.11, 1.49)0.724 −0.30 ( −1.22, 0.63)0.5150.91 ( −0.32, 2.15)0.136Q 4 −1.22 ( −2.74, 0.29)0.110 −0.89 ( −1.80, 0.03)0.0580.87 ( −0.61, 2.35)0.229P for trend0.0860.0430.201PhenoAgeContinuous −0.13 ( −0.21, −0.04)0.006 −0.13 ( −0.19, −0.07)<0.001 −0.05 ( −0.14, 0.04)0.293Q 1RefRefRefQ 2 −1.19 ( −3.62, 1.23)0.321 −1.13 ( −2.62, 0.36)0.130 −0.48 ( −2.14, 1.18)0.548Q 3 −0.77 ( −2.89, 1.35)0.462 −0.76 ( −2.29, 0.76)0.3100.26 ( −1.60, 2.11)0.774Q 4 −2.88 ( −4.82, −0.95)0.005 −2.59 ( −3.81, −1.37)<0.001 −1.02 ( −2.60, 0.56)0.189P for trend0.006<0.0010.382GrimAgeContinuous −0.18 ( −0.25, −0.11)<0.001 −0.18 ( −0.23, −0.13)<0.001 −0.12 ( −0.18, −0.05)0.001Q 1RefRefRefQ 2 −1.17 ( −3.17, 0.83)0.240 −1.08 ( −2.17, 0.01)0.052 −0.51 ( −1.71, 0.69)0.382Q 3 −1.49 ( −3.16, 0.18)0.078 −1.44 ( −2.43, −0.45)0.007 −0.70 ( −1.78, 0.38)0.186Q 4 −3.80 ( −5.32, −2.28)<0.001 −3.50 ( −4.55, −2.45)<0.001 −2.32 ( −3.66, −0.99)0.002P for trend<0.001<0.0010.002GrimAge2Continuous −0.20 ( −0.27, −0.13)<0.001 −0.19 ( −0.24, −0.14)<0.001 −0.14 ( −0.20, −0.07)<0.001Q 1RefRefRefQ 2 −1.33 ( −3.37, 0.70)0.190 −1.21 ( −2.42, 0.01)0.051 −0.68 ( −2.02, 0.66)0.296Q 3 −1.57 ( −3.15, 0.01)0.051 −1.46 ( −2.46, −0.45)0.007 −0.81 ( −1.99, 0.38)0.167Q 4 −4.19 ( −5.72, −2.65)<0.001 −3.81 ( −4.92, −2.70)<0.001 −2.71 ( −4.16, −1.26)0.001P for trend<0.001<0.0010.001ZhangAgeContinuous −0.02 ( −0.05, 0.00)0.061 −0.02 ( −0.04, −0.01)0.0020.01 ( −0.02, 0.03)0.467Q 1RefRefRefQ 2 −0.22 ( −1.14, 0.69)0.616 −0.21 ( −0.69, 0.27)0.3700.05 ( −0.48, 0.59)0.837Q 3 −0.15 ( −0.86, 0.56)0.673 −0.16 ( −0.51, 0.18)0.3370.28 ( −0.18, 0.75)0.216Q 4 −0.58 ( −1.17, 0.02)0.057 −0.48 ( −0.83, −0.12)0.0100.16 ( −0.35, 0.67)0.512P for trend0.0560.0130.323LinAgeContinuous −0.03 ( −0.11, 0.05)0.387 −0.04 ( −0.11, 0.04)0.3160.08 ( −0.02, 0.18)0.100Q 1RefRefRefQ 2 −0.07 ( −2.57, 2.43)0.954 −0.00 ( −1.37, 1.36)0.9940.94 ( −0.67, 2.55)0.231Q 30.30 ( −1.63, 2.23)0.7530.31 ( −1.18, 1.79)0.6711.93 (0.07, 3.79)0.043Q 4 −1.05 ( −2.87, 0.76)0.244 −0.77 ( −2.21, 0.67)0.2791.40 ( −0.59, 3.39)0.155P for trend0.2920.4110.104WeidnerAgeContinuous −0.01 ( −0.11, 0.09)0.864 −0.01 ( −0.11, 0.08)0.8000.05 ( −0.07, 0.16)0.390Q 1RefRefRefQ 2 −0.90 ( −2.34, 0.54)0.209 −0.86 ( −2.45, 0.73)0.274 −0.36 ( −2.15, 1.44)0.679Q 3 −1.02 ( −2.91, 0.87)0.278 −1.01 ( −2.72, 0.71)0.238 −0.26 ( −2.14, 1.61)0.768Q 4 −0.54 ( −2.44, 1.36)0.564 −0.39 ( −2.24, 1.45)0.6630.57 ( −1.71, 2.86)0.600P for trend0.5890.6670.608VidalBraloAgeContinuous −0.01 ( −0.06, 0.05)0.806 −0.02 ( −0.06, 0.02)0.2930.02 ( −0.04, 0.08)0.578Q 1RefRefRefQ 2 −0.64 ( −1.92, 0.64)0.317 −0.75 ( −1.66, 0.17)0.104 −0.43 ( −1.49, 0.62)0.396Q 30.02 ( −1.16, 1.20)0.973 −0.16 ( −1.08, 0.77)0.7270.35 ( −0.71, 1.41)0.494Q 4 −0.58 ( −1.66, 0.49)0.274 −0.74 ( −1.51, 0.03)0.059 −0.06 ( −1.21, 1.09)0.911P for trend0.5520.2530.682HorvathTeloContinuous0.00 (0.00, 0.01)0.0040.01 (0.00, 0.01)<0.0010.00 (0.00, 0.01)0.023Q 1RefRefRefQ 20.05 ( −0.02, 0.12)0.1590.06 (0.00, 0.12)0.0440.05 ( −0.02, 0.11)0.138Q 30.07 ( −0.00, 0.14)0.0520.08 (0.03, 0.14)0.0030.07 (0.00, 0.13)0.042Q 40.10 (0.03, 0.17)0.0070.11 (0.06, 0.17)<0.0010.08 (0.00, 0.16)0.043P for trend0.004<0.0010.031DunedinPoAmContinuous −0.00 ( −0.00, −0.00)<0.001 −0.00 ( −0.00, −0.00)<0.001 −0.00 ( −0.00, −0.00)0.001Q 1RefRefRefQ 2 −0.01 ( −0.03, 0.00)0.083 −0.01 ( −0.03, 0.00)0.106 −0.01 ( −0.03, 0.01)0.264Q 3 −0.01 ( −0.03, 0.01)0.153 −0.01 ( −0.03, 0.01)0.182 −0.01 ( −0.04, 0.01)0.354Q 4 −0.05 ( −0.07, −0.03)<0.001 −0.05 ( −0.07, −0.03)<0.001 −0.04 ( −0.07, −0.02)0.002P for trend<0.001<0.0010.003β: beta coefficient; CI: confidence interval.Model 1 was unadjusted,Model 2 was adjusted for age, gender, and race,Model 3 was adjusted for age, gender, race, education, marital status, PIR, and energy intake.Bold indicates significant at P < 0.05 Multivariate linear regression analysis of the association between epigenetic age and OBS. β: beta coefficient; CI: confidence interval. Model 1 was unadjusted, Model 2 was adjusted for age, gender, and race, Model 3 was adjusted for age, gender, race, education, marital status, PIR, and energy intake. Bold indicates significant at P < 0.05 Distributions of epigenetic age residuals stratified by OBS quartiles Figure 3 presents the distribution of residuals derived from regressing epigenetic age on chronological age, stratified by OBS quartiles. For PhenoAge, residuals decreased progressively across increasing OBS quartiles (P = 0.004), suggesting that higher OBS levels are associated with decelerated biological aging as measured by PhenoAge. A similar pattern was observed for GrimAge and GrimAge2 residuals across OBS quartiles (both P < 0.001), indicating a significant slowdown in the aging process. Likewise, untransformed DunedinPoAm values showed a declining pattern with higher OBS levels (P < 0.001), suggesting that the pace of biological aging similarly slows down as OBS levels increase. In contrast, no statistically significant differences were found across OBS quartiles for the remaining epigenetic clocks (all P > 0.05).Fig. 3Distribution of epigenetic age residuals across oxidative balance score (OBS) quartiles. Box plots show the interquartile range (IQR) and median; points represent individual samples. Positive residuals indicate epigenetic age acceleration (EAA). HorvathTelo and DunedinPoAm are presented as untransformed values. Group comparisons were performed using one-way ANOVA, with p-values shown in the figure. Distribution of epigenetic age residuals across oxidative balance score (OBS) quartiles. Box plots show the interquartile range (IQR) and median; points represent individual samples. Positive residuals indicate epigenetic age acceleration (EAA). HorvathTelo and DunedinPoAm are presented as untransformed values. Group comparisons were performed using one-way ANOVA, with p-values shown in the figure. Association between EAA and OBS Table 4 shows the results of the multivariable logistic regression analyses of the associations between EAA and OBS. Consistent with the preceding findings, a significant inverse association was observed between OBS and biological aging acceleration as measured by GrimAge and GrimAge2. In Model 3, each unit increase in OBS was associated with a 5% reduction in the risk of biological aging acceleration (GrimAge Accel: OR = 0.95, 95% CI: (0.92, 0.98), P = 0.004; GrimAge2 Accel: OR = 0.95, 95% CI: (0.93, 0.98), P = 0.003). Compared with Q1, participants in Q4 showed a 59% lower risk of GrimAge accel (OR = 0.41, 95% CI: (0.20, 0.82), P = 0.015) and a 56% lower risk of GrimAge2 accel (OR = 0.44, 95% CI: (0.24, 0.83), P = 0.014). The trend test was also significant (P for trend < 0.05). To further explore the dose-response relationship between EAA and OBS, we also conducted RCS analysis based on the fully adjusted model (Fig. 4). The results revealed a significant negative linear relationship between OBS and EAA measured by GrimAge and GrimAge2.Table 4Multivariate logistic regression analysis of the association between EAA and OBS.EAAOBSModel 1PModel 2PModel 3POR (95%CI)OR (95%CI)OR (95%CI)HorvathAge AccelContinuous0.99 (0.97, 1.01)0.1880.98 (0.96, 1.01)0.1311.00 (0.97, 1.02)0.690Q 1RefRefRefQ 20.91 (0.61, 1.37)0.6510.89 (0.59, 1.35)0.5790.98 (0.64, 1.50)0.937Q 30.75 (0.52, 1.08)0.1160.72 (0.49, 1.06)0.0960.85 (0.56, 1.28)0.417Q 40.85 (0.55, 1.32)0.4530.81 (0.52, 1.27)0.3471.04 (0.63, 1.73)0.870P for trend0.3240.2430.985HannumAge AccelContinuous0.98 (0.96, 1.01)0.1900.98 (0.96, 1.00)0.1121.00 (0.97, 1.03)0.975Q 1RefRefRefQ 21.01 (0.64, 1.60)0.9710.97 (0.61, 1.55)0.8841.10 (0.68, 1.79)0.669Q 30.87 (0.58, 1.32)0.5100.82 (0.53, 1.28)0.3651.02 (0.61, 1.69)0.941Q 40.79 (0.49, 1.28)0.3290.74 (0.46, 1.20)0.2101.06 (0.58, 1.95)0.836P for trend0.2130.1200.939SkinBloodAge AccelCategories0.98 (0.97, 1.00)0.0130.98 (0.97, 0.99)0.0080.99 (0.97, 1.02)0.582Q 1RefRefRefQ 21.16 (0.79, 1.69)0.4301.28 (0.83, 1.96)0.2451.15 (0.78, 1.69)0.471Q 31.20 (0.83, 1.73)0.3181.44 (0.92, 2.25)0.1021.18 (0.82, 1.71)0.355Q 40.78 (0.59, 1.04)0.0921.04 (0.62, 1.76)0.8680.77 (0.58, 1.02)0.071P for trend0.0950.0590.903PhenoAge AccelContinuous0.98 (0.96, 1.00)0.0200.98 (0.95, 1.00)0.0210.98 (0.96, 1.01)0.155Q 1RefRefRefQ 20.84 (0.56, 1.25)0.3780.83 (0.56, 1.24)0.3540.88 (0.58, 1.32)0.514Q 30.86 (0.52, 1.43)0.5510.85 (0.51, 1.43)0.5270.94 (0.54, 1.61)0.798Q 40.60 (0.41, 0.88)0.0110.59 (0.40, 0.88)0.0110.70 (0.46, 1.06)0.088P for trend0.0130.0150.118GrimAge AccelContinuous0.95 (0.94, 0.97)<0.0010.95 (0.93, 0.97)<0.0010.95 (0.92, 0.98)0.004Q 1RefRefRefQ 20.93 (0.70, 1.25)0.6410.93 (0.68, 1.26)0.6120.97 (0.66, 1.45)0.891Q 30.80 (0.56, 1.13)0.1920.78 (0.54, 1.14)0.1850.78 (0.48, 1.28)0.301Q 40.41 (0.27, 0.64)<0.0010.39 (0.24, 0.61)<0.0010.41 (0.20, 0.82)0.015P for trend<0.001<0.0010.009GrimAge2 AccelContinuous0.95 (0.94, 0.97)<0.0010.95 (0.94, 0.97)<0.0010.95 (0.93, 0.98)0.003Q 1RefRefRefQ 20.94 (0.70, 1.27)0.6950.95 (0.69, 1.32)0.7701.00 (0.68, 1.48)0.989Q 30.84 (0.59, 1.20)0.3340.85 (0.58, 1.25)0.3940.86 (0.53, 1.40)0.521Q 40.42 (0.28, 0.61)<0.0010.41 (0.28, 0.62)<0.0010.44 (0.24, 0.83)0.014P for trend<0.001<0.0010.008ZhangAge AccelContinuous0.98 (0.96, 1.00)0.0250.98 (0.96, 0.99)0.0060.98 (0.96, 1.01)0.221Q 1RefRefRefQ 21.08 (0.68, 1.71)0.7311.03 (0.65, 1.64)0.8871.11 (0.67, 1.84)0.672Q 31.10 (0.71, 1.69)0.6691.03 (0.65, 1.63)0.8881.17 (0.67, 2.04)0.553Q 40.70 (0.49, 1.02)0.0620.64 (0.44, 0.94)0.0230.79 (0.46, 1.33)0.347P for trend0.0240.0060.279LinAge AccelContinuous1.01 (0.99, 1.03)0.3401.01 (0.99, 1.03)0.4011.02 (0.99, 1.04)0.137Q 1RefRefRefQ 20.94 (0.63, 1.41)0.7580.93 (0.61, 1.42)0.7340.99 (0.62, 1.60)0.981Q 31.12 (0.78, 1.62)0.5211.12 (0.76, 1.65)0.5621.25 (0.82, 1.90)0.281Q 41.09 (0.71, 1.66)0.6871.06 (0.68, 1.65)0.7771.21 (0.71, 2.05)0.455P for trend0.5070.5820.312WeidnerAge AccelContinuous1.00 (0.99, 1.02)0.6081.00 (0.98, 1.02)0.7161.00 (0.98, 1.02)0.936Q 1RefRefRefQ 20.81 (0.55, 1.18)0.2560.80 (0.54, 1.18)0.2430.79 (0.51, 1.23)0.273Q 30.83 (0.52, 1.33)0.4340.83 (0.51, 1.34)0.4220.78 (0.46, 1.34)0.348Q 41.06 (0.73, 1.53)0.7491.04 (0.70, 1.53)0.8560.94 (0.60, 1.47)0.774P for trend0.6280.7200.926VidalBraloAge AccelContinuous1.00 (0.98, 1.02)0.9450.99 (0.97, 1.01)0.5160.99 (0.97, 1.01)0.380Q 1RefRefRefQ 21.01 (0.70, 1.46)0.9370.95 (0.64, 1.41)0.7820.94 (0.61, 1.45)0.769Q 31.12 (0.83, 1.52)0.4451.03 (0.74, 1.46)0.8361.00 (0.70, 1.43)0.985Q 40.97 (0.64, 1.47)0.8790.84 (0.53, 1.32)0.4310.80 (0.47, 1.36)0.382P for trend0.9900.4970.384EAA: epigenetic age acceleration. OR: odds ratio. CI: confidence interval."HorvathAge Accel" refers to epigenetic age acceleration calculated using the HorvathAge clock. The same applies to all other epigenetic clock acceleration metrics shown in this article.Fig. 4Restricted cubic spline analyses for associations between epigenetic age acceleration and OBS. A GrimAge Accel, B GrimAge2 Accel. Models were adjusted for age, gender, race, education, marital status, PIR, and energy intake. OR: Odds Ratio; CI: confidence interval Table 4Multivariate logistic regression analysis of the association between EAA and OBS.EAAOBSModel 1PModel 2PModel 3POR (95%CI)OR (95%CI)OR (95%CI)HorvathAge AccelContinuous0.99 (0.97, 1.01)0.1880.98 (0.96, 1.01)0.1311.00 (0.97, 1.02)0.690Q 1RefRefRefQ 20.91 (0.61, 1.37)0.6510.89 (0.59, 1.35)0.5790.98 (0.64, 1.50)0.937Q 30.75 (0.52, 1.08)0.1160.72 (0.49, 1.06)0.0960.85 (0.56, 1.28)0.417Q 40.85 (0.55, 1.32)0.4530.81 (0.52, 1.27)0.3471.04 (0.63, 1.73)0.870P for trend0.3240.2430.985HannumAge AccelContinuous0.98 (0.96, 1.01)0.1900.98 (0.96, 1.00)0.1121.00 (0.97, 1.03)0.975Q 1RefRefRefQ 21.01 (0.64, 1.60)0.9710.97 (0.61, 1.55)0.8841.10 (0.68, 1.79)0.669Q 30.87 (0.58, 1.32)0.5100.82 (0.53, 1.28)0.3651.02 (0.61, 1.69)0.941Q 40.79 (0.49, 1.28)0.3290.74 (0.46, 1.20)0.2101.06 (0.58, 1.95)0.836P for trend0.2130.1200.939SkinBloodAge AccelCategories0.98 (0.97, 1.00)0.0130.98 (0.97, 0.99)0.0080.99 (0.97, 1.02)0.582Q 1RefRefRefQ 21.16 (0.79, 1.69)0.4301.28 (0.83, 1.96)0.2451.15 (0.78, 1.69)0.471Q 31.20 (0.83, 1.73)0.3181.44 (0.92, 2.25)0.1021.18 (0.82, 1.71)0.355Q 40.78 (0.59, 1.04)0.0921.04 (0.62, 1.76)0.8680.77 (0.58, 1.02)0.071P for trend0.0950.0590.903PhenoAge AccelContinuous0.98 (0.96, 1.00)0.0200.98 (0.95, 1.00)0.0210.98 (0.96, 1.01)0.155Q 1RefRefRefQ 20.84 (0.56, 1.25)0.3780.83 (0.56, 1.24)0.3540.88 (0.58, 1.32)0.514Q 30.86 (0.52, 1.43)0.5510.85 (0.51, 1.43)0.5270.94 (0.54, 1.61)0.798Q 40.60 (0.41, 0.88)0.0110.59 (0.40, 0.88)0.0110.70 (0.46, 1.06)0.088P for trend0.0130.0150.118GrimAge AccelContinuous0.95 (0.94, 0.97)<0.0010.95 (0.93, 0.97)<0.0010.95 (0.92, 0.98)0.004Q 1RefRefRefQ 20.93 (0.70, 1.25)0.6410.93 (0.68, 1.26)0.6120.97 (0.66, 1.45)0.891Q 30.80 (0.56, 1.13)0.1920.78 (0.54, 1.14)0.1850.78 (0.48, 1.28)0.301Q 40.41 (0.27, 0.64)<0.0010.39 (0.24, 0.61)<0.0010.41 (0.20, 0.82)0.015P for trend<0.001<0.0010.009GrimAge2 AccelContinuous0.95 (0.94, 0.97)<0.0010.95 (0.94, 0.97)<0.0010.95 (0.93, 0.98)0.003Q 1RefRefRefQ 20.94 (0.70, 1.27)0.6950.95 (0.69, 1.32)0.7701.00 (0.68, 1.48)0.989Q 30.84 (0.59, 1.20)0.3340.85 (0.58, 1.25)0.3940.86 (0.53, 1.40)0.521Q 40.42 (0.28, 0.61)<0.0010.41 (0.28, 0.62)<0.0010.44 (0.24, 0.83)0.014P for trend<0.001<0.0010.008ZhangAge AccelContinuous0.98 (0.96, 1.00)0.0250.98 (0.96, 0.99)0.0060.98 (0.96, 1.01)0.221Q 1RefRefRefQ 21.08 (0.68, 1.71)0.7311.03 (0.65, 1.64)0.8871.11 (0.67, 1.84)0.672Q 31.10 (0.71, 1.69)0.6691.03 (0.65, 1.63)0.8881.17 (0.67, 2.04)0.553Q 40.70 (0.49, 1.02)0.0620.64 (0.44, 0.94)0.0230.79 (0.46, 1.33)0.347P for trend0.0240.0060.279LinAge AccelContinuous1.01 (0.99, 1.03)0.3401.01 (0.99, 1.03)0.4011.02 (0.99, 1.04)0.137Q 1RefRefRefQ 20.94 (0.63, 1.41)0.7580.93 (0.61, 1.42)0.7340.99 (0.62, 1.60)0.981Q 31.12 (0.78, 1.62)0.5211.12 (0.76, 1.65)0.5621.25 (0.82, 1.90)0.281Q 41.09 (0.71, 1.66)0.6871.06 (0.68, 1.65)0.7771.21 (0.71, 2.05)0.455P for trend0.5070.5820.312WeidnerAge AccelContinuous1.00 (0.99, 1.02)0.6081.00 (0.98, 1.02)0.7161.00 (0.98, 1.02)0.936Q 1RefRefRefQ 20.81 (0.55, 1.18)0.2560.80 (0.54, 1.18)0.2430.79 (0.51, 1.23)0.273Q 30.83 (0.52, 1.33)0.4340.83 (0.51, 1.34)0.4220.78 (0.46, 1.34)0.348Q 41.06 (0.73, 1.53)0.7491.04 (0.70, 1.53)0.8560.94 (0.60, 1.47)0.774P for trend0.6280.7200.926VidalBraloAge AccelContinuous1.00 (0.98, 1.02)0.9450.99 (0.97, 1.01)0.5160.99 (0.97, 1.01)0.380Q 1RefRefRefQ 21.01 (0.70, 1.46)0.9370.95 (0.64, 1.41)0.7820.94 (0.61, 1.45)0.769Q 31.12 (0.83, 1.52)0.4451.03 (0.74, 1.46)0.8361.00 (0.70, 1.43)0.985Q 40.97 (0.64, 1.47)0.8790.84 (0.53, 1.32)0.4310.80 (0.47, 1.36)0.382P for trend0.9900.4970.384EAA: epigenetic age acceleration. OR: odds ratio. CI: confidence interval."HorvathAge Accel" refers to epigenetic age acceleration calculated using the HorvathAge clock. The same applies to all other epigenetic clock acceleration metrics shown in this article. Multivariate logistic regression analysis of the association between EAA and OBS. EAA: epigenetic age acceleration. OR: odds ratio. CI: confidence interval. "HorvathAge Accel" refers to epigenetic age acceleration calculated using the HorvathAge clock. The same applies to all other epigenetic clock acceleration metrics shown in this article. Restricted cubic spline analyses for associations between epigenetic age acceleration and OBS. A GrimAge Accel, B GrimAge2 Accel. Models were adjusted for age, gender, race, education, marital status, PIR, and energy intake. OR: Odds Ratio; CI: confidence interval Subgroup analysis Subgroup analyses and interaction tests were performed to evaluate the consistency of the association between EAA and OBS across different populations. The results indicated that in most subgroups, the association between OBS and EAA remained statistically significant and inverse (Fig. 5). No significant interaction was observed between OBS and any of the predefined subgroups (all P for interaction > 0.05).Fig. 5Forest plot of stratified analysis of the associations between OBS and epigenetic age acceleration. A GrimAge Accel, B GrimAge2 Accel. Models were adjusted for age, gender, race, education, marital status, PIR, and energy intake. Forest plot of stratified analysis of the associations between OBS and epigenetic age acceleration. A GrimAge Accel, B GrimAge2 Accel. Models were adjusted for age, gender, race, education, marital status, PIR, and energy intake. Sensitivity analysis Sensitivity analyses were performed to assess the associations of the dietary OBS and lifestyle OBS with epigenetic age. Whether dietary OBS is treated as a continuous or categorical variable, the negative correlation between dietary OBS and epigenetic age remains significant. Similarly, there were significant negative associations found between lifestyle OBS and epigenetic age. More detailed results are shown in Table 5.Table 5Results of multivariate linear regression analysis of the association between epigenetic age and dietary/ lifestyle OBS.Epigenetic ageOBSModel 1PModel 2PModel 3Pβ (95%CI)β (95%CI)β (95%CI)GrimAgeDietaryContinuous −0.19 ( −0.27, −0.11)<0.001 −0.15 ( −0.21, −0.10)<0.001 −0.08 ( −0.15, −0.01)0.021Q 1RefRefRefQ 2 −2.06 ( −4.21, 0.09)0.060 −1.48 ( −2.74, −0.21)0.024 −0.95 ( −2.30, 0.41)0.157Q 3 −1.61 ( −3.54, 0.32)0.099 −1.15 ( −2.42, 0.12)0.074 −0.18 ( −1.64, 1.28)0.797Q 4 −3.58 ( −5.20, −1.97)<0.001 −2.85 ( −3.90, −1.80)<0.001 −1.43 ( −2.70, −0.15)0.031P for trend<0.001<0.0010.035GrimAgeLifestyleContinuous −0.39 ( −0.77, −0.01)0.047 −0.79 ( −1.04, −0.54)<0.001 −0.68 ( −0.95, −0.42)<0.001Q 1RefRefRefQ 2 −0.62 ( −2.34, 1.10)0.465 −2.45 ( −3.74, −1.16)<0.001 −2.35 ( −3.50, −1.20)<0.001Q 3 −1.68 ( −3.56, 0.20)0.078 −2.99 ( −4.14, −1.85)<0.001 −2.73 ( −3.91, −1.56)<0.001Q 4 −1.03 ( −2.82, 0.77)0.252 −3.39 ( −4.60, −2.17)<0.001 −2.90 ( −4.18, −1.63)<0.001P for trend0.203<0.001<0.001GrimAge2DietaryContinuous −0.20 ( −0.27, −0.12)<0.001 −0.16 ( −0.21, −0.10)<0.001 −0.09 ( −0.16, −0.01)0.025Q 1RefRefRefQ 2 −2.22 ( −4.41, −0.03)0.047 −1.63 ( −3.02, −0.25)0.023 −1.10 ( −2.58, 0.37)0.133Q 3 −1.80 ( −3.76, 0.15)0.069 −1.29 ( −2.65, 0.06)0.060 −0.32 ( −1.95, 1.30)0.677Q 4 −3.74 ( −5.34, −2.14)<0.001 −2.95 ( −4.05, −1.84)<0.001 −1.51 ( −2.94, −0.08)0.039P for trend<0.001<0.0010.049GrimAge2LifestyleContinuous −0.55 ( −0.93, −0.18)0.006 −0.95 ( −1.21, −0.70)<0.001 −0.83 ( −1.09, −0.56)<0.001Q 1RefRefRefQ 2 −0.86 ( −2.59, 0.86)0.314 −2.63 ( −3.97, −1.28)<0.001 −2.49 ( −3.70, −1.29)<0.001Q 3 −2.12 ( −3.99, −0.24)0.028 −3.36 ( −4.53, −2.19)<0.001 −3.07 ( −4.27, −1.87)<0.001Q 4 −1.75 ( −3.53, 0.03)0.054 −4.04 ( −5.25, −2.84)<0.001 −3.49 ( −4.72, −2.26)<0.001P for trend0.036<0.001<0.001HorvathTeloDietaryContinuous0.01 (0.00, 0.01)0.0030.01 (0.00, 0.01)<0.0010.00 (0.00, 0.01)0.048Q 1Q 20.08 (0.01, 0.15)0.0260.08 (0.02, 0.13)0.0110.06 ( −0.00, 0.13)0.060Q 30.08 (0.02, 0.15)0.0160.09 (0.04, 0.13)0.0010.06 (0.01, 0.12)0.033Q 40.11 (0.04, 0.18)0.0040.11 (0.06, 0.16)<0.0010.08 (0.00, 0.15)0.048P for trend0.003<0.0010.046HorvathTeloLifestyleContinuous0.00 ( −0.01, 0.01)0.7010.01 (0.00, 0.03)0.0060.01 (0.00, 0.02)0.019Q 1RefRefRefQ 20.02 ( −0.04, 0.07)0.6110.07 (0.02, 0.12)0.0120.07 (0.02, 0.12)0.012Q 30.02 ( −0.04, 0.09)0.4980.06 (0.01, 0.12)0.0300.06 (0.00, 0.11)0.036Q 40.00 ( −0.05, 0.06)0.9940.07 (0.02, 0.13)0.0080.06 (0.01, 0.11)0.024P for trendDunedinPoAmDietaryContinuous −0.00 ( −0.00, −0.00)<0.001 −0.00 ( −0.00, −0.00)<0.001 −0.00 ( −0.00, −0.00)0.047Q 1RefRefRefQ 2 −0.02 ( −0.04, −0.00)0.045 −0.02 ( −0.04, 0.00)0.076 −0.01 ( −0.03, 0.01)0.219Q 3 −0.01 ( −0.03, 0.02)0.574 −0.01 ( −0.03, 0.02)0.6430.00 ( −0.03, 0.03)0.874Q 4 −0.03 ( −0.05, −0.02)<0.001 −0.03 ( −0.05, −0.01)0.001 −0.02 ( −0.04, 0.00)0.092P for trend<0.0010.0020.120DunedinPoAmLifestyleContinuous −0.02 ( −0.02, −0.01)<0.001 −0.02 ( −0.02, −0.01)<0.001 −0.02 ( −0.02, −0.01)<0.001Q 1RefRefRefQ 2 −0.03 ( −0.06, −0.01)0.007 −0.04 ( −0.06, −0.01)0.005 −0.04 ( −0.06, −0.01)0.005Q 3 −0.04 ( −0.07, −0.02)<0.001 −0.05 ( −0.07, −0.02)<0.001 −0.05 ( −0.07, −0.02)<0.001Q 4 −0.07 ( −0.09, −0.05)<0.001 −0.07 ( −0.09, −0.05)<0.001 −0.07 ( −0.09, −0.05)<0.001P for trend< 0.001< 0.001< 0.001β: beta coefficient; CI: confidence interval.Model 1 was unadjusted,Model 2 was adjusted for age, gender, and race,Model 3 was adjusted for age, gender, race, education, marital status, PIR, and energy intake. Table 5Results of multivariate linear regression analysis of the association between epigenetic age and dietary/ lifestyle OBS.Epigenetic ageOBSModel 1PModel 2PModel 3Pβ (95%CI)β (95%CI)β (95%CI)GrimAgeDietaryContinuous −0.19 ( −0.27, −0.11)<0.001 −0.15 ( −0.21, −0.10)<0.001 −0.08 ( −0.15, −0.01)0.021Q 1RefRefRefQ 2 −2.06 ( −4.21, 0.09)0.060 −1.48 ( −2.74, −0.21)0.024 −0.95 ( −2.30, 0.41)0.157Q 3 −1.61 ( −3.54, 0.32)0.099 −1.15 ( −2.42, 0.12)0.074 −0.18 ( −1.64, 1.28)0.797Q 4 −3.58 ( −5.20, −1.97)<0.001 −2.85 ( −3.90, −1.80)<0.001 −1.43 ( −2.70, −0.15)0.031P for trend<0.001<0.0010.035GrimAgeLifestyleContinuous −0.39 ( −0.77, −0.01)0.047 −0.79 ( −1.04, −0.54)<0.001 −0.68 ( −0.95, −0.42)<0.001Q 1RefRefRefQ 2 −0.62 ( −2.34, 1.10)0.465 −2.45 ( −3.74, −1.16)<0.001 −2.35 ( −3.50, −1.20)<0.001Q 3 −1.68 ( −3.56, 0.20)0.078 −2.99 ( −4.14, −1.85)<0.001 −2.73 ( −3.91, −1.56)<0.001Q 4 −1.03 ( −2.82, 0.77)0.252 −3.39 ( −4.60, −2.17)<0.001 −2.90 ( −4.18, −1.63)<0.001P for trend0.203<0.001<0.001GrimAge2DietaryContinuous −0.20 ( −0.27, −0.12)<0.001 −0.16 ( −0.21, −0.10)<0.001 −0.09 ( −0.16, −0.01)0.025Q 1RefRefRefQ 2 −2.22 ( −4.41, −0.03)0.047 −1.63 ( −3.02, −0.25)0.023 −1.10 ( −2.58, 0.37)0.133Q 3 −1.80 ( −3.76, 0.15)0.069 −1.29 ( −2.65, 0.06)0.060 −0.32 ( −1.95, 1.30)0.677Q 4 −3.74 ( −5.34, −2.14)<0.001 −2.95 ( −4.05, −1.84)<0.001 −1.51 ( −2.94, −0.08)0.039P for trend<0.001<0.0010.049GrimAge2LifestyleContinuous −0.55 ( −0.93, −0.18)0.006 −0.95 ( −1.21, −0.70)<0.001 −0.83 ( −1.09, −0.56)<0.001Q 1RefRefRefQ 2 −0.86 ( −2.59, 0.86)0.314 −2.63 ( −3.97, −1.28)<0.001 −2.49 ( −3.70, −1.29)<0.001Q 3 −2.12 ( −3.99, −0.24)0.028 −3.36 ( −4.53, −2.19)<0.001 −3.07 ( −4.27, −1.87)<0.001Q 4 −1.75 ( −3.53, 0.03)0.054 −4.04 ( −5.25, −2.84)<0.001 −3.49 ( −4.72, −2.26)<0.001P for trend0.036<0.001<0.001HorvathTeloDietaryContinuous0.01 (0.00, 0.01)0.0030.01 (0.00, 0.01)<0.0010.00 (0.00, 0.01)0.048Q 1Q 20.08 (0.01, 0.15)0.0260.08 (0.02, 0.13)0.0110.06 ( −0.00, 0.13)0.060Q 30.08 (0.02, 0.15)0.0160.09 (0.04, 0.13)0.0010.06 (0.01, 0.12)0.033Q 40.11 (0.04, 0.18)0.0040.11 (0.06, 0.16)<0.0010.08 (0.00, 0.15)0.048P for trend0.003<0.0010.046HorvathTeloLifestyleContinuous0.00 ( −0.01, 0.01)0.7010.01 (0.00, 0.03)0.0060.01 (0.00, 0.02)0.019Q 1RefRefRefQ 20.02 ( −0.04, 0.07)0.6110.07 (0.02, 0.12)0.0120.07 (0.02, 0.12)0.012Q 30.02 ( −0.04, 0.09)0.4980.06 (0.01, 0.12)0.0300.06 (0.00, 0.11)0.036Q 40.00 ( −0.05, 0.06)0.9940.07 (0.02, 0.13)0.0080.06 (0.01, 0.11)0.024P for trendDunedinPoAmDietaryContinuous −0.00 ( −0.00, −0.00)<0.001 −0.00 ( −0.00, −0.00)<0.001 −0.00 ( −0.00, −0.00)0.047Q 1RefRefRefQ 2 −0.02 ( −0.04, −0.00)0.045 −0.02 ( −0.04, 0.00)0.076 −0.01 ( −0.03, 0.01)0.219Q 3 −0.01 ( −0.03, 0.02)0.574 −0.01 ( −0.03, 0.02)0.6430.00 ( −0.03, 0.03)0.874Q 4 −0.03 ( −0.05, −0.02)<0.001 −0.03 ( −0.05, −0.01)0.001 −0.02 ( −0.04, 0.00)0.092P for trend<0.0010.0020.120DunedinPoAmLifestyleContinuous −0.02 ( −0.02, −0.01)<0.001 −0.02 ( −0.02, −0.01)<0.001 −0.02 ( −0.02, −0.01)<0.001Q 1RefRefRefQ 2 −0.03 ( −0.06, −0.01)0.007 −0.04 ( −0.06, −0.01)0.005 −0.04 ( −0.06, −0.01)0.005Q 3 −0.04 ( −0.07, −0.02)<0.001 −0.05 ( −0.07, −0.02)<0.001 −0.05 ( −0.07, −0.02)<0.001Q 4 −0.07 ( −0.09, −0.05)<0.001 −0.07 ( −0.09, −0.05)<0.001 −0.07 ( −0.09, −0.05)<0.001P for trend< 0.001< 0.001< 0.001β: beta coefficient; CI: confidence interval.Model 1 was unadjusted,Model 2 was adjusted for age, gender, and race,Model 3 was adjusted for age, gender, race, education, marital status, PIR, and energy intake. Results of multivariate linear regression analysis of the association between epigenetic age and dietary/ lifestyle OBS. β: beta coefficient; CI: confidence interval. Model 1 was unadjusted, Model 2 was adjusted for age, gender, and race, Model 3 was adjusted for age, gender, race, education, marital status, PIR, and energy intake. Discussion To our knowledge, this study is the first to assess the relationship between OBS and epigenetic clocks based on a relatively large sample from NHANES. We found significant inverse associations between OBS levels and both epigenetic age and accelerate biological aging. RCS analysis revealed a negative dose-response relationship between OBS and accelerated biological aging as measured by GrimAge and GrimAge2. The association between OBS and biological aging acceleration was homogeneous across subgroups. Sensitivity analysis showed that both dietary OBS and lifestyle OBS were negatively correlated with epigenetic age. These findings indicate that higher OBS levels are associated with reduced epigenetic age and decelerated aging. This evidence underscores that maintaining a high level of antioxidative status through diet and lifestyle may represent a viable strategy for anti-aging interventions. We conducted a comprehensive study on the relationship between OBS and 12 different epigenetic clocks among middle-aged and older adults in the U.S. Our results indicate that GrimAge, GrimAge2, HorvathTelo, and DunedinPoAm exhibit stronger associations with OBS compared with other clocks. Epigenetic clock estimates biological age by applying machine learning algorithms to DNA methylation patterns at specific cytosine-guanine dinucleotides (CpG) sites [16–27]. However, heterogeneity in the composition of epigenetic clocks including selected CpGs and clinical biomarkers contributed to different association strengths with outcomes. The development of first generation epigenetic clocks was primarily based on DNA methylation data of selected CpGs, such as Hannum clock (71 CpGs) and Horvath clock (353 CpGs) [16, 17]. In contrast, the second generation clocks incorporated not only more CpGs but also clinical biomarkers. For example, Levine’s clock PhenoAge incorporates 513 CpGs, chronological age, and a set of biomarkers strongly associated with mortality [19]. The Horvath group’s GrimAge is a composite biomarker based on seven DNA methylation-based estimators of plasma proteins and a DNA methylation -based estimator of smoking pack years [20]. DunedinPoAm, a novel algorithm developed using aging data and whole-genome methylation data to predict the rate of aging, may actually be the first of the third generation of clocks [27, 31]. Previous study found that the second generation epigenetic clocks were more strongly associated with an increased risk of both arsenic-related cardiovascular disease incidence and mortality than the first generation epigenetic clocks [32]. Consistently, our results suggest that the second generation epigenetic clocks are more effective in capturing the impact of oxidative stress on aging. Being the first study addressing OBS and epigenetic clocks, we can only compare our results with previous research on oxidative stress and other biological ages. Phenotypic age is a novel indicator of biological aging, calculated based on a combination of chronological age and nine clinical chemistry biomarkers [33]. Wu et al. reported that higher OBS levels are significantly associated with slower phenotypic age acceleration, suggesting that higher levels of OBS may be associated with a slower biological aging process, which is consistent with our findings [34]. In addition, our subgroup analyses demonstrated a consistent association between OBS and biological aging measured by GrimAge and GrimAge2 across all subgroups. This suggests that populations with varying demographic characteristics may still delay aging by enhancing their OBS levels. Telomere length, a well-established indicator of biological aging, exhibits a negative association with the aging process. Zhang et al. reported a positive association between OBS and telomere length measured by qPCR (qPCRTL) in females but not males [29]. In our study, OBS also showed positive associations with DNA methylation-based telomere length (DNAmTL) in all models. Previous study demonstrated that DNAmTL exhibited a stronger correlation with chronological age compared to qPCRTL (r = −0.58 vs. r = −0.25) [28]. Additionally, ROC analysis showed that DNAmTL was superior to qPCRTL for predicting long-term all-cause mortality (AUC: 0.74 vs. 0.62, P < 0.001) and CVD mortality (AUC: 0.75 vs. 0.64, P < 0.001). Therefore, compared with qPCRTL, DNAmTL may be a more valuable biomarker for aging. The association between OBS and epigenetic clocks may be explained by biological mechanisms involving oxidative stress and DNA methylation. DNA methylation is a significant epigenetic alteration during the aging process, with CpG being the primary targets of DNA methylation modifications in mammals [35]. Mounting evidence indicates that oxidative stress dynamically regulates methylation/demethylation patterns of genes and their expression. O'Hagan et al. discovered that oxidative damage induces the relocalization of DNMT1, DNMT3B, SIRT1, and EZH2 from low GC-content regions togene promoter regions containing CpG islands, forming a large multi-protein complex(es) that promotes DNA methylation [36]. Oxidative stress induced aberrant DNA methylation patterns through DNA strand breakage and damage, while also regulating DNA demethylation by DNA oxidation or TET-mediated hydroxymethylation [37, 38]. In addition, oxidative stress promote DNA methyltransferase 1 (DNMT) expression, which leads to site-specific hypermethylation in neurodegeneration [39]. A recently research showed that aging and oxidative stress lead to aberrant methylation of the CpG island in the LEDGF promoter, resulting in the silencing of LEDGF [40]. Additionally, chronic inflammation is also one of the Hallmarks of aging, and oxidative stress can lead to age-related tissue damage by promoting inflammation [41]. Considering the critical role of oxidative stress in modulating DNA methylation, the OBS, as a comprehensive measure of oxidative homeostasis, may capture the DNA methylation modifications underlying the construction of epigenetic clocks. Subsequent research needs to s further investigate aging-related gene loci and their methylation patterns, particularly focusing on their regulatory mechanisms by oxidative stress. Diet and lifestyle factors represent practical interventions for modulating aging processes, with accumulating evidence linking them to epigenetic aging. For example, higher dietary intake of creatine was associated with reduced biological age acceleration and mortality risk as estimated by epigenetic biomarkers [42]. Inadequate intake of specific nutrients, including vitamin B12 and vitamin C, has been associated with accelerated epigenetic aging [43]. Exercise training has been shown to shift epigenetic and transcriptomic patterns toward a younger profile in human skeletal muscle, while muscle disuse accelerates aging-related changes [44]. A 6-month endurance exercise training intervention demonstrated that GrimAge decreased by 7.44 months relative to the expected trajectory, and this change was correlated with improvements in cardiorespiratory fitness [45]. In a thymic regeneration study, the average epigenetic age was approximately 1.5 years lower than baseline after one year of treatment, along with improvements in risk indices for many age-related diseases [46]. A retrospective analysis reported that 7-month supplementation with α-ketoglutarate turned back epigenetic age by 8 years [47]. Wang et al. reported that an antiinflammatory and/or antioxidant diet may mitigate biological aging [48]. In the present study, further analyses showed that both the dietary OBS and the lifestyle OBS were significantly negatively correlated with epigenetic age measured by four epigenetic clocks. Our findings suggest that higher antioxidant dietary and lifestyle exposures are associated with attenuated biological aging, consistent with the existing literature. In mice, the overexpression of antioxidant enzymes can reduce ROS production and protect DNA from its harmful effects, which is associated with increased lifespan [49]. Moreover, it is reported that increased intake of dietary antioxidants, such as vitamin E, are associated with reduced risk of atherosclerotic diseases and Alzheimer's disease [10]. Bouzid et al. reported that regular physical activity can enhance antioxidant defenses and reduce lipid peroxidation levels in both young and elderly individuals, highlighting the importance of regular exercise in slowing down aging-related damage [50]. To our knowledge, no prior study has specifically examined the association between the oxidative balance score and multiple epigenetic aging markers. Given the results of this study, we recommend that future research on aging consider using epigenetic clocks as efficacy indicators for antioxidant-based interventions. This study has several strengths. First, the complex sampling design used in NHANES makes our findings generalizable to the U.S. population. Second, by using both the OBS and epigenetic age as indicators, we can more intuitively understand the relationship between an individual’s oxidative stress levels and aging. Third, we conducted rigorous and comprehensive statistical analyses, demonstrating the robustness of our key findings. There are also several limitations to this study. First, the cross-sectional design limits the ability to draw causal inferences between OBS and epigenetic age. Second, the dietary and lifestyle data used to calculate OBS come from recall interviews, which may introduce recall bias. Third, although OBS is a comprehensive indicator of oxidative stress, certain antioxidant components such as flavonoids and resveratrol were not included in its calculation. Fourth, despite identifying a linear dose-response relationship between OBS and biological aging through RCS analysis, further clinical validation is needed to confirm the impact of interventions targeting OBS on biological aging. Fifth, despite adjusting for multiple covariates, residual confounding due to unmeasured factors (e.g., genetic background, medication use) cannot be completely excluded. Conclusion In conclusion, our study is the first to reveal a negative correlation between OBS and biological aging measured by multiple epigenetic clocks. Antioxidant-rich diets and lifestyles may reduce epigenetic age and delay biological aging, emphasizing the importance of antioxidants in anti-aging strategies. These findings need to be further validated in prospective studies with larger samples and different populations in the future. Electronic Supplementary Material Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1