Abstract
Oxidative balance score (OBS) integrates dietary/lifestyle factors to assess cumulative oxidative stress, yet its association with thyroid hormone sensitivity remains unexplored. This study aimed to investigate the relationship between OBS and thyroid hormone sensitivity through central and peripheral regulation mechanisms. This cross-sectional study analyzed data from the National Health and Nutrition Examination Survey (2007–2012), involving 5536 participants. OBS was computed using a tertile-based scoring system, with higher scores indicating a predominance of antioxidant exposures. Central sensitivity was evaluated via thyroid-stimulating hormone index, thyrotroph thyroxine resistance index, and thyroid feedback quantile-based index; peripheral sensitivity was assessed by the free triiodothyronine to free thyroxine ratio. Weighted multivariable linear regression and restricted cubic splines were used to characterize dose–response relationships, supported by subgroup and sensitivity analyses. Threshold-dependent associations emerged: central sensitivity indices (thyroid-stimulating hormone index, thyrotroph thyroxine resistance index, and thyroid feedback quantile-based index) showed a positive linear association with OBS below inflection points (8.275, 8.275, and 8.847, respectively), plateauing above these points. Peripheral sensitivity (free triiodothyronine/free thyroxine ratio) was negatively associated with OBS scores above 15.681 (β = −0.013, P = .001). Subgroup analyses showed consistent patterns across demographics (all P for interaction > .05), and sensitivity analyses confirmed the robustness of the findings. Our findings suggest a dual redox-thyroid regulatory mechanism, wherein both prooxidant dominance and antioxidant excess may be associated with impaired thyroid hormone sensitivity through central and peripheral pathways, respectively. These results underscore the potential importance of maintaining an optimal oxidative balance zone for thyroid health and provide a composite framework for future research into personalized diet and lifestyle interventions.
Keywords: NHANES, oxidative balance score, oxidative stress, sensitivity to thyroid hormones, thyroid hormone action
1. Introduction
Thyroid hormones (THs) are fundamental regulators of systemic metabolism, cardiovascular function, and neurodevelopment.[1] Classical physiology posits that TH homeostasis is maintained by a negative feedback loop between the pituitary and thyroid gland. However, emerging epidemiological evidence reveals a non-negligible prevalence of individuals with concurrent elevations in both thyroid-stimulating hormone (TSH) and TH levels.[2] This discordance challenges the conventional feedback model and implies the heterogeneity in tissue sensitivity to THs.
This paradox has spurred the conceptual framework of thyroid hormone sensitivity, which provides a more dynamic assessment of hormonal responsiveness beyond static hormone concentrations.[3] The framework distinguishes 2 regulatory dimensions: central sensitivity, quantified by indices like the thyroid-stimulating hormone index (TSHI), thyrotroph thyroxine resistance index (TT4RI), and thyroid feedback quantile-based index (TFQI), which reflect the set-point of the hypothalamic–pituitary–thyroid (HPT) axis; and peripheral sensitivity, gauged by the free triiodothyronine/free thyroxine ratio (FT3/FT4), which serves as a proxy for the efficiency of extrathyroidal thyroxine-to-triiodothyronine (T4-to-T3) conversion.[4,5] This dual-axis framework resolves discordances between circulating THs levels and tissue responses, providing mechanistic insights beyond conventional hormone measurements.[6] Impairments in these sensitivity indices are clinically significant, being linked to a higher risk of metabolic disorders, cardiovascular disease, and all-cause mortality.[7,8] However, the systemic factors that modulate this sensitivity framework remain largely unexplored.
Among potential modulators, oxidative stress, defined as an imbalance between the production of reactive oxygen species and the body’s antioxidant defenses,[9] is implicated in the pathogenesis of various thyroid disorders.[10–12] Experimental studies have shown that redox status can influence thyroid hormone receptor function and deiodinase activity.[13,14] However, existing research has predominantly focused on isolated factors, such as smoking, specific nutrients (e.g., selenium, iron), or alcohol intake,[15–17] neglecting the complex, cumulative redox environment in vivo where these factors interact synergistically or antagonistically.
To address this limitation, the oxidative balance score (OBS) was developed as a validated composite metric that holistically integrates multiple dietary and lifestyle components based on their pro- or antioxidant properties.[18] A higher OBS indicates a predominance of antioxidant exposures[18] and has been validated against biomarkers of oxidative stress and inflammation, such as F2-isoprostanes, C-reactive protein, and gamma-glutamyl transferase.[19–21] Its clinical utility is further evidenced by population studies linking a higher OBS to a graded reduction in the risk of several chronic conditions, including kidney stone,[22] cardiovascular disease,[23] and metabolic syndrome.[24] However, its specific role in regulating central and peripheral TH sensitivity remains unclear. Therefore, leveraging data from the National Health and Nutrition Examination Survey (NHANES), this study aimed to investigate the associations between OBS and indices of both central and peripheral thyroid hormone sensitivity in a nationally representative adult population.
2. Methods
2.1. Data source and study population
NHANES, administered by the National Center for Health Statistics, is a nationwide research program designed to systematically track the health and nutritional profiles of the noninstitutionalized U.S. civilian population biannually. Employing a stratified, multistage probability cluster sampling framework with oversampling of minority subgroups, NHANES integrates 3 core data modalities: in-person demographic and health-related questionnaires; standardized physical examinations and biospecimen collection through mobile examination centers; and comprehensive laboratory analyses spanning clinical chemistry, hematology, and nutritional biomarkers.
This cross-sectional study analyzed 3 consecutive NHANES cycles (2007–2012) selected for comprehensive thyroid data availability. Initially, a total of 30,442 participants were included. Exclusions were made for the following criteria: missing thyroid indicators (n = 19,998), incomplete OBS components (n = 3028), or unavailable covariates information (n = 421). Additional exclusions applied to pregnant women (n = 73), individuals with a history of thyroid disease (n = 499) or thyroid cancer (n = 3), and those using thyroid-related medications,[25] including levothyroxine, liothyronine, liotrix, or β-blockers (n = 56). To ensure physiological plausibility, participants with extreme energy intake (<500 or >5000 kcal/d for women; <500 or >8000 kcal/d for men; n = 33)[26] and those exhibiting elevated thyroid autoantibodies (thyroid peroxidase antibodies > 9 IU/mL [n = 602] or thyroglobulin antibodies > 4 IU/mL [n = 193]).[27,28] The final analytical cohort thus comprised 5536 adults aged ≥ 18 years with complete data for all variables of interest, constituting a complete-case analysis (Fig. 1). All participants provided written informed consent under National Center for Health Statistics Institutional Review Board protocols. The present study is a secondary analysis of publicly available, de-identified data and therefore was deemed exempt from additional review at our institution.
Figure 1.
A detailed flowchart of participant recruitment.
2.2. Exposure: oxidative balance scores
The OBS was constructed by integrating 16 dietary nutrients and 4 lifestyle factors, comprising 15 antioxidant and 5 prooxidant exposures.[18] Dietary data for the nutrient components were derived from the average of two 24-hour recall interviews, conducted using NHANES’s computer-assisted dietary interview system and quantified with the University of Texas Food Intake Analysis System Nutrients Database. Serum cotinine levels quantified both active and passive smoke exposure. Certified health technicians recorded anthropometric measurements (height and weight) for body mass index calculation. Physical activity was assessed via the Global Physical Activity Questionnaire and converted to metabolic equivalent (MET)-minutes/week. Components were scored using gender-stratified tertiles. Antioxidants were assigned 2, 1, and 0 points from the highest to lowest tertile, while prooxidants were inversely scored (0, 1, and 2).[18] Cut-points for alcohol intake[29] and physical activity[30] were adopted from established OBS scoring methodologies used in prior research to ensure comparability across studies: alcohol intake was categorized as nondrinker (2 points), moderate (men ≤ 30 g/d; women ≤ 15 g/d; 1 point), or heavy drinker (0 points); physical activity was stratified as low (<400 MET-min/wk; 0 points), moderate (400–1000 MET-min/wk; 1 point), or high (>1000 MET-min/wk; 2 points). The composite OBS was derived by summing all component scores, with a higher total score indicating a greater predominance of antioxidant exposures. Detailed scoring criteria are presented in Table 1.
Table 1.
Scheme for assigning oxidative balance scores based on data from 5536 participants.
| OBS | Men | Women | ||||
|---|---|---|---|---|---|---|
| 0 | 1 | 2 | 0 | 1 | 2 | |
| Dietary OBS components | ||||||
| *Dietary fiber (g/d) | <12.75 | 12.75–20.25 | >20.25 | <10.80 | 10.80–16.20 | >16.20 |
| *β-carotene (mcg/d) | <549.33 | 549.33–1915.00 | >1915.00 | <549.17 | 547.19–1932.16 | >1932.16 |
| *Riboflavin (Vitamin B2) (mg/d) | <1.74 | 1.74–2.56 | >2.56 | <1.37 | 1.37–1.99 | >1.99 |
| *Niacin (mg/d) | <21.58 | 21.58–31.54 | >31.54 | <15.69 | 15.69–22.75 | >22.75 |
| *Vitamin B6 (mg/d) | <1.67 | 1.67–2.51 | >2.51 | <1.24 | 1.24–1.83 | >1.83 |
| *Total folate (mcg/d) | <323.00 | 323.00–487.00 | >487.00 | <256.50 | 256.50–380.00 | >380.00 |
| *Vitamin B12 (mcg/d) | <3.63 | 3.63–6.22 | >6.22 | <2.69 | 2.69–4.63 | >4.63 |
| *Vitamin C (mg/d) | <40.17 | 40.17–102.58 | >102.58 | <38.10 | 38.10–88.23 | >88.23 |
| *Vitamin E as alpha-tocopherol (mg/d) | <5.46 | 5.46–8.75 | >8.75 | <4.34 | 4.34–7.13 | >7.13 |
| *Calcium (mg/d) | <702.50 | 702.50–1092.83 | >1092.83 | <598.67 | 598.83–900.67 | >900.67 |
| *Magnesium (mg) | <250.00 | 250.00–354 | >354.00 | <201.33 | 201.33–279.50 | >279.50 |
| *Zinc (mg/d) | <9.70 | 9.70–14.44 | >14.44 | <7.23 | 7.23–10.34 | >10.34 |
| *Copper (mg/d) | <1.06 | 1.06–1.51 | >1.51 | <0.86 | 0.86–1.21 | >1.21 |
| *Selenium (mcg/d) | <99.03 | 99.03–140.95 | >140.95 | <72.70 | 70.70–101.78 | >101.78 |
| †Total fat (g/d) | >98.62 | 64.96–98.62 | <64.96 | >73.04 | 48.76–73.04 | <48.76 |
| †Iron (mg/d) | >18.29 | 12.40–18.29 | <12.40 | >14.07 | 9.81–14.07 | <9.81 |
| Lifestyle OBS components | ||||||
| *Leisure time physical activity (MET-minute/week) | <400 | 400–1000 | >1000 | <400 | 400–1000 | >1000 |
| †Alcohol (g/d) | >30 | 0–30 | 0 | >15 | 0–15 | 0 |
| †Cotinine (ng/mL) | >2.86 | 0.03–2.86 | <0.03 | >0.19 | 0.02–0.19 | <0.02 |
| †Body mass index (kg/m2) | >30.20 | 25.87–30.20 | <25.87 | >31.37 | 25.34–31.37 | <25.34 |
The dietary components did not include nutrients obtained from dietary supplements or medications.
ATE = alpha-tocopherol equivalent, MET = metabolic equivalent, OBS = oxidative balance score.
Antioxidant.
Prooxidant.
2.3. Outcomes: indices of thyroid hormone sensitivity
Thyroid hormones (TSH, FT3, and FT4) were measured at the University of Washington using standardized immunoenzymatic assays. Validated indices were employed to assess thyroid hormone sensitivity. Central sensitivity, reflecting HPT axis feedback, was quantified by: TFQI, where positive values indicate resistance; TSHI; and the TT4RI, with higher values of both TSHI and TT4RI denoting reduced central sensitivity. Peripheral sensitivity, indicative of hormone conversion efficiency, was estimated by the FT3/FT4 ratio, where a lower ratio suggests impaired conversion. The detailed calculation formulas are provided in File S1, Supplemental Digital Content, https://links.lww.com/MD/R688.
2.4. Covariates
Covariates were selected a priori based on their established or potential role as confounders in the relationship between oxidative balance and thyroid sensitivity, as evidenced by previous literature.[26,30] These included: age (years), gender (men or women), race/ethnicity (non-Hispanic Black, non-Hispanic White, Mexican American, and other races), educational level (under high school, high school or equivalent, and college graduate or above), poverty-income ratio (PIR; <1.3, 1.3–3.5, and >3.5), serum 25-hydroxyvitamin D (25(OH)D) (insufficient: <75 nmol/L; sufficient: ≥75 nmol/L), and total energy intake (kcal/d).
2.5. Statistical analysis
All analyses adhered to NHANES analytical guidelines, incorporating sample weights, strata, and primary sampling units to account for the complex survey design and ensure national representativeness. Continuous variables were assessed for normality following Box-Cox transformation and presented as weighted mean ± standard deviation or median (interquartile range), as appropriate. Categorical variables are summarized as unweighted counts (weighted percentages). The OBS was analyzed both as a continuous variable and categorized into sex-specific quartiles (Q1–Q4). Differences in baseline characteristics across OBS quartiles were evaluated using analysis of variance, Rao-Scott χ2 tests, or Kruskal–Wallis tests, depending on the variable distribution.
The association between OBS and indices of thyroid hormone sensitivity was examined using 3 sequential weighted multivariable linear regression models: Model I was unadjusted; Model II was adjusted for age, gender, race/ethnicity, educational level, and the PIR; and Model III was further adjusted for serum 25(OH)D level and total energy intake. Potential nonlinear dose–response relationships were explored using weighted restricted cubic splines (RCS) with 4 knots placed at the 5th, 35th, 65th, and 95th percentiles. The likelihood ratio test was used to assess overall nonlinearity, and inflection points were identified via recursive partitioning. Subgroup analyses were performed to evaluate the influence of age (20–39, 40–59, ≥60), gender (men/women), race (non-Hispanic Black, non-Hispanic White, Mexican American, and other races), educational level (under high school, high school or equivalent, and college graduate or above), PIR (<1.3, 1.3–3.5, >3.5), and serum 25(OH)D level (insufficient and sufficient). Interaction tests were used to examine the presence of significant interactions of these stratum variables with the association between OBS and indices of thyroid hormone sensitivity. Additionally, sensitivity analyses were performed by recategorizing the OBS into tertiles to assess the robustness of the primary findings. All statistical analyses were performed using R software (version 4.2.3; R Foundation for Statistical Computing, Vienna, Austria). A two-tailed P-value <.05 was considered statistically significant.
3. Results
3.1. Baseline characteristics of the study participants
The analytical cohort comprised 5536 adults (mean age 45.489 ± 16.351 years; 46.628% women) stratified into OBS quartiles (Table 2). Participants in the highest OBS quartile (Q4: OBS ≥ 14.69) were significantly younger, more likely to be of Mexican American ethnicity, and had higher educational attainment and PIR compared to those in the lowest quartile (Q1: OBS < 8.85) (all P < .01). Biochemically, the Q4 group exhibited significantly higher serum 25(OH)D levels and total energy intake (both P < .001). Regarding thyroid sensitivity indices, a significant trend across OBS quartiles was observed for TFQI (P = .043), with the Q4 group showing a lower value compared to Q1. Nonsignificant trends toward lower TSHI and TT4RI values were also observed in higher OBS quartiles.
Table 2.
Weighted characteristics of participants based on quartiles of oxidative balance score, NHANES 2007 to 2012.
| Variables | Overall | Q1 (<8.85) | Q2 (8.85–12.12) | Q3 (12.12–14.69) | Q4 (>14.69) | P-value |
|---|---|---|---|---|---|---|
| n | 5536 | 1507 | 1395 | 1255 | 1379 | |
| OBS | 12.388 (3.788) | 6.981 (1.399) | 10.857 (0.910) | 13.673 (0.727) | 16.762 (1.192) | <.001 |
| Age (continuous) | 45.489 (16.351) | 46.306 (16.862) | 45.634 (16.716) | 46.460 (16.689) | 43.972 (15.244) | .002 |
| Age group (%) | .002 | |||||
| 20 ≤ age ≤ 39 | 1996 (40.256) | 474 (38.521) | 501 (40.5076) | 468 (39.9905) | 553 (41.5791) | |
| 40 ≤ age ≤ 59 | 1780 (38.091) | 473 (37.6839) | 431 (36.1681) | 390 (35.3249) | 486 (42.2147) | |
| Age ≥ 60 | 1760 (21.653) | 560 (23.7951) | 463 (23.3243) | 397 (24.6846) | 340 (16.2062) | |
| Gender (%) | .541 | |||||
| Women | 2528 (46.628) | 700 (48.8287) | 646 (46.2349) | 552 (45.6125) | 630 (46.0885) | |
| Men | 3008 (53.373) | 807 (51.1713) | 749 (53.7651) | 703 (54.3875) | 749 (53.9115) | |
| Race/ethnicity (%) | <.001 | |||||
| Non-Hispanic Black | 867 (8.130) | 224 (8.126) | 209 (8.237) | 210 (8.549) | 224 (7.708) | |
| Non-Hispanic White | 553 (4.963) | 160 (5.938) | 131 (4.473) | 131 (5.245) | 131 (4.405) | |
| Mexican American | 2557 (69.541) | 617 (62.632) | 634 (67.859) | 634 (70.773) | 708 (75.226) | |
| Other race | 1559 (17.367) | 506 (23.304) | 421 (19.431) | 421 (15.433) | 316 (12.661) | |
| Education level (%) | <.001 | |||||
| Under high school | 1565 (18.851) | 594 (28.399) | 414 (20.425) | 307 (17.035) | 250 (11.709) | |
| High school or equivalent | 1264 (22.864) | 398 (29.742) | 333 (24.110) | 273 (22.095) | 260 (17.198) | |
| College graduate or above | 2707 (58.285) | 515(41.860) | 648 (55.465) | 675 (60.870) | 869 (71.094) | |
| PIR (%) | <.001 | |||||
| <1.3 | 1749 (21.115) | 631 (32.476) | 428 (20.396) | 353 (18.158) | 337 (15.419) | |
| 1.3–3.5 | 2123 (35.744) | 589 (38.876) | 570 (39.404) | 482 (35.812) | 482 (30.244) | |
| >3.5 | 1664 (43.142) | 287 (28.648) | 397 (40.200) | 420 (46.030) | 560 (54.337) | |
| 25(OH)D (nmol/L) | 67.648 (26.390) | 61.547 (29.517) | 65.157 (25.251) | 70.415 (26.092) | 72.170 (23.791) | <.001 |
| 25(OH)D group (%) | <.001 | |||||
| Insufficient | 3967 (63.914) | 1164 (71.815) | 1054 (69.556) | 861 (61.091) | 888 (55.435) | |
| Sufficient | 1569 (36.085) | 343 (28.185) | 341 (30.445) | 394 (38.909) | 491 (44.566) | |
| Dietary energy intake (kcal/d) | 2162.791 (856.826) | 1523.917 (530.354) | 1954.457 (600.085) | 2237.566 (739.924) | 2763.967 (910.223) | <.001 |
| FT3 (pmol/L) | 0.897 (0.032) | 0.897 (0.034) | 0.898 (0.033) | 0.895 (0.032) | 0.897 (0.030) | .081 |
| FT4 (pmol/L) | 1.514 (0.066) | 1.518 (0.066) | 1.520 (0.068) | 1.512 (0.064) | 1.508 (0.066) | .077 |
| TSH (mIU/L) | 0.468 (0.637) | 0.446 (0.674) | 0.451 (0.631) | 0.516 (0.662) | 0.461 (0.588) | .069 |
| FT3/FT4 | ‐0.719 (0.188) | ‐0.727 (0.184) | ‐0.729 (0.196) | ‐0.721 (0.187) | ‐0.704 (0.184) | .163 |
| TSHI | 1.800 (1.418–2.164) | 1.788 (1.386–2.154) | 1.802 (1.443–2.202) | 1.821 [1.459–2.205] | 1.772 [1.387–2.121] | .334 |
| TT4RI | 3.618 (0.989) | 3.597 (1.049) | 3.616 (0.984) | 3.682 (1.017) | 3.583 (0.920) | .199 |
| TFQI | 0.058 (‐0.183–0.323) | 0.063 (‐0.189–0.326) | 0.082 (‐0.155–0.359) | 0.054 (‐0.206–0.339) | 0.040 [-0.190–0.277] | .043 |
Values are presented as unweighted n (weighted %) for categorical variables and weighted mean ± standard deviation or median (IQR) for continuous variables, as appropriate. Race/ethnicity values are survey-weighted and may differ from unweighted counts due to NHANES oversampling and analytic exclusions.
25(OH)D = serum 25OHD2 + 25OHD3, FT3 = free triiodothyronine, FT3/FT4 = free triiodothyronine to free thyroxine ratio, FT4 = free thyroxine, IQR = interquartile range, NHANES = National Health and Nutrition Examination Survey, PIR = poverty income ratio, Q1 = quartile 1, Q2 = quartile 2, Q3 = quartile 3, Q4 = quartile 4, TFQI = thyroid feedback quantile-based index, TgAb = thyroglobulin antibodies, TPOAb = thyroid peroxidase antibodies, TSH = thyroid stimulating hormone, TSHI = thyroid-stimulating hormone index, TT4RI = thyrotrophic thyroxine resistance index.
3.2. Associations between OBS and TH sensitivity indices
Weighted multivariable linear regression analyses revealed distinct patterns of association between OBS and thyroid hormone sensitivity indices (Table 3). In the fully adjusted model (Model III), when OBS was treated as a continuous variable, each unit increase in OBS was significantly associated with an increase in TSHI (β = 0.008, 95% CI: 0.002–0.014, P = .012) and TT4RI (β = 0.014, 95% CI: 0.004–0.024, P = .009).When OBS was analyzed in quartiles, a significant positive trend was observed for both TSHI (P for trend = .029) and TT4RI (P for trend = .020) in Model III. Compared to Q1, the strongest associations were observed in Q3 for TSHI (β = 0.0815, P = .014) and TT4RI (β = 0.1405, P = .011), with attenuated but still significant associations in Q4. No significant linear associations were found between OBS and FT3/FT4 ratio or TFQI in Model III.
Table 3.
Association of OBS with sensitivity-related indicators in US adult population, NHANES 2007 to 2012.
| Model I | Model II | Model III | |||||||
|---|---|---|---|---|---|---|---|---|---|
| β | P-value | 95% CI | β | P-value | 95% CI | β | P-value | 95% CI | |
| FT3/FT4 | |||||||||
| OBS (continuous) | 0.0018 | .1030 | –0.0004–0.0039 | 0.0017 | .0838 | ‐0.0002 to 0.0036 | ‐0.0018 | .1785 | ‐0.0045 to 0.0009 |
| OBS category | |||||||||
| Q1 | Reference | ||||||||
| Q2 | –0.0024 | .8180 | –0.0230–0.0183 | –0.0017 | .8568 | ‐0.0212 to 0.0177 | ‐0.0109 | .2892 | ‐0.0313 to 0.0096 |
| Q3 | 0.0057 | .5270 | –0.0123–0.0237 | 0.0093 | .2867 | ‐0.0081 to 0.0266 | ‐0.0051 | .5829 | ‐0.0239 to 0.0136 |
| Q4 | 0.0224 | .0390 | 0.0012–0.0436 | 0.0211 | .0338 | 0.0017 to 0.0405 | ‐0.0050 | .6988 | ‐0.0308 to 0.0209 |
| P for trend | 0.0079 | .0319 | 0.0007–0.0151 | 0.0077 | .0259 | 0.0010 to 0.0145 | ‐0.0006 | .8791 | ‐0.0092 to 0.0079 |
| TSHI | |||||||||
| OBS (continuous) | –0.0005 | .8268 | –0.0047–0.0037 | –0.0026 | .2745 | ‐0.0074 to 0.0022 | 0.0082 | .0118 | 0.0019 to 0.0144 |
| OBS category | |||||||||
| Q1 | Reference | ||||||||
| Q2 | 0.0187 | .4790 | –0.0340–0.0714 | 0.0075 | .7619 | ‐0.0388 to 0.0597 | 0.0423 | .1145 | ‐0.0103 to 0.0929 |
| Q3 | 0.0456 | .1200 | –0.0124–0.1035 | 0.0218 | .4400 | ‐0.0306 to 0.0837 | 0.0815 | .0139 | 0.0136 to 0.1419 |
| Q4 | –0.0103 | .6480 | –0.0553–0.0347 | –0.0317 | .1950 | ‐0.0696 to 0.0262 | 0.0692 | .0299 | 0.0071 to 0.1284 |
| P for trend | –0.0021 | .7730 | –0.0168–0.0126 | –0.0095 | .2618 | ‐0.0265 to 0.0074 | 0.0239 | .0289 | 0.0026 to 0.0452 |
| TT4RI | |||||||||
| OBS (continuous) | –0.00002 | .9950 | –0.0070–0.00691 | –0.0033 | .3783 | ‐0.0109 to 0.0043 | 0.0137 | .0092 | 0.0036 to 0.0238 |
| OBS category | |||||||||
| Q1 | Reference | ||||||||
| Q2 | 0.0189 | .6747 | –0.0711–0.1089 | 0.0009 | .9830 | ‐0.0781 to 0.0901 | 0.0556 | .2249 | ‐0.0357 to 0.1469 |
| Q3 | 0.0845 | .0905 | –0.0139–0.1829 | 0.0463 | .3331 | ‐0.0423 to 0.1506 | 0.1405 | .0114 | 0.0336 to 0.2473 |
| Q4 | –0.0146 | .6854 | –0.0868–0.0576 | –0.0483 | .2165 | ‐0.1086 to 0.0446 | 0.1106 | .0314 | 0.0104 to 0.2107 |
| P for trend | –0.0007 | .9542 | –0.0241–0.0228 | –0.0122 | .3647 | ‐0.0394 to 0.0148 | 0.0407 | .0204 | 0.0067 to 0.0747 |
| TFQI | |||||||||
| OBS (continuous) | –0.0030 | .1320 | –0.0068–0.0009 | –0.0043 | .0324 | ‐0.0082 to ‐0.0038 | 0.0036 | .1188 | ‐0.0001 to 0.0083 |
| OBS category | |||||||||
| Q1 | Reference | ||||||||
| Q2 | 0.0232 | .1090 | –0.0054–0.0518 | 0.0163 | .2331 | ‐0.0109 to 0.0434 | 0.0417 | .0056 | 0.0130 to 0.0703 |
| Q3 | 0.0094 | .6500 | –0.0322–0.0510 | –0.0053 | .7899 | ‐0.0457 to 0.0350 | 0.0380 | .0931 | ‐0.0067 to 0.0827 |
| Q4 | –0.0300 | .1340 | –0.0695–0.0096 | –0.0430 | .0334 | ‐0.0824 to ‐0.0036 | 0.0307 | .1935 | ‐0.0163 to 0.0778 |
| P for trend | –0.0117 | .0837 | –0.0695–0.0096 | –0.0163 | .0195 | ‐0.0297 to ‐0.0028 | 0.0079 | .3298 | ‐0.0083 to 0.0240 |
Model I without adjustments, Model II adjusted for age, gender, race/ethnicity, educational level, PIR; Model III further adjusted for 25(OH)D and total energy intakes.
Bold values indicate the P for trend (overall P-values) for OBS categories.
CI = confidence interval, FT3/FT4 = free triiodothyronine to free thyroxine ratio, NHANES = National Health and Nutrition Examination Survey, OBS = oxidative balance score, PIR = poverty income ratio, Q1 = quartile 1, Q2 = quartile 2, Q3 = quartile 3, Q4 = quartile 4, TFQI = thyroid feedback quantile-based index, TSHI = thyroid-stimulating hormone index, TT4RI = thyrotrophic thyroxine resistance index.
3.3. Nonlinear association between OBS and TH sensitivity indices
RCS analyses uncovered significant nonlinearity and critical inflection points, reconciling the apparent contradictions from linear models (Table 4, Fig. S1, Supplemental Digital Content, https://links.lww.com/MD/R686). In Model III, for central sensitivity indices, a significant positive association was observed below an inflection point of OBS = 8.275 for both TSHI (β = 0.029, 95% CI: 0.008–0.049, P = .006) and TT4RI (β = 0.048, 95% CI: 0.015–0.081, P = .005). Above this threshold, the associations became nonsignificant (P > .05). A similar pattern was observed for TFQI, with a positive association below OBS = 8.847 (β = 0.014, 95% CI: 0.004–0.025, P = .007) that vanished above it. Conversely, for peripheral sensitivity (FT3/FT4 ratio), no significant association was observed at lower OBS levels. However, a significant negative association emerged after OBS exceeded a threshold of 15.681 (β = −0.013, 95% CI: ‐0.020 to ‐0.005, P = .001).
Table 4.
Association of OBS with TH sensitivity indices (n = 5536).
| Outcome | Model I | Model III | ||||||
|---|---|---|---|---|---|---|---|---|
| Cut points | β (95% CI) | P-value | Log likelihood ratio | Cut points | β (95% CI) | P-value | Log likelihood ratio | |
| FT3/FT4 | <6.497 | ‐0.017 (‐0.030 to ‐0.003) | .014 | 0.007 | <15.681 | ‐0.001 (‐0.003 to 0.000) | .143 | 0.008 |
| >6.497 | 0.002 (0.001 to 0.004) | .001 | >15.681 | ‐0.013 (‐0.020 to ‐0.005) | .001 | |||
| TSHI | <8.275 | 0.030 (0.009 to 0.050) | .005 | 0.002 | <8.275 | 0.029 (0.008 to 0.049) | .006 | 0.030 |
| >8.275 | ‐0.006(‐0.012 to 0.000) | .047 | >8.275 | 0.002 (‐0.004 to 0.009) | .487 | |||
| TT4RI | <8.275 | 0.050 (0.016 to 0.083) | .004 | 0.003 | <8.275 | 0.048 (0.015 to 0.081) | .005 | 0.023 |
| >8.275 | ‐0.009 (‐0.019 to 0.000) | .056 | >8.275 | 0.003 (‐0.008 to 0.014) | .586 | |||
| TFQI | <6.497 | 0.038 (0.012 to 0.065) | .004 | 0.004 | <8.847 | 0.014 (0.004 to 0.025) | .007 | 0.035 |
| >6.497 | ‐0.004 (‐0.007 to ‐0.001) | .003 | >8.847 | 0.001 (‐0.004 to 0.005) | .757 | |||
Model I without adjustments, Model III adjusted for age, gender, race/ethnicity, educational level, PIR, 25(OH)D and total energy intakes.
FT3/FT4 = free triiodothyronine to free thyroxine ratio, PIR = poverty income ratio, TFQI = thyroid feedback quantile-based index, TSHI = thyroid stimulating hormone index, TT4RI = thyrotrophic thyroxine resistance index.
3.4. Subgroup and sensitivity analysis
Subgroup analyses demonstrated that the associations between OBS and the primary thyroid sensitivity indices were consistent across all predefined subgroups (all P for interaction > .05) (Figs. S2–S5, Supplemental Digital Content, https://links.lww.com/MD/R686). Sensitivity analyses using OBS tertiles yielded results consistent with the primary analysis, confirming the robustness of the findings (Table S1, Supplemental Digital Content, https://links.lww.com/MD/R687).
4. Discussion
In this large, cross-sectional study of a nationally representative US adult population, we move beyond conventional linear assumptions to uncover significant nonlinear associations between the OBS and thyroid hormone sensitivity. The key finding is a dual-threshold pattern: a prooxidant-dominant status (OBS ≤ 8.275 for TSHI/TT4RI; ≤8.847 for TFQI) was associated with impaired central sensitivity, whereas an antioxidant-rich status (OBS > 15.681) was associated with reduced peripheral conversion efficiency, as indicated by a lower FT3/FT4 ratio. The initial absence of significant linear associations for the FT3/FT4 ratio and TFQI in multivariate models underscores the limitation of linear approaches and necessitated the use of RCS to characterize these complex, threshold-dependent relationships.
The impairment of central sensitivity at low OBS levels may be explained by experimental evidence showing that prooxidants can induce oxidative damage to TRH and TSH-secreting neurons or alter thyroid hormone receptor expression and function.[13,31] For instance, the prooxidant Di2-ethylhexyl phthalate has been shown to activate the Ras/Akt pathway, upregulate TRH receptor expression, and lead to dysregulated HPT feedback,[13] a mechanism that parallels our observation of elevated central resistance indices in the low OBS range. The plateau effect beyond the inflection point (OBS = 8.275) suggests a saturation of benefit, where further increases in antioxidant exposure yield diminishing returns, potentially due to nutrient redundancy or the establishment of a new redox homeostasis.
Conversely, the novel finding of reduced peripheral sensitivity at high OBS levels suggests a potential downside of excessive antioxidant intake. This phenomenon may be explained by 2 interconnected biological mechanisms. First, supraphysiological levels of certain antioxidants can paradoxically exhibit prooxidant effects through autoxidation processes,[32] and elevated reactive oxygen species are known to directly inhibit type 1 deiodinase activity, the key enzyme for T4-to-T3 conversion.[14] Second, analogous to xenobiotic-induced hepatic enzyme induction, a high dietary antioxidant load may upregulate phase II detoxification enzymes (e.g., UDP-glucuronosyltransferases and sulfotransferases), thereby accelerating the metabolic clearance and inactivation of thyroid hormones.[13] This dual-mechanism framework explains the counterproductive endocrine effects observed at the upper extreme of the OBS and underscores the principle that “more is not always better.” Taken together, the divergent associations between oxidative balance and central versus peripheral sensitivity delineated above highlight a complex, tissue-specific regulatory role that warrants further mechanistic investigation.
Our findings reconcile apparent paradoxes in previous research that focused on isolated nutrients or lifestyle factors.[33–36] For example, while selenium is essential for antioxidant enzymes, its supplementation exhibits a U-shaped relationship with thyroid function.[35] Similarly, moderate alcohol intake may have antioxidant properties, but heavy consumption is a prooxidant linked to altered thyroid profiles.[34] The OBS framework integrates these opposing and synergistic effects, demonstrating that the net systemic redox balance, rather than any single component, is a critical determinant of thyroid sensitivity. It is noteworthy that participants with a higher OBS tended to have a more favorable socioeconomic and lifestyle profile, including younger age, higher educational attainment, and income. This pattern is sociologically plausible, as higher socioeconomic status is generally associated with greater access to health-promoting resources and nutritional knowledge, which facilitate the maintenance of healthier dietary habits and lifestyles and a superior oxidative balance.[37] However, the persistence of the OBS-thyroid sensitivity associations after rigorous adjustment for these confounders indicates that the OBS captures a dimension of redox balance that is independently associated with endocrine function. From a public health perspective, this underscores the importance of holistic lifestyle and dietary interventions aimed at maintaining an optimal oxidative balance for endocrine health.
The major strengths of this study include the use of a large, nationally representative sample, a comprehensive OBS that integrates dietary and lifestyle components, the application of novel thyroid sensitivity indices for a refined assessment of central and peripheral thyroid hormones regulation, and the use of RCS to characterize complex nonlinear relationships, thereby providing novel insights into threshold effects. However, several limitations warrant consideration. First, the cross-sectional design of the NHANES data poses a significant limitation in our ability to infer causality. While we can identify associations between variables, establishing a clear cause-and-effect relationship is challenging without longitudinal data. Second, although we adjusted for a broad range of covariates based on prior literature and a priori knowledge, the potential for residual confounding cannot be entirely ruled out. There may be unmeasured or unaccounted-for variables that could influence the observed associations, and this residual confounding could potentially affect the robustness of our findings. It is important to note that our stringent exclusion criteria (excluding participants with known thyroid disease, cancer, pregnancy, those taking medications that affect thyroid function, and individuals with abnormal thyroid peroxidase antibodies) (were implemented specifically to mitigate the influence of major clinical conditions that could confound or mediate the relationship between OBS and thyroid sensitivity). Third, the OBS, derived from self-reported data, is susceptible to measurement error. Furthermore, the equal weighting of its components may not precisely reflect their biological contributions, although prior evidence suggests that weighted and unweighted scores yield similar associations with health outcomes.[38] Despite these limitations, the consistency of our findings across extensive sensitivity and subgroup analyses strengthens the credibility of our conclusions.
5. Conclusions
This study reveals complex, nonlinear associations between oxidative balance and thyroid hormone sensitivity. These findings suggest a dual-threshold pattern, wherein both pro-oxidant dominance and antioxidant excess may be associated with impairments in central and peripheral thyroid regulation, respectively. This OBS framework may offer a useful approach for future research aimed at identifying individuals at risk of redox-mediated thyroid dysfunction and exploring the potential of holistic lifestyle interventions for thyroid health.
Acknowledgments
The authors thank all team members and participants in the NHANES study.
Author contributions
Conceptualization: Zhu-Zhu Wang, Yu-Han Zhang, Qin Xu.
Data curation: Yi Wang.
Formal analysis: Zhu-Zhu Wang, Yu-Han Zhang, Qin Xu.
Funding acquisition: Jing-Fang Hong.
Project administration: Jing-Fang Hong.
Supervision: Jing-Fang Hong.
Validation: Zhu-Zhu Wang, Yu-Han Zhang, Qin Xu.
Visualization: Zhu-Zhu Wang.
Writing – review & editing: Zhu-Zhu Wang, Ji Zhou, Yong-Xia Song, Xiao-Qing Lv.
Supplementary Material
Abbreviations:
- 25(OH)D
- 25-hydroxyvitamin D
- BMI
- body mass index
- HPT axis
- hypothalamic–pituitary–thyroid axis
- MET
- metabolic equivalent
- NCHS
- National Center for Health Statistics
- NHANES
- National Health and Nutrition Examination Survey
- OBS
- oxidative balance score
- PIR
- poverty-income ratio
- RCS
- restricted cubic splines
- TFQI
- thyroid feedback quantile-based index
- THs
- thyroid hormones
- TSH
- thyroid-stimulating hormone
- TSHI
- thyroid-stimulating hormone index
- TT4RI
- thyrotroph thyroxine resistance index
This study was conducted according to the guideline laid down in the Declaration of Helsinki, and all procedures involving study participants were approved by the Institutional Review Board of the NCHS. Ethical review and approval were waived for this study as it solely used publicly available data for research and publication. Informed consent was obtained from all subjects involved in the NHANES.
The authors have no conflicts of interest to disclose.
This work was supported by the National Natural Science Foundation of China (82272926); Humanities and Social Sciences Research of Anhui Provincial Higher Education Institutions (SK2020ZD13); Promoting scientific research cooperation and high-level talent training projects with Canada, Australia, New Zealand, and Latin America of the National Scholarship Foundation ((2022)1007).
The datasets generated during and/or analyzed during the current study are publicly available.
Supplemental Digital Content is available for this article.
How to cite this article: Wang Z-Z, Zhang Y-H, Xu Q, Zhou J, Song Y-X, Wang Y, Lv X-Q, Hong J-F. Nonlinear associations of oxidative balance score with central and peripheral thyroid hormone sensitivity: A cross-sectional study of NHANES 2007 to 2012. Medicine 2026;105:16(e48277).
ZZW, YHZ, and QX contributed to this article equally.
Contributor Information
Zhu-Zhu Wang, Email: wynursing@outlook.com.
Yu-Han Zhang, Email: 2345011189@stu.ahmu.edu.cn.
Qin Xu, Email: 2345011273@stu.ahmu.edu.cn.
Ji Zhou, Email: zjanhmu@163.com.
Yong-Xia Song, Email: 2015500020@ahmu.edu.cn.
Yi Wang, Email: wynursing@outlook.com.
Xiao-Qing Lv, Email: 1107815747@qq.com.
References
- [1].Shahid MA, Ashraf MA, Sharma S. Physiology, thyroid hormone. StatPearls. StatPearls Publishing, 2024. http://www.ncbi.nlm.nih.gov/books/NBK500006/ [accessed April 6, 2024]. [PubMed] [Google Scholar]
- [2].Laclaustra M, Moreno-Franco B, Lou-Bonafonte JM, et al. Impaired sensitivity to thyroid hormones is associated with diabetes and metabolic syndrome. Diabetes Care. 2019;42:303–10. [DOI] [PubMed] [Google Scholar]
- [3].Xiao X-Q, Fu F-S, Xiang C, Yan H-C. Sensitivity to thyroid hormones is associated with sleep duration in the euthyroid population with depression degree lower than moderate. Sci Rep. 2024;14:6583. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [4].Taylor PN, Razvi S, Pearce SH, Dayan CM. A review of the clinical consequences of variation in thyroid function within the reference range. J Clin Endocrinol Metab. 2013;98:3562–71. [DOI] [PubMed] [Google Scholar]
- [5].Yu L, Li Z, Yang R, et al. Impaired sensitivity to thyroid hormones is associated with elevated blood glucose in coronary heart disease. Front Endocrinol. 2022;13:895843. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [6].Sun Y, Teng D, Zhao L, et al. Impaired sensitivity to thyroid hormones is associated with hyperuricemia, obesity, and cardiovascular disease risk in subjects with subclinical hypothyroidism. Thyroid. 2022;32:376–84. [DOI] [PubMed] [Google Scholar]
- [7].Cui C, Sui H, Wang Z, et al. Thyroid hormone sensitivity and diabetes onset: a longitudinal cross-lagged cohort. Front Endocrinol. 2023;14:1267612. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [8].Yu G, Liu S, Song C, et al. Association of sensitivity to thyroid hormones with all-cause mortality in euthyroid US adults: a nationwide cohort study. Eur Thyroid J. 2024;13:e230130. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [9].Veal EA, Day AM, Morgan BA. Hydrogen peroxide sensing and signaling. Mol Cell. 2007;26:1–14. [DOI] [PubMed] [Google Scholar]
- [10].De Groot LJ. Non-thyroidal illness syndrome is a manifestation of hypothalamic–pituitary dysfunction, and in view of current evidence, should be treated with appropriate replacement therapies. Crit Care Clin. 2006;22:57–86, vi. [DOI] [PubMed] [Google Scholar]
- [11].Chopra IJ. Euthyroid sick syndrome: is it a misnomer? J Clin Endocrinol Metab. 1997;82:329–34. [DOI] [PubMed] [Google Scholar]
- [12].Mancini A, Raimondo S, Segni CD, et al. Non-thyroidal illness: physiopathology and clinical implications. Current topics in hypothyroidism with focus on development. IntechOpen, 2013. [Google Scholar]
- [13].Ye H, Ha M, Yang M, Yue P, Xie Z, Liu C. Di2-ethylhexyl phthalate disrupts thyroid hormone homeostasis through activating the ras/akt/TRHr pathway and inducing hepatic enzymes. Sci Rep. 2017;7:40153. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [14].Chen K, Yan B, Wang F, et al. Type 1 5′-deiodinase activity is inhibited by oxidative stress and restored by alpha-lipoic acid in HepG2 cells. Biochem Biophys Res Commun. 2016;472:496–501. [DOI] [PubMed] [Google Scholar]
- [15].Klasson CL, Sadhir S, Pontzer H. Daily physical activity is negatively associated with thyroid hormone levels, inflammation, and immune system markers among men and women in the NHANES dataset. PLoS One. 2022;17:e0270221. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [16].Garofalo V, Condorelli RA, Cannarella R, Aversa A, Calogero AE, La Vignera S. Relationship between iron deficiency and thyroid function: a systematic review and meta-analysis. Nutrients. 2023;15:4790. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [17].Wang X, Gao X, Han Y, et al. Causal association between serum thyrotropin and obesity: a bidirectional, Mendelian randomization study. J Clin Endocrinol Metab. 2021;106:e4251–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [18].Hernández-Ruiz A, García-Villanova B, Guerra-Hernández EJ, et al. Oxidative Balance Scores (OBSs) integrating nutrient, food and lifestyle dimensions: development of the NutrientL-OBS and FoodL-OBS. Antioxidants (Basel, Switzerland). 2022;11:300. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [19].Sohouli MH, Baniasadi M, Hernández-Ruiz A, et al. Adherence to oxidative balance scores is associated with a reduced risk of breast cancer; a case-control study. Nutr Cancer. 2023;75:164–73. [DOI] [PubMed] [Google Scholar]
- [20].Li H, Song L, Cen M, et al. Oxidative balance scores and depressive symptoms: mediating effects of oxidative stress and inflammatory factors. J Affect Disord. 2023;334:205–12. [DOI] [PubMed] [Google Scholar]
- [21].Zhan F, Lin G, Duan K, Huang B, Chen L, Ni J. Higher oxidative balance score decreases risk of stroke in US adults: evidence from a cross-sectional study. Front Cardiovasc Med. 2023;10:1264923. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [22].Ke R, He Y, Chen C. Association between oxidative balance score and kidney stone in United States adults: analysis from NHANES 2007-2018. Front Physiol. 2023;14:1275750. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [23].Cheng S, Han Y, Jiang L, Lan Z, Liao H, Guo J. Associations of oxidative balance score and visceral adiposity index with risk of ischaemic heart disease: a cross-sectional study of NHANES, 2005–2018. BMJ Open. 2023;13:e072334. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [24].Xu Z, Lei X, Chu W, Weng L, Chen C, Ye R. Oxidative balance score was negatively associated with the risk of metabolic syndrome, metabolic syndrome severity, and all-cause mortality of patients with metabolic syndrome. Front Endocrinol. 2023;14:1233145. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [25].Obsekov V, Ghassabian A, Mukhopadhyay S, Trasande L. Manganese and thyroid function in the national health and nutrition examination survey, 2011–2012. Environ Res. 2023;222:115371. [DOI] [PubMed] [Google Scholar]
- [26].Song L, Zhou H, Yang Q, et al. Association between the oxidative balance score and thyroid function: Results from the NHANES 2007–2012 and Mendelian randomization study. PLoS One. 2024;19:e0298860. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [27].Walsh M. Thyroid Peroxidase Antibodies. NCHS Lab. Proced. Man; 2006. [Google Scholar]
- [28].Walsh M. Thyroglobulin Antibodies. NCHS Lab. Proced. Man; 2006. [Google Scholar]
- [29].Zhang W, Peng S-F, Chen L, Chen H-M, Cheng X-E, Tang Y-H. Association between the oxidative balance score and telomere length from the national health and nutrition examination survey 1999–2002. Oxid Med Cell Longev. 2022;2022:1345071. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [30].Li Y, Sun J, Jiao Y, Li N, Zhao W. Impaired sensitivity to thyroid hormones is associated with decreased vitamin D levels in the euthyroid population. J Clin Endocrinol Metab. 2024;109:691–700. [DOI] [PubMed] [Google Scholar]
- [31].Kim H, Ji K. Exposure to humidifier disinfectants induces developmental effects and disrupts thyroid endocrine systems in zebrafish larvae. Ecotoxicol Environ Saf. 2019;184:109663. [DOI] [PubMed] [Google Scholar]
- [32].Sotler R, Poljšak B, Dahmane R, et al. Prooxidant activities of antioxidants and their impact on health. Acta Clin Croat. 2019;58:726–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [33].Vieira IH, Rodrigues D, Paiva I. Vitamin D and autoimmune thyroid disease-cause, consequence, or a vicious cycle? Nutrients. 2020;12:2791. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [34].Gruppen EG, Kootstra-Ros J, Kobold AM, et al. Cigarette smoking is associated with higher thyroid hormone and lower TSH levels: the PREVEND study. Endocrine. 2020;67:613–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [35].Huwiler VV, Maissen-Abgottspon S, Stanga Z, et al. Selenium supplementation in patients with hashimoto thyroiditis: a systematic review and meta-analysis of randomized clinical trials. Thyroid. 2024;34:295–313. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [36].Zhao G, Wang Z, Ji J, Cui R. Effect of coffee consumption on thyroid function: NHANES 2007-2012 and Mendelian randomization. Front Endocrinol. 2023;14:1188547. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [37].Pantoja-Ruiz C, Akinyemi R, Lucumi-Cuesta DI, et al. Socioeconomic status and stroke: a review of the latest evidence on inequalities and their drivers. Stroke. 2025;56:794–805. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [38].Kong SY, Goodman M, Judd S, Bostick RM, Flanders WD, McClellan W. Oxidative balance score as predictor of all-cause, cancer, and noncancer mortality in a biracial US cohort. Ann Epidemiol. 2015;25:256–62.e1. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.