Data were derived from the NHANES, a large-scale, cross-sectional study conducted by the U.S. National Institutes of Health (NIH) and the Centers for Disease Control and Prevention (CDC), designed to assess the nutritional and health status of the U.S. population. NHANES protocols were approved by the National Center for Health Statistics (NCHS) Research Ethics Review Board, and informed consent was obtained from all participants before participation. This study utilized publicly available data from the NHANES 2005–2006 cycle. Initially, a total of 20,497 participants were included. Participants were excluded if they met any of the following criteria: (1) missing AR data; (2) incomplete data for OBS calculation; or (3) missing data on included covariates. A detailed description of the inclusion and exclusion process is provided in Figure 2.

The OBS was calculated by summing scores from 20 components, including 16 dietary and 4 lifestyle factors (13, 16). The dietary components were dietary fiber, carotene, riboflavin, niacin, vitamin B6, total folate, vitamin B12, vitamin C, vitamin E, calcium, magnesium, zinc, copper, selenium, total fat, and iron, while the four lifestyle components included physical activity, alcohol consumption, body mass index (BMI), and serum cotinine. Dietary data were obtained from the first 24-h dietary recall interview conducted in person at the Mobile Examination Center (MEC), and lifestyle information was collected through standardized questionnaires and laboratory tests. Based on prior evidence, total fat, alcohol, BMI, and cotinine were considered pro-oxidants, whereas the remaining components were treated as antioxidants. Antioxidants were scored from 0 to 2 (higher values indicating more favorable exposure), while pro-oxidants were inversely scored (2 to 0) to reflect risk.

Except for BMI, all other dietary and lifestyle components were stratified by sex where appropriate, to account for differences in daily energy intake and activity patterns between men and women. BMI was the only variable for which we did not apply gender-specific cut-offs, as the WHO adult classification uses the same thresholds for both sexes. Physical activity (antioxidant) was also divided into sex-specific tertiles and scored 0/1/2 from the lowest to the highest tertile within sex. Alcohol intake (pro-oxidant) was scored according to the Dietary Guidelines for Americans, 2020–2025 (9th ed.): non-drinkers = 2 points; moderate drinking = 1 point (≤14 g/day for women; ≤28 g/day for men); heavy drinking = 0 points (>14 g/day for women; >28 g/day for men). BMI was scored using WHO adult categories with the same thresholds for both sexes: <25.0 kg/m² = 2 points; 25.0–29.9 kg/m² = 1 point; ≥30.0 kg/m² = 0 points. Serum cotinine (pro-oxidant) was divided into sex-specific tertiles and inversely scored (2/1/0 from the lowest to the highest tertile within sex). The total OBS was derived by summing all component scores, with higher scores representing more favorable oxidative balance. Full scoring details are provided in Supplementary Table S1.

The diagnosis of AR was based on a combination of questionnaire data and serum allergen-specific IgE measurements. Relevant data were obtained from the 2005–2006 cycle of the NHANES, which included both allergy-related questionnaire responses and laboratory-based IgE test results for 19 specific allergens.

Two questionnaire items were used: AGQ030 (“During the past 12 months, have you had an episode of hay fever?”) and AGQ100 (“During the past 12 months, have you had a problem with sneezing, or a runny, or blocked nose when (2) did not have a cold or the flu?”). Serum-specific IgE levels were considered positive if they were ≥0.35 kU/L.

Participants were classified as having AR if they answered “yes” to either AGQ030 or AGQ100 and had a positive serum-specific IgE result. The control group included participants who answered “no” to both questions and had negative IgE results. Individuals who did not meet either criterion were excluded from the analysis (17, 18).

Covariates included in the analysis were age, gender, BMI, race/ethnicity, education, poverty-income ratio (PIR), smoking status, serum total IgE antibody, serum C-reactive protein (CRP), hypertension, diabetes, and cardiovascular disease (CVD). Race/ethnicity was categorized into five groups: Mexican American, Non-Hispanic Black, Non-Hispanic White, Other Hispanic, and Other Race. Education level was divided into three categories: above high school (college graduate or higher), high school or equivalent (high school diploma or GED), and below high school (including less than 9th grade, 9–11th grade, or no diploma). Smoking status was classified into three categories: never smokers (smoked fewer than 100 cigarettes in lifetime and currently do not smoke), former smokers (smoked at least 100 cigarettes in lifetime but no longer smoke), and current smokers (smoked at least 100 cigarettes in lifetime and currently smoke every day or on some days). Dietary supplement use was obtained from self-reported questionnaires, while total energy was assessed at the MEC using 24-h dietary recall interviews. Hypertension was defined by self-reported physician diagnosis, elevated average blood pressure (systolic ≥130 mmHg and/or diastolic ≥85 mmHg), or current use of antihypertensive medication. Diabetes was defined based on any of the following: fasting plasma glucose ≥7.0 mmol/L (126 mg/dL), 2-h oral glucose tolerance test (OGTT) result ≥11.1 mmol/L (200 mg/dL), hemoglobin A1c (HbA1c) ≥6.5%, or current use of insulin or oral hypoglycemic medications. CVD was defined using responses to the medical condition’s questionnaire. Participants were classified as having CVD if they reported ever being diagnosed by a physician with any of the following: coronary heart disease, congestive heart failure, stroke, myocardial infarction, or angina.

Continuous variables were summarized as means with standard deviations (SD), and categorical variables as frequencies and percentages. The OBS was examined in three ways: per unit increase, per 1-SD increase, and as a quartile-based categorical variable, and all forms were applied in the models to evaluate its association with AR.

To assess the robustness and national representativeness of our finding, additional weighted analyses were performed in line with NHANES analytic guidelines. These accounted for the complex survey design by incorporating the 2-year MEC examination weight (WTMEC2YR), strata (SDMVSTRA), and primary sampling units (SDMVPSU). Baseline differences between groups after applying survey weights were assessed using standardized mean differences (SMDs).

Three multivariable logistic regression models were constructed. Model 1 was unadjusted. Model 2 was adjusted for age, sex, and race/ethnicity. Model 3 included additional adjustments for education level, PIR, serum CRP, smoking status, dietary supplement, total energy, hypertension, diabetes, and CVD. Multicollinearity was assessed using the variance inflation factor (VIF). Dose–response relationships were explored using restricted cubic spline (RCS) models based on Model 3, with knot placement optimized via Akaike information criterion (AIC). If nonlinearity was detected, associations on either side of the inflection point were estimated separately.

Subgroup analyses and interaction tests were conducted to explore effect modification across different population strata. To address the issue of multiple comparisons, the Benjamini–Hochberg method was applied to control the false discovery rate (FDR). To evaluate the robustness of the results, four sensitivity analyses were conducted. First, participants with asthma were excluded. Second, considering the potential associations between dietary supplement use and participants’ dietary habits or behaviors, participants who reported dietary supplement use were excluded. Third, dietary components of the OBS were energy-adjusted using the residual method based on total energy intake estimated from the first 24-h dietary recall interview conducted in person at the MEC. Fourth, alternative definitions of AR were applied, including defining AR as the presence of any sIgE ≥0.35 kU/L, and using a higher cutoff of ≥0.7 kU/L for sIgE.

All statistical tests were two-sided, with p-values <0.05 considered statistically significant. All analyses were performed using R software (version 4.5.0).

Among the 1,491 participants, 513 were classified as having AR, and 978 as controls. Table 1 presents the baseline characteristics of the study population stratified by AR status. Compared with the control group, participants with AR were significantly younger. There were no significant differences in gender distribution, BMI, serum CRP levels, or the prevalence of diabetes, hypertension, and CVD between the two groups. However, race/ethnicity, education level, PIR, and smoking status differed significantly between groups. Serum total IgE levels were markedly higher in the AR group. Participants with AR also exhibited significantly higher total and dietary OBS, whereas lifestyle OBS did not differ between groups. In addition, the AR group reported higher total energy intake and a greater prevalence of dietary supplement use. After weighting, the differences between groups in gender, BMI, PIR, serum CRP, hypertension, and diabetes were minimal (all SMD <0.1). Detailed results are presented in Supplementary Table S2.

VIF analysis revealed no evidence of multicollinearity, with all VIF values below 5 (Supplementary Table S3). Table 2 shows the associations between OBS and AR across three logistic regression models.

For total OBS, significant positive associations with AR were consistently observed. In the fully adjusted Model 3, each unit increase in total OBS was associated with a 2% higher likelihood of AR (OR = 1.02; 95% CI: 1.01, 1.04; p = 0.003). When expressed per SD increase, total OBS was associated with a 19% higher likelihood of AR (OR = 1.19; 95% CI: 1.06, 1.34; p = 0.003). Quartile-based analysis indicated a dose–response relationship, with participants in the highest OBS quartile (Q4) having significantly increased odds of AR compared to the lowest quartile (Q1) (OR = 1.67; 95% CI: 1.20, 2.33; p = 0.003). The trend across quartiles was statistically significant (p for trend = 0.003). Dietary OBS demonstrated a similar pattern. In Model 3, each unit increase in dietary OBS corresponded to a 3% higher likelihood of AR (OR = 1.03; 95% CI: 1.01, 1.04; p = 0.003). When expressed per SD increase, dietary OBS was associated with a 19% higher likelihood of AR (OR = 1.19; 95% CI: 1.06, 1.34; p = 0.003). Quartile analysis reinforced this association, with the highest dietary OBS quartile showing increased odds (OR = 1.60; 95% CI: 1.16, 2.22; p = 0.005) and a significant trend (p for trend = 0.005). Lifestyle OBS did not show significant associations with AR in either unit (OR = 1.01; 95% CI: 0.93–1.08; p = 0.900), per SD (OR = 1.01; 95% CI: 0.90–1.14; p = 0.900), or quartile analyses (all p > 0.05). RCS analyses (Figure 3) revealed significant linear relationships between OBS and AR, including total OBS (p for overall = 0.006) and dietary OBS (p for overall = 0.001).

For total OBS, significant positive associations with AR were also observed after applying survey weights. In the fully adjusted Model 3, each unit increase in total OBS was associated with a 2% higher likelihood of AR (OR = 1.02; 95% CI: 1.00, 1.03; p = 0.047). When expressed per SD increase, total OBS was associated with a 12% higher likelihood of AR (OR = 1.12; 95% CI: 1.00, 1.26; p = 0.047). Quartile-based analysis further supported a dose–response relationship, with participants in the highest OBS quartile (Q4) having significantly increased odds of AR compared to the lowest quartile (Q1) (OR = 1.65; 95% CI: 1.08, 2.51; p = 0.029). The trend across quartiles remained statistically significant (p for trend = 0.017). For dietary OBS and lifestyle OBS, no significant associations with AR were observed after applying survey weights. Detailed results are presented in Supplementary Table S4.

In sensitivity analyses, the results were largely consistent with the primary findings. First, after excluding participants with asthma, total OBS and dietary OBS were still robustly associated with AR, with clear dose–response relationships across quartiles (p for trend <0.001; Supplementary Table S5). Second, after excluding participants who reported dietary supplement use yielded comparable results, reinforcing the robustness of the associations (Supplementary Table S6). Third, dietary components of the OBS were energy-adjusted using the residual method (assignment scheme shown in Supplementary Table S7). After energy adjustment, both total and dietary OBS remained significantly associated with AR with consistent dose–response patterns, whereas lifestyle OBS showed no significant association (Supplementary Table S8). Finally, applying alternative definitions of AR (≥0.35 kU/L or ≥0.7 kU/L for sIgE positivity) did not materially change the results, and positive associations between OBS and AR were consistently observed (Supplementary Table S9).

Exploratory subgroup analyses were performed to assess whether the association between total OBS (per SD) and AR was consistent across demographic and clinical subgroups (Figure 4). After controlling the FDR, the positive association was largely consistent across subgroups. The association appeared stronger in younger participants, males, and those with lower educational attainment. However, none of the interaction terms reached statistical significance (all p for interaction >0.05).