The mediating role of BMI in the association between composite dietary antioxidant index and infertility in women: a nationwide population-based study.

Of the 1806 participants in this study cohort, median age [IQR] was 33 years [ 26 , 39 ], 1495 (82.78%) had high school education and above, 1044 (57.8%) were married/cohabiting, 321 (17.8%) were Mexican Americans, 198 (11%) other Hispanics, 581 (32.3%) non-Hispanic whites, 415 (23%) non-Hispanic blacks, and 291 (16.1%) “other” races or ethnicities (including American Indians/Alaska Natives/Pacific Islanders, Asians, and multiracial people). The median [IQR] of BMI for participants was 28.4 [23.5, 34.4] kg/m 2 . Among all participants, 211 (11.68%) were in the infertile group, and BMI for participants in the infertile group was significantly higher than that in the non-infertile group (31.5 [24.9, 39.1] vs. 28 [23.4, 38.8] kg/m 2 , P < 0.001). On the other hands, the CDAI level in the infertile group was lower than that in the non-infertile group (−1.9 [−3.4, 0.2] vs. −1.0 [−2.8, 1.7], P < 0.001). Moreover, compared with the non-infertile group, the infertile group had lower intakes of vitamin E, vitamin A, vitamin D, α-carotene, zinc, and selenium ( P < 0.05). In addition, there were significant differences between the groups in age, marital status, smoking, drinking, physical activity level, hypertension, diabetes, and history of pelvic infection, estrogen use, time of menarche, and menstrual cycles within one year. (Table 1 ) Table 1 Characteristics of the participants Overall( n  = 1806) Infertile( n  = 211) Non-Infertile( n  = 1595) P -Value Age, years, median [IQR] 33.0[27,40] 36[31,41] 33[26,39]  40 458(25.4) 69(32.7) 389(24.4) Race, n (%) 0.063  Mexican American 321(17.8) 37(17.5) 284(17.8)  Other Hispanic 198(11.0) 18(8.5) 180(11.3)  Non-Hispanic White 581(32.2) 86(40.8) 495(31.0)  Non-Hispanic Black 415(23.0) 41(19.4) 374(23.4)  Other Race 291(16.1) 29(13.7) 262(16.4) Education Level, n (%) 0.721  Less than high school 311(17.2) 34(16.1) 277(17.4)  High school or above 1495(82.8) 177(83.9) 1318(82.6) Marital status, n (%) < 0.001  Yes 1044(57.8) 152(72.0) 892(55.9)  No 762(42.2) 59(28.0) 703(44.1) Pregnancy status at exam, n (%) < 0.001  Yes 93(5.4) 9(4.6) 8(5.5)  No 1584(91.6) 170(86.7) 1414(92.2) Ratio of family income to poverty, median [IQR] 1.8[0.1,3.4] 2.02[1.1,3.9] 1.7[0.9,3.4] 0.0504 Ratio of family income to poverty, n (%) 0.263  < 1.3 93(5.4) 65(33) 574(38.9)  1.3–3.49 1584(91.6) 79(40.1) 551(37.4)  ≥ 3.5 52(3.0) 53(26.9) 350(23.7) Smoked at least 100 cigarettes in life, n (%) 0.021  Yes 524(29.0) 76(36) 448(28.1)  No 1281(71.0) 135(64) 1146(71.9) Ever had a drink of any kind of alcohol, n (%) 0.293  Yes 1197(90.3) 135(87.7) 1062(90.7)  No 128(9.7) 19(12.3) 109(9.3) Physical activities, n (%) 0.046  Sufficient 593(32.8) 56(26.5) 537(33.7)  Insufficient 1213(67.2) 155(73.5) 1058(66.3) Hypertension, n (%) 0.006  Yes 267(14.8) 45(21.3) 222(13.9)  No 1538(85.2) 166(78.7) 1372(86.1) Diabetes, n (%) 0.001  Yes 74(4.2) 18(8.7) 56(3.6)  No 1701(95.8) 189(91.3) 1512(96.4) Age when first menstrual period occurred, median [IQR] 12[11,13] 12[11,13] 12[11,13] 0.415 Had regular periods in past 12 months, n (%) 0.265  Yes 1611(89.2) 183(86.7) 1428(89.5)  No 195(10.8) 28(13.3) 167(10.5) Ever treated for a pelvic infection/PID, n (%) 0.004  Yes 93(5.2) 20(9.5) 73(4.6)  No 1702(94.8) 190(90.5) 1512(95.4) Ever use female hormones, n (%) 0.140  Yes 66(3.7) 12(5.7) 54(3.4)  No 1738(96.3) 199(94.3) 1539(96.6) BMI (kg/m 2 ), median [IQR] 28.4[23.5,34.4] 31.5[24.9,39.1] 28[23.4,33.8] < 0.001 BMI (kg/m 2 ) group, n (%) < 0.001  Normal 54(26.2) 524(33.1)  Overweight 34(16.5) 394(24.9)  Obese 118(57.3) 664(42.0) Energy (kcal), median [IQR] 1681[1216,2194] 1651[1151,2110] 1683[1226,2223] 0.175 Vitamin E (mg), median [IQR] 6.3[4.2,9.8] 5.6[4.0,8.7] 6.5[4.2,9.9] 0.021 Vitamin A (mcg), median [IQR] 405.5[224,707] 339[179,573.5] 420[228,724] < 0.001 Alpha-carotene (mcg), median [IQR] 41[7,182.8] 33[4,116] 43[8,192] 0.028 Beta-carotene (mcg), median [IQR] 712[257,2221.8] 659[206.5,1886.5] 719[261.5,2253.5] 0.134 Vitamin D (mcg), median [IQR] 48[16.9,108.4] 32.2[13.9,79] 50.3[18.0,111.5] < 0.001 Zinc (mg), median [IQR] 8.085[5.4,11.5] 7.49[5.1,11.1] 8.2[5.4,11.7] 0.039 Selenium (mcg), median [IQR] 92.3[61.6,127.4] 83.6[52.1,112.3] 93.5[62.1,129.1] 0.002 CDAI, median [IQR] −1.1[−2.9,1.4] −1.9[−3.4,0.2] −1.0[−2.8,1.7] < 0.001 Abbreviations : IQR interquartile range, PID pelvic inflammatory disease, BMI body mass index, kcal kilocalorie, mg milligram, mcg  microgram Characteristics of the participants Abbreviations : IQR interquartile range, PID pelvic inflammatory disease, BMI body mass index, kcal kilocalorie, mg milligram, mcg  microgram Table  2 shows the relationship between CDAI and infertility based on logistic regression analysis. Without adjusting for covariates (Model A), For every unit increase in CDAI, the odds of female infertility decreased by 11% (OR = 0.89, 95%CI: 0.84 to 0.94, P  < 0.001). When CDAI was used as a quartile variable, the likelihood of infertility in the Q4 group was significantly reduced compared with the Q1 group (OR = 0.40, 95%CI: 0.22 to 0.73, P  = 0.003), P trend < 0.001. Table 2 Associations between CDAI and the odds of infertility among women a Model A Model B Model C OR 95% CI p-Value OR 95% CI p-Value OR 95% CI p-Value Continuous b 0.89 0.84–0.94 < 0.001 0.89 0.84–0.94 < 0.001 0.91 0.86–0.97 0.004 Q1 Reference Reference Reference Q2 0.63 0.37–1.07 0.088 0.58 0.34 − 0.10 0.047 0.62 0.34–1.11 0.102 Q3 0.68 0.43–1.07 0.095 0.64 0.40–1.04 0.069 0.72 0.44–1.16 0.167 Q4 0.40 0.22–0.73 0.003 0.41 0.23–0.75 0.005 0.48 0.26–0.90 0.023 p-trend < 0.001 < 0.001 < 0.001 Abbreviations : CI confidence interval, OR odds ratio.  Model A did not adjust for any covariates Model B adjusted for age, race, education level, marital status, the ratio of family income to poverty Model C further adjusted for alcohol use, smoking—cigarette use, and physical activity hypertension, diabetes, reproductive history, had regular periods, treated for a pelvic infection/PID, use female hormones based on Model B a The associations between CDAI and the odds of infertility in women are presented as ORs (95% CI) b ORs represent the change in infertility risk per 1-point increase in CDAI Associations between CDAI and the odds of infertility among women a Abbreviations : CI confidence interval, OR odds ratio. Model A did not adjust for any covariates Model B adjusted for age, race, education level, marital status, the ratio of family income to poverty Model C further adjusted for alcohol use, smoking—cigarette use, and physical activity hypertension, diabetes, reproductive history, had regular periods, treated for a pelvic infection/PID, use female hormones based on Model B a The associations between CDAI and the odds of infertility in women are presented as ORs (95% CI) b ORs represent the change in infertility risk per 1-point increase in CDAI After adjusted for age, race, education level, marital status, and PIR (Model B), the odds of female infertility decreased by 11% for every unit score increase in CDAI (OR = 0.89, 95%CI: 0.84 to 0.94, P  < 0.001). When CDAI was used as a quartile variable, the likelihood in the Q4 group was reduced by 59% (OR = 0.41, 95%CI: 0.23 to 0.75, P  = 0.005), P trend < 0.001. After further adjustment for factors such as drinking, smoking, physical activity, hypertension, diabetes, history of pelvic infection, estrogen use, time of menarche, and menstrual cycles within one year (Model C), the odds of female infertility decreased by 9% for every unit score increase in CDAI (OR = 0.91, 95%CI: 0.86 to 0.97, P  = 0.004). When CDAI was used as a quartile variable, the odds in the Q4 group reduced by 52% (OR = 0.48, 95%CI: 0.26 to 0.90, P  = 0.023), P trend < 0.001. Restricted cubic spline analysis revealed a linear dose-response relationship between CDAI and infertility (P overall = 0.001, P for non-linearity = 0.786), supporting our primary linear model. (Fig.  2 ) Fig. 2 Dose-response relationship between CDAI and female infertility Dose-response relationship between CDAI and female infertility Table  3 shows the association analysis between BMI and female infertility likelihood based on logistic regression analysis. In Model A without adjusting any covariates, when BMI was used as a continuous variable, the odds of female infertility increased by 6% for every unit score increase in BMI (OR = 1.06, 95%CI: 1.03 to 1.09, P  < 0.001). When BMI was used as a quartile variable, the likelihood of infertility in the obese group was significantly increased compared with the normal BMI group (OR = 2.48, 95%CI: 1.48 to 4.14, P  < 0.001), P trend < 0.001. After adjusted for age, race, education level, marital status, and family income and poverty ratio in Model B, the odds of infertility increased significantly for every unit score increase in BMI (OR = 1.06, 95%CI: 1.03 to 1.09, P  < 0.001). When BMI was used as a quartile variable, the odds of infertility in the obese group was significantly increased compared with the normal BMI group (OR = 2.46, 95%CI: 1.44 to 4.20, P  = 0.001), P trend < 0.001. After further adjusted for lifestyle and health-related factors such as alcohol use, smoking, physical activity, hypertension, diabetes, history of pelvic infection, estrogen use, time of menarche, and menstrual cycles within one year (Model C), the odds of infertility increased with each unit score increase in BMI (OR = 1.05, 95%CI: 1.03 to 1.08, P  < 0.001). When BMI was used as a quartile variable, the likelihood of infertility in the obese group was positively correlated with the likelihood of infertility in the normal BMI group (OR = 2.23, 95%CI: 1.30–3.83, P  = 0.005), P trend < 0.001. Additionally, associations between BMI and the odds of infertility among women taking higher levels (Q4) of CDAI and its components was shown in Supplementary Table 1. Table 3 Associations between BMI and the odds of infertility among women a Model A Model B Model C OR 95% CI p-Value OR 95% CI p-Value OR 95% CI p-Value Continuous b 1.06 1.03–1.09 < 0.001 1.06 1.03–1.09 < 0.001 1.050 1.03–1.08 < 0.001 Normal Reference Reference Reference Overweight 0.73 0.42–1.25 0.244 0.65 0.35–1.18 0.148 0.590 0.31–1.09 0.090 Obese 2.48 1.48–4.14 < 0.001 2.46 1.44–4.20 0.001 2.230 1.30–3.83 0.005 p-trend < 0.001 < 0.001 < 0.001 Abbreviations : CI Confidence interval, OR Odds ratio Model A did not adjust for any covariates Model B adjusted for age, race, education level, marital status, the ratio of family income to poverty Model C further adjusted for alcohol use, smoking—cigarette use, and physical activity hypertension, diabetes, reproductive history, had regular periods, treated for a pelvic infection/PID, use female hormones based on Model B a The associations between BMI and the odds of infertility in women are presented as ORs (95% CI) b ORs represent the change in infertility risk per 1 kg/m² increase in BMI Associations between BMI and the odds of infertility among women a Abbreviations : CI Confidence interval, OR Odds ratio Model A did not adjust for any covariates Model B adjusted for age, race, education level, marital status, the ratio of family income to poverty Model C further adjusted for alcohol use, smoking—cigarette use, and physical activity hypertension, diabetes, reproductive history, had regular periods, treated for a pelvic infection/PID, use female hormones based on Model B a The associations between BMI and the odds of infertility in women are presented as ORs (95% CI) b ORs represent the change in infertility risk per 1 kg/m² increase in BMI Figure 3 shows that CDAI has a significant total effect (β=−0.005383, P  = 0.048) and direct effect (β=−0.007600, P  < 0.001) on the odds of female infertility, and BMI showed a significant indirect effect in the above association (indirect effect: −0.000500, P  = 0.012, 95%CI: −0.00110 to −0.00010), and the mediating effect of BMI accounts for 6.2% of the total effect. When the BMI of the subjects 1 year ago was included in the mediation model, the total effect of CDAI on female infertility was significant (β=−0.121513, P  = 0.019), as was the direct effect (β= −0.007743, P  < 0.001). Furthermore, BMI demonstrated a significant indirect effect in the aforementioned association (indirect effect: −0.000521 P  = 0.001, 95% CI: −0.001120 to −0.0000978), with the mediating effect of BMI accounting for 6.3% of the total effect. On the risk difference scale, the mediated effect represented an absolute risk reduction of 0.45% (95% CI: 0.15% to 0.82%) per 1-unit CDAI increase. Fig. 3 Mediating role of BMI in the association between CDAI and female infertility Mediating role of BMI in the association between CDAI and female infertility Supplementary Fig. 2 presents the mediation effect analysis of three different age groups (20–30 years old, 30–40 years old, > 40 years old). In women aged 20–30 years, BMI can mediate the relationship between CDAI and the likelihood of female infertility (indirect effect: −0.000585, P  = 0.042, 95%CI: −0.001625 to −0.000006), accounting for 9.8% of the total effect. In women aged 30–40 years, BMI can mediate the relationship between CDAI and the odds of female infertility (indirect effect: −0.000618, P  = 0.040, 95%CI: −0.001505 to −0.000026), with a mediating effect ratio of 9.96%. However, in women aged > 40 years, the mediating effect of BMI in the relationship between CDAI and the odds of female infertility is not significant. Furthermore, the mediating role of BMI in the association between CDAI and infertility was also validated in those with higher education levels or with middle family income levels (PIR 1.3 to 3.5) (Supplementary Fig. 3–4). In women with regular menstruation, BMI mediated this relationship with an (indirect effect of −0.000447 P  = 0.036, 95% CI: −0.00115 to −0.0000186), accounting for 5.44% of the total effect. For women with irregular menstruation, the (indirect effect was − 0.000887, P  = 0.018, 95% CI: −0.002140 to −0.000104), constituting 11.3% of the total effect. In those who have used female hormones (indirect effect was − 0.000708, P  = 0.008, 95% CI: −0.001466 to −0.000128), representing 7.32% of the total effect. Among women without pelvic infection, BMI mediated the relationship with an (indirect effect of −0.000537, P  = 0.002, 95% CI: −0.001063 to −0.000139), accounting for 6.13%. In women with menarche age ≥ 11 years, the indirect effect was − 0.0057 ( P  = 0.002, 95% CI: −0.001206 to −0.000136), making up 7.97% of the total effect. In education of high school or above, the (indirect effect was − 0.00064 P  = 0.006, 95%CI: −0.001320 to −0.000137), constituting 7.98% of the total effect. However, in women with menarche age < 11 years, not using hormones, and with pelvic infection, BMI did not significantly mediate this relationship. (Supplementary Fig. 5–8) In separate relationship between six antioxidants and infertility, BMI can mediate the relationship between vitamin A, α-carotene, β-carotene and infertility (Supplementary Fig. 9). BMI can partially mediate the relationship between vitamin A and infertility (indirect effect: −0.000003, P  < 0.001, 95%CI: −0.00008 to −0.00000004), with a mediating effect ratio of 6.5%. BMI can also partially mediate the relationship between α-carotene and infertility (indirect effect: −0.000003, P  < 0.001, 95%CI: −0.000003 to −0.000006), with a mediating effect ratio of 39.6%. Moreover, BMI can partially mediate the relationship between β-carotene and infertility (indirect effect β=−0.000001, P  < 0.001, 95%CI: −0.000002 to −0.0000003), with a mediating effect ratio of 22.5%. After imputation for missing values in covariates, we still found that BMI can partially mediate the relationship between CDAI and female infertility (indirect effect: −0.000353, P  = 0.002, 95%CI: (−0.000691, −0.000099), accounting for 10.23% of the total effect, supporting the primary analysis (Supplementary Fig. 10). Sensitivity analyses using nutrient density-adjusted models (antioxidant intake per 1,000 kcal) yielded similar results: indirect effect: −0.00110, P  < 0.001, 95%CI: (−0.002134, −0.000424), accounting for 10.07% of the total effect. (Supplementary Fig. 11). After adjusting for comorbidity burden using CCI, the mediating effect of BMI persisted (indirect effect: −0.0003, 95% CI: −0.0023 to −0.0001), with mediated proportion 5.9% (Supplementary Fig. 12). Sensitivity analysis using BMI measurements from one year prior demonstrated consistent mediation effects (indirect effect: −0.000521, 95% CI: −0.001120 to −0.0000978), supporting the robustness of our findings to temporal variations in BMI assessment (Supplementary Fig. 13). Unmeasured confounding using the “medsens” function revealed that our results were robust to moderate unmeasured confounding (ρ < 0.12) (Supplementary Fig. 14).

This study utilized data from the National Health and Nutrition Examination Survey (NHANES, 2013–2018). NHANES is a national health survey conducted by the Centers for Disease Control and Prevention (CDC) and the National Center for Health Statistics (NCHS) to analyze the health and nutritional status of a nationally representative, non-institutionalized sample of US citizens. The NHANES protocol was approved by the NCHS Research Ethics Review Board, and all participants provided informed consent. From 2013 to 2018, NHANES collected data on a total of 29,400 participants, after excluding males ( n  = 14452), people aged under 20 or over 45 ( n  = 11093), respondents with incomplete data on infertility ( n  = 604), participants with lack of CDAI component information ( n  = 542), and participants with incomplete BMI assessment or any other covariates ( n  = 903). Finally, a total of 1806 participants were included in the analysis (Fig.  1 ). Fig. 1 Flowchart of the participants selection Flowchart of the participants selection Dietary data were collected via two 24-hour recall interviews, and the average intake of antioxidants across both days was used to calculate CDAI. To account for variation in total energy intake (a potential confounder of dietary antioxidant associations), we first adjusted each antioxidant component (vitamin A, C, E, selenium, zinc, carotenoids) for energy using the residual method (kcal) prior to CDAI calculation. This method regresses each antioxidant’s intake on total energy, then uses the residuals (representing energy-adjusted antioxidant intake) in the CDAI formula. In this study, we calculated the CDAI based on six dietary antioxidants: vitamin A, vitamin C, vitamin E, zinc, selenium, and carotenoids, according to the formula for CDAI proposed by Wright et al. [ 11 ]: \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{CDAI}\;=\;\sum_{\mathrm i=1}^{\mathrm n=6}\limits\;\mathrm{Individual}\;\mathrm{intake}\;-\;\mathrm{Mean}/\mathrm{SD}$$\end{document} Infertility was determined based on self-reported response to a specific question on infertility in the reproductive health questionnaire (“Have you ever tried to get pregnant but failed to get pregnant for at least one year?”). Women who reported “yes” were considered to have a history of infertility. Infertility was defined as the inability to achieve pregnancy after one year of unprotected intercourse, consistent with clinical guidelines and prior studies [ 23 ]. Women who responded “yes” were classified as the “infertile group”, while those who responded “no” were classified as the “non-infertile group”. While NHANES collects data on infertility history, the survey does not systematically capture detailed etiological information that would allow for stratification by specific causes (e.g., tubal factors, ovulatory disorders, or endometriosis). Therefore, our analysis utilizes the broad definition of infertility as the inability to conceive after 12 months of unprotected intercourse. Subjects were measured for body height and body weight at the mobile medical examination center according to standardized procedures. During the assessment, participants only had to wear minimal clothing, including underwear, disposable paper robes, and foam slippers. BMI was calculated using the formula: BMI = weight (kg)/height² (m²). According to the WHO classification, BMI of 18.5–24.9 kg/m² was defined as the normal range, BMI 25.0–29.9 kg/m² was defined as overweight, and BMI ≥ 30.0 kg/m² was defined as obesity [ 20 ]. In addition, all subjects provided self-reported weight information from 1 year ago, which was used to calculate BMI from 1 year ago. Based on previous studies on infertility [ 3 , 10 ], we included the following covariates. Sociodemographic factors included age (20–30, 30–40, >40 years old), race/ethnicity (non-Hispanic white, non-Hispanic black, Mexican American, and other race/ethnicity), family income-to-poverty ratio (PIR), education (less than high school, high school or above), marital status (married/living with a partner, single), sufficient physical activity (yes/no), smoking (yes/no), alcohol consumption (yes/no), hypertension (yes/no), diabetes (yes/no), history of pelvic infection (yes/no), estrogen use (yes/no), contraceptive use (yes/no), pregnancy and childbirth history (yes/no), time of menarche, and menstrual cycle within one year. Total physical activity was estimated by adding the minutes of moderate-intensity leisure activities to twice the minutes of vigorous-intensity leisure activities. According to the 2018 Physical Activity Guidelines for Americans, participants were classified as follows: “inactive” if there was no leisure-time physical activity (LTPA) in the previous week; “insufficiently active” if LTPA was greater than 0 min/wk but less than 150 min/wk; and “sufficiently active” if LTPA was 150 min/wk or more [ 24 , 25 ], we divided the participants into the sufficient physical activity (LTPA ≥ 150 min/wk) and insufficient physical activity group (inactive or LTPA < 150 min/wk). The PIR was used to measure family income and was divided into three categories (3.5). Participants’ smoking status was assessed with “Have you smoked at least 100 cigarettes in your lifetime?”. Participants’ drinking status was recorded with the answer to “Have you ever consumed 12 glasses of alcohol in any one year?” Given the lack of direct measures for reproductive conditions, we included self-reported “history of pelvic infection”, “irregular menstruation” and “estrogen use” as proxies for underlying reproductive dysfunction (e.g., irregular menstruation may correlate with PCOS-related hormonal imbalances). These variables were incorporated to minimize residual confounding to the extent possible with available NHANES data. NHANES 2013–2018 includes six survey cycles; to ensure consistency in dietary data collection across cycles, we harmonized the 24-hour recall protocols (e.g., standardized food coding, portion size estimation) using NHANES’ cycle-specific documentation. This harmonization ensured that antioxidant intake estimates were comparable across all study participants, regardless of survey cycle. First, the subjects were divided into infertile group and non-infertile group according to the presence of infertility. All numerical variables were tested for normality. Variables that conformed to normal distribution were described as means (standard deviations, SD), and two-sample t-tests were used for inter-group comparisons. Variables that did not conform to normal distribution were described as medians [interquartile ranges, IQR], and Mann-Whitney U test was used for inter-group comparisons. Categorical data were described as frequencies (percentages), and Chi-square tests were used for inter-group comparisons. Three weighted logistic regression models were used to explore the independent associations of CDAI and BMI with women infertility. Model A was the baseline model without adjusting for any covariates. Model B was adjusted for demographic variables (age, race, education level, marital status) and socioeconomic indicators (PIR) on the basis of Model (A) Model C further included lifestyle factors (drinking, smoking, physical activity), history of metabolic diseases (hypertension, diabetes), reproductive health-related variables (history of pelvic infection, time of menarche, and menstrual cycles within one year,), and exogenous hormone use (female hormone drug use history) on the basis of Model (B) Directed acyclic graph (DAG) was used to illustrate assumed causal relationships and covariate adjustment strategy (Supplementary Fig. 1). We have conducted restricted cubic spline analysis with three knots to examine potential nonlinear dose-response relationships between CDAI and infertility. Further, fully adjusted mediation analysis was used to explore the mediating effect of BMI in the relationship between CDAI and infertility. The indirect effect and 95% CI were estimated based on the Bootstrap method. The 95% CI of the indirect effect was defined as significant if it did not include 0. In addition, we conducted subgroup analysis on women of different age groups to explore whether the mediating effect of BMI differed among women of different age groups. Subgroups analyses were also conducted based on the socioeconomic conditions (education, PIR) and reproductive factors including use of estrogen, regular menstruation, age of menarche (< 11, ≥ 11 years old) [ 26 ], and pelvic infection. Finally, we further explored the mediating effect of BMI in the relationship between the six antioxidants components of CDAI and infertility. To address the potential for selection bias due to participant exclusion, we performed sensitivity analyses using multiple imputation (MI) for missing covariates (e.g., age, race, income). Missing covariates were imputed using chained equations (MICE) with 20 imputations, incorporating all study variables. Moreover, additional sensitivity analyses were performed using nutrient density models (intake per 1,000 kcal). To address potential residual confounding by chronic conditions, we calculated the Charlson Comorbidity Index (CCI) [ 27 ] for each participant based on self-reported medical conditions. The CCI was included as a covariate in sensitivity analyses. We also performed sensitivity analysis for unmeasured confounding using the “medsens” function. Primary analyses employed the residual method for energy adjustment (kcal). Sensitivity analyses used nutrient densities (intake per 1,000 kcal). Finally, to address potential measurement timing issues, we utilized self-reported BMI from one year prior to survey participation, available for a subset of participants (n = 1753). This allowed us to examine whether BMI measurements preceding infertility diagnosis yielded similar results. All statistical analyses were performed using R (4.4.1), and the mediation analysis was based on the R package “mediation”. Mediation effects are reported on the log-odds scale. For clinical interpretability, we also calculated risk differences using the mediation package’s ‘transform’ option to convert to the probability scale. Analyses incorporated NHANES sampling weights to account for complex survey design, using the R package “survey” to ensure nationally representative estimates. All analyses incorporated NHANES sampling weights (WTDR2D), strata (SDMVSTRA), and primary sampling units (SDMVPSU) using the R “survey” package. In primary analysis, all P values ​​were defined as significant when the two-sided test was < 0.05. For exploratory subgroup analyses, we applied Bonferroni correction, considering P < 0.008 as significant for the six antioxidant components.

Based on a nationally representative sample, this study explored the relationship between CDAI and the likelihood of female infertility, as well as the mediating role of BMI. The results showed that higher levels of CDAI were associated with a reduced odds of female infertility, while higher BMI levels were associated with an increased likelihood of female infertility, and CDAI may affect female infertility by affecting BMI. The results of subgroup analysis showed that the mediating effect was particularly evident in infertile women at the peak of childbearing age (20–40 years old), with regular menstruation, without pelvic infection, at menarche age ≥ 11 years old, and using female hormones. It is worth noting that among specific antioxidants, BMI can mediate the relationship between energy, vitamin A, α-carotene, β-carotene and infertility. This study showed that women with higher CDAI levels had a significantly lower likelihood of infertility, which is consistent with previous studies. Oxidative stress (OS) is a state characterized by an imbalance between pro-oxidant molecules (including reactive oxygen and nitrogen species) and antioxidant defenses. It has been identified as playing a key role in the pathogenesis of female subfertility [ 28 ]. The negative effects of OS on female reproductive health have been clearly manifested in follicular development disorders and ovarian dysfunction. A study by Smits et al. on the assisted reproductive population showed that antioxidants can improve the reproductive function of infertile patients by antagonizing oxidative stress [ 29 ]. Iranian reproductive health cohort showed that insufficient intake of key antioxidant nutrients (potassium, magnesium, copper, vitamin C and dietary fiber) may reduce the likelihood of infertility [ 30 ]. Based on the above mechanism, antioxidants have both antioxidation and anti-inflammatory properties and show unique clinical value in key reproductive links such as improving ovarian reserve function and regulating endometrial receptivity. On the other hand, consistent with previous works, obese women have a significantly increased odds of infertility. A prospective cohort study confirmed that in women of childbearing age with patent fallopian tubes and normal ovulation, every 1 unit score increase in BMI >29 kg/m² would lead to a 4% decrease in the natural pregnancy rate [ 31 ]. In a survey of obese people with normal metabolism conducted by Tang et al., the results showed that metabolically healthy obesity is also associated with an increased prevalence of infertility in women. Regardless of metabolic health status, obesity itself is associated with a higher prevalence of infertility [ 32 ]. Its pathological mechanism mainly involves a triple action pathway of hypothalamic-pituitary-ovarian axis dysfunction, oocyte mitochondrial dysfunction, and impaired endometrial decidualization [ 21 , 33 ]. Specifically, obesity-induced metabolic disorders (dyslipidemia, imbalance of cholesterol homeostasis) and reproductive dysfunction (hypothalamic-pituitary-gonadal axis disorder, reduced uterine receptivity) form a bidirectional pathological association [ 34 – 36 ]. Our primary contribution lies in demonstrating BMI’s mediating role, particularly in women aged 20–40 years. In the report of Bhardwaj et al., clarified that dietary vitamins, natural antioxidant ingredients such as carotenoids, and active substances from medicinal plants showed biological effects on improving female fertility [ 37 ]. On the other hand, in a cross-sectional study of 25,553 obese people included in Wang et al., CDAI was inversely correlated with overall obesity and abdominal obesity. Maintaining a high CDAI level may be a protective factor against obesity-related inflammation [ 38 ]. The study of Hosseini et al. showed that low levels of antioxidant trace elements (carotenoids, vitamins E and C, zinc, magnesium, and selenium) were associated with obesity [ 39 ]. A cross-sectional study of Mexican women showed that vitamins A was associated with obesity, fat deposition, and leptin concentrations [ 40 ]. All of these studies support our findings that higher CDAI levels may reduce the likelihood of infertility in women by reducing obesity levels, thereby increasing the probability of pregnancy. However, it is worth noting that the mediating effect of obesity accounted for only 6.2%, indicating that the mediating effect of obesity can only explain part of the association between CDAI and infertility, and there may be other mediating factors that play a synergistic role, such as imbalance between reactive oxygen species (ROS) and antioxidants, leading to cell damage [ 41 ]. This modest proportion also aligns with infertility’s multifactorial etiology. Even small mediation effects may be clinically meaningful given infertility’s complex pathophysiology. The finding suggests weight management could partially enhance fertility benefits from antioxidant-rich diets, particularly in younger women. Our findings suggest combined nutritional and weight management strategies may optimize fertility outcomes. Clinicians could consider: (1) dietary counseling to increase antioxidant intake, (2) BMI screening and weight loss interventions for overweight/obese women seeking conception, particularly those aged 20–40 years. Public health policies promoting both healthy diets and weight maintenance may help reduce population-level infertility burdens. Combining weight loss with an antioxidant diet may enhance the benefits of intervention [ 42 ]. In the future, randomized controlled trials targeting combined interventions are needed to verify the long-term effects in the infertile population. This study offers several notable advantages. Firstly, it is the first to demonstrate that BMI played a mediating role in the relationship between CDAI and infertility, providing a targeted intervention pathway for enhancing reproductive health management in women. Secondly, we accounted for multiple confounding factors in our theoretical model, and our study population is representative of the broader national demographic. However, this study also has limitations. First, as NHANES is fundamentally designed as a repeated cross-sectional survey rather than a longitudinal cohort study, our analysis is inherently limited to examining associations rather than establishing causal relationships. Our findings should therefore be interpreted as generating hypotheses about potential mechanisms rather than providing definitive causal evidence. In addition, while BMI measurements may not perfectly align with the period of infertility onset, our sensitivity analysis using BMI data from one year prior yielded consistent results. Nevertheless, the cross-sectional design remains an inherent limitation for causal inference. To address the causal questions raised by our findings, we recommend prospective cohort studies with repeated measures of dietary antioxidants and BMI prior to infertility onset. Ideally, randomized controlled trials testing combined antioxidant supplementation and weight management interventions in women with infertility could provide more definitive evidence. Such trials could specifically target women with low CDAI and high BMI to maximize potential benefits and establish causal efficacy. Another key limitation is the absence of data on specific reproductive conditions (e.g., PCOS, endometriosis, thyroid disease) and detailed parity in NHANES. These factors are established correlates of both BMI regulation and infertility risk, and their omission may introduce residual confounding. Future studies using datasets with comprehensive reproductive phenotyping should validate our findings while adjusting for these variables. Additionally, the dietary assessment was based on two-day 24-hour recall questionnaire, which may not fully capture the subjects’ long-term dietary habits. A more comprehensive and prolonged dietary evaluation would be beneficial. Third, while we recognize the clinical importance of infertility subtyping, NHANES does not collect detailed information on specific etiologies of infertility. The inability to stratify by causes such as tubal factors, ovulatory disorders, or endometriosis represents a limitation of our study. Future studies with more detailed phenotyping of infertility causes would be valuable to explore whether the associations we observed vary by infertility etiology. Finally, our findings are specific to the American population and may not be generalizable to other populations worldwide. Future research should include diverse populations from various countries and ethnic backgrounds to further validate our results.

This study demonstrates that women with higher CDAI levels experience a lower prevalence of infertility. Additionally, we provide novel evidence of an indirect association involving BMI in the relationship between CDAI and female infertility. These findings enhance our understanding of potential associative pathways linking antioxidant-rich dietary patterns to the likelihood of infertility. Furthermore, they offer valuable insights for developing targeted prevention strategies aimed at reducing female infertility.

Infertility is the inability to become pregnant after regular unprotected sexual intercourse for 12 months [ 1 ]. Female infertility has become a global health challenge [ 2 , 3 ]. It is estimated that infertility affects up to 186 million people worldwide [ 2 ]. Infertility has a profound impact on individuals and their families, and imposes a huge economic burden on society [ 3 ]. In the United States, the prevalence of infertility in women of childbearing age is approximately 15.5% [ 4 ]. The causes of infertility are multifactorial, including reproductive system disorders or diseases, nutritional deficiencies, environmental factors, genetic factors, and diet and physical activity lifestyle [ 5 – 10 ]. The Composite Dietary Antioxidant Index (CDAI) has been recently introduced as an innovative metric to assess the antioxidant potential of an individual’s dietary intake [ 11 ]. This index provides a comprehensive score that encompasses a range of dietary antioxidants, including vitamins A, C, and E, as well as selenium, zinc, and carotenoids [ 11 ]. In previous studies, CDAI was a protective factor for depression, cardiovascular disease, HPV infection, colorectal cancer and metabolic disease [ 12 – 16 ]. Meanwhile, preliminary evidence showed that CDAI also has a significant impact on women’s reproductive health [ 17 ]. Li et al. conducted a cross-sectional analysis of 8263 American women aged 20–45 years and found that infertile women had lower CDAI levels compared with fertile women [ 18 ]. Shao et al. found in a population-based survey that a higher intake of dietary antioxidants may be associated with a lower prevalence of female infertility [ 19 ]. However, further evidence on the association between CDAI and female infertility, and the potential mechanisms in the relationship between CDAI and female infertility are still unclear. According to the Centers for Disease Control and Prevention (CDC) in the United States, the obesity rate (BMI ≥ 30 kg/m²) among women of childbearing age is approximately 29% [ 20 ]. Obesity causes multi-level damage to the reproductive system by interfering with the function of the hypothalamus-pituitary-ovarian axis, impairing oocyte quality, and reducing endometrial receptivity, while significantly increasing the prevalence of adverse pregnancy outcomes [ 21 ]. A nationwide cross-sectional study by Yang et al. also showed that obesity was correlated with increased prevalence of female infertility [ 22 ]. However, whether obesity can mediate the relationship between CDAI and female infertility has not been studied. To fill the gap in the above works, this study aims to explore the relationship between CDAI and infertility in American women based on a large nationally representative sample, as well as the mediating role of obesity. Our findings may help further understand how antioxidant diets affect female infertility and provide precise guidance for infertility prevention in female population.

Supplementary Material 1 Supplementary Material 1