Association between composite dietary antioxidant index and pulmonary function in middle-aged and older adults: results from two population-based studies Authors: Rui Qiang Li, Yu Hui Lin, Jun Tao Kan, Jiao Wang, Ya Li Jin, Wei Sen Zhang, Lin Xu, Emma Yun-Zhi Huang Published: 2026 ABSTRACT AbstractBackgroundThe decline in pulmonary function among older adults is a widespread issue, lacking specific dietary guidelines for optimising respiratory health.MethodsData from the Guangzhou Biobank Cohort Study (GBCS) and National Health and Nutrition Examination Survey (NHANES) were analysed to assess the association of the Composite Dietary Antioxidant Index (CDAI) with key pulmonary function indicators in individuals aged 50 and older.ResultsThe study included 8640 participants from GBCS and 4029 from NHANES, with a mean age of 62.5 years. After adjusting for confounders, higher CDAI scores were significantly linked to increased forced expiratory volume in 1 s (FEV1) and FEV1 percentage predicted (FEV1pp) in GBCS (β=0.004 L, 95% CI 0.001 to 0.008; β=0.193%, 95% CI 0.045% to 0.342%). Similar trends were observed in NHANES for FEV1pp (adjusted β=0.353%, 95% CI 0.059% to 0.646%). Subgroup analyses showed that women exhibited consistent positive associations between CDAI and both FEV1pp and forced vital capacity (FVC) percentage predicted (FVCpp) in both cohorts. In GBCS, stronger associations were found for participants over 65 years.ConclusionsHigher CDAI scores correlate with improved pulmonary function, particularly in women and older adults, underscoring the need for dietary recommendations that emphasise antioxidant-rich foods to enhance respiratory health and support healthy ageing. WHAT IS ALREADY KNOWN ON THIS TOPIC Oxidative stress is a critical driver of age-related pulmonary decline and respiratory diseases, yet most research has focused on the effects of individual antioxidants rather than the synergistic impact of overall dietary antioxidant capacity. Oxidative stress is a critical driver of age-related pulmonary decline and respiratory diseases, yet most research has focused on the effects of individual antioxidants rather than the synergistic impact of overall dietary antioxidant capacity. WHAT THIS STUDY ADDS By analysing data from two large-scale, multi-ethnic cohorts (Guangzhou Biobank Cohort Study and National Health and Nutrition Examination Survey), this study provides the first evidence that higher Composite Dietary Antioxidant Index scores are significantly associated with improved pulmonary function indicators, with these protective associations being most pronounced in women and individuals aged 65 and older. By analysing data from two large-scale, multi-ethnic cohorts (Guangzhou Biobank Cohort Study and National Health and Nutrition Examination Survey), this study provides the first evidence that higher Composite Dietary Antioxidant Index scores are significantly associated with improved pulmonary function indicators, with these protective associations being most pronounced in women and individuals aged 65 and older. HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY These findings suggest that public health strategies and dietary guidelines should emphasise holistic, antioxidant-rich dietary patterns rather than single-nutrient supplementation to effectively mitigate age-related respiratory decline and support healthy ageing. These findings suggest that public health strategies and dietary guidelines should emphasise holistic, antioxidant-rich dietary patterns rather than single-nutrient supplementation to effectively mitigate age-related respiratory decline and support healthy ageing. Introduction The ageing population faces an increased burden of pulmonary diseases such as chronic obstructive pulmonary disease (COPD) and lung cancer, which significantly impact health and quality of life and impose considerable economic burdens.1 2 Globally, chronic respiratory diseases affected approximately 569.2 million people and caused 4.2 million deaths in 2023, remaining the third leading cause of death worldwide.3 The absolute burden continues to grow due to population ageing and demographic changes, despite declines in age-standardised rates over recent decades.4 The decline in pulmonary function in older adults is particularly concerning, as it impairs respiratory capacity and triggers systemic inflammation, elevating the risk of cardiovascular diseases such as atherosclerosis, myocardial infarction and stroke.5 6 In addressing the respiratory dysfunction related to ageing, the role of diet, particularly dietary antioxidants, has garnered growing interest.7 8 Oxidative stress is involved in the aetiology of numerous chronic diseases, and dietary antioxidants have been shown to mitigate these risks by neutralising reactive oxygen species (ROS) and bolstering the endogenous antioxidant defences.9 10 The Composite Dietary Antioxidant Index (CDAI) was proposed to evaluate the overall impact of antioxidants on health outcomes, encapsulating a spectrum of dietary antioxidants, including vitamins (A, C, E), minerals (selenium, zinc, manganese) and phytochemicals (carotenoids).11 The CDAI is calculated by summing the standardised intakes of these components, with a higher score reflecting a greater overall antioxidant capacity. Unlike single-nutrient approaches, the CDAI captures the synergistic effects of multiple antioxidants, offering a more comprehensive assessment of the diet’s ability to combat oxidative stress.10 Although the potential benefits of dietary antioxidants on respiratory health have been indicated, most studies have focused on the effects of individual antioxidants.12–14 Previous studies indicated that the combined effects of vitamin C, vitamin E and carotenoids were more effective in improving pulmonary function than single components.15 16 Moreover, intervention with compound nutrients significantly improved pulmonary function in patients with COPD, while supplementation with single vitamins showed no significant effect.17 However, these interventional studies were often short-term and might not fully capture the benefits of antioxidants from a regular diet.18–20 Observational studies may provide complementary evidence to support dietary guidelines for the general population. Despite these potential benefits, significant research gaps remain in the current literature. Most observational studies have focused on individual nutrients rather than the overall dietary antioxidant capacity. Specifically, the association between the CDAI and pulmonary function has not been adequately explored in large-scale populations, particularly among older adults who are most vulnerable to age-related pulmonary decline. Additionally, there is a lack of cross-cultural evidence to determine whether these associations are consistent across populations with diverse ethnic backgrounds and dietary patterns. Therefore, this study aims to examine the associations of the consumption of dietary antioxidants, measured using CDAI, with various pulmonary function indicators including forced expiratory volume in 1 s (FEV1), FEV1 percentage predicted (FEV1pp), forced vital capacity (FVC), FVC percentage predicted (FVCpp) and FEV1/FVC ratio, taking advantage of data obtained from two well-organised large-scale studies: the cross-sectional National Health and Nutrition Examination Survey (NHANES) conducted in the USA and the prospective Guangzhou Biobank Cohort Study (GBCS) conducted in China. Materials and methods Study samples The GBCS is a collaborative project involving the Guangzhou Twelfth People’s Hospital, the University of Hong Kong and the University of Birmingham, UK.21 Participants were recruited from the Guangzhou Health and Happiness Association for the Respectable Elders, a community organisation with branches in all ten districts of Guangzhou. The study targets Guangzhou residents aged 50 and above, with a nominal monthly fee of 4 RMB (approximately 50 US cents). Baseline information was collected through face-to-face computer-assisted interviews conducted by trained nurses from Guangzhou Twelfth People’s Hospital. Data collection includes demographic characteristics, lifestyle factors, family and personal medical history and anthropometric measurements following a standardised protocol. The inclusion criteria were participants aged 50 years or older who had completed the baseline survey. Exclusion criteria included: (1) missing data on lung function parameters; (2) missing nutrient data; (3) lack of essential baseline status information and (4) self-reported history of asthma. After applying these criteria, a total of 8640 participants were included in the final analysis (online supplemental figure 1A). The NHANES is a recurring survey overseen by the National Center for Health Statistics (NCHS) in the USA, with the primary objective of evaluating the health and nutritional status of the American populace.22 This survey furnishes nationally representative data encompassing a range of domains, including but not limited to health status, dietary patterns, nutritional intake, lifestyle choices and environmental influences, and its research methodology has received approval from the Research Ethics Review Board of the NCHS. In the current study, data from the years 2007–2008, 2009–2010 and 2011–2012 were used. To ensure comparability, participants aged 50 years or older were included. Exclusion criteria were: (1) missing lung function data; (2) missing nutrient data; (3) missing baseline status data and (4) a history of asthma. In total, 4029 participants were included in the analysis (online supplemental figure 1B). Assessment of CDAI The CDAI was assessed in the GBCS and the NHANES using validated methodologies to estimate the overall antioxidant potential of diets.11 23 24 In the GBCS, the CDAI was determined by summing the antioxidant scores of various foods consumed. The calculation of CDAI involved standardising the intake of manganese, selenium, zinc and vitamins A, C and E. This was achieved by subtracting the gender-specific mean intake and dividing by the gender-specific SD, followed by summing these standardised intakes.25 26 Dietary intake information, representing the participants’ habitual dietary intake over the past 7 days, was obtained through a validated Food Frequency Questionnaire.27 Using data from the Chinese Food Composition Tables, the nutrient content of food items was quantified by multiplying the consumption of each item by its standard portion size of 100 g. The nutrients quantified include energy (kcal), vitamin A (μg RE), vitamin C (mg), vitamin E (mg), zinc (mg), selenium (μg) and manganese (mg). In the NHANES, the CDAI was calculated based on the intake of zinc, selenium, carotenoids and vitamins A, C and E. Each of these dietary antioxidants was standardised by subtracting the gender-specific mean and dividing by the gender-specific SD, and the CDAI was then derived by summing these standardised intakes.26 Dietary information was collected using 24-hour dietary recalls, which capture the actual intake during the 24-hour period immediately prior to the interview. If two recalls were conducted, data from the first recall was used for analysis. Participants provided detailed descriptions of all food items, beverages and water consumed within the previous day, including types, amounts, timings and sources. The nutritional analysis was performed using the food composition database maintained by the US Department of Agriculture. The initial use of different antioxidant components in calculating the CDAI was due to the inherent differences in dietary data between the GBCS and NHANES datasets. To optimise data utilisation, our primary analysis included all CDAI-related antioxidant components documented in each dataset. Potential confounders We selected covariates theoretically related to CDAI and pulmonary function parameters but do not lie on the causal pathway connecting the two (ie, potential confounding factors). In the GBCS, these covariates included age (<65 years, ≥65 years), sex (men, women), education (primary or below, middle school, college or above), occupation (manual, non-manual, others), personal annual income, CNY/year (<10 000, 10 000–14 999, ≥15 000, Don’t know), body mass index (BMI) (<18.5 kg/m2, 18.5–24.9 kg/m2, 25.0–27.4 kg/m2, ≥27.5 kg/m2), energy, smoking status (never, former, current), physical activity (inactive, moderate, active), self-reported diabetes (no, yes), self-reported hypertension (no, yes) and self-rated health (poor, good). In the NHANES, the covariates were age, sex, education, race, energy, poverty income ratio (PIR), BMI (<18.5 kg/m2, 18.5–24.9 kg/m2, 25.0–30.0 kg/m2, ≥30.0 kg/m2), cotinine, physical activity, self-rated health (no, yes), self-reported diabetes (no, yes) and self-reported hypertension (no, yes). Physical activity was evaluated using a physical activity questionnaire, which included questions such as ‘Have you done any vigorous-intensity activities in the past 30 days?’ and ‘Have you done any moderate-intensity activities in the past 30 days?’ Study outcomes In the GBCS, spirometry procedures were conducted following established protocols.28 Specifically, at least three manoeuvres were performed without bronchodilators to obtain optimal measurements of forced expiratory volume in 1 s (FEV1), FVC and the FEV1/FVC ratio. The predicted values for FEV1 and FVC were calculated based on norms for the Chinese population.29 In the NHANES, spirometry assessments were conducted using Ohio 822/827 dry-rolling volume seal spirometers prior to bronchodilator administration, adhering to the guidelines of the American Thoracic Society.30 Measurements included FEV1 and FVC, with only manoeuvres meeting or exceeding the American Thoracic Society quality grade C (indicating at least partial adherence to standards) being included in the analysis. Original measurements were converted to percentage-predicted values using the Hankinson equations.31 Statistical analysis Characteristics of the study population were described using frequency (percentage) for categorical variables and mean±SD and median (IQR) for continuous variables in both the GBCS and NHANES. The CDAI was treated as a continuous variable in the linear regression models. A series of generalised linear regression models were constructed to estimate the β coefficients and 95% CIs between CDAI and CDAI components and different pulmonary function-related outcomes (including FEV1, FEV1pp, FVC, FVCpp and FEV1/FVC). These models were adjusted for age, sex, education, race, energy, PIR, BMI, cotinine levels, physical activity, self-rated health and self-reported hypertension in the NHANES; and for age, sex, education, occupation, energy, personal annual income, BMI, smoking status, physical activity, self-rated health, self-reported diabetes and self-reported hypertension in the GBCS. As per the NHANES analysis guidelines, all analyses were conducted with the integration of sample dietary weights.32 To further visualise the relationship between CDAI and each pulmonary function indicator, restricted cubic spline (RCS) was also performed. Furthermore, we examined potential age and sex-related interactions in the association between CDAI and pulmonary function indicators. Furthermore, we employed the Extreme Gradient Boosting (XGBoost) machine learning algorithm model to predict the relative importance of CDAI and its components on pulmonary function parameters. The XGBoost model was used to analyse the contribution of each variable to pulmonary function parameters using SHapley Additive exPlanations (SHAP) values. SHAP, rooted in cooperative game theory, provides a unified and mathematically consistent framework for model interpretability. Notably, SHAP values address the bias of traditional importance metrics towards continuous variables and offer both global and local interpretability. Additionally, using the penalty spline method, an XGBoost was constructed with a fully adjusted smooth curve to explore the potential association of the CDAI and CDAI components with pulmonary function parameters.33 In this study, the dataset was randomly split into a training set (70%) and an independent test set (30%). All variable selection and model training were performed solely on the training set. The test set was held out for final evaluation to assess the model’s generalisation performance. Variable selection in the XGBoost model was guided by their importance determined through Variance Inflation Factor and Relief-F analysis. The model parameters were set to a maximum of 50 iterations, a tree depth of 10 and a learning rate (eta) of 0.1. Cross-validation was employed in each boosting iteration to prevent overfitting. All data analyses were conducted using R V.4.3.2 with a significance level set at 0.05. Results Demographic characteristics The study included 8640 participants from the GBCS and 4029 participants from the NHANES, with mean ages (SD) of 62.5 (6.8) years and 62.5 (8.1) years, respectively. Compared with NHANES, the GBCS cohort had a lower proportion of participants aged 65 years and above (37.9% vs 39.1%), men (27.4% vs 50.2%) and participants with a college education or higher (9.2% vs 52.1%). Conversely, the GBCS participants had a higher prevalence of normal weight (62.6% vs 22.1%), being physically active (42.1% vs 12.5%), self-reported good health status (84.2% vs 76.8%) and absence of hypertension (70.5% vs 50.9%). The CDAI showed statistically significant differences across various participant characteristics, including sex, age, race, education level, occupation, household income, smoking status, physical activity, self-reported health status and self-reported hypertension status (P ranging from <0.001 to 0.049) (table 1). Table 1Characteristics of participants aged 50 years and above in the Guangzhou Biobank Cohort Study (N=8640) and NHANES (N=4029)CharacteristicsGBCSNHANESStudy population, N (%)CDAI,median (IQR)Study population,N (%)CDAI,median (IQR)Age, years <655369 (62.1)−0.49 (−2.90, 2.18)2454 (60.9)−0.50 (−2.43, 2.04) ≥653271 (37.9)−0.88 (−3.06, 2.08)1575 (39.1)−0.97 (−2.70, 1.17) P value0.019<0.001Sex Men2364 (27.4)0.25 (−2.23, 3.31)2021 (50.2)0.11 (−1.99, 2.67) Women6276 (72.6)−0.96 (−3.18, 1.73)2008 (49.8)−1.35 (−3.00, 0.61) P value<0.001<0.001Race/ethnicity Mexican American––553 (13.7)−0.85 (−2.76, 1.29) Other Hispanic––389 (9.7)−1.19 (−2.90, 1.06) Non-Hispanic White––2031 (50.4)−0.44 (−2.28, 1.98) Non-Hispanic Black––820 (20.4)−0.93 (−3.03, 1.51) Other/multiracial––236 (5.8)−0.40 (−2.51, 1.99) P value<0.001Education Primary or below3844 (44.5)−1.16 (−3.39, 1.50)466 (11.6)−1.69 (−3.29, 0.71) Middle school4005 (46.4)−0.28 (−2.59, 2.58)1464 (36.3)−1.01 (−2.80, 1.22) College or above791 (9.2)0.05 (−2.24, 3.35)2099 (52.1)−0.23 (−2.07, 2.35) P value<0.001<0.001Occupation Manual5183 (60.0)−0.95 (−3.18, 1.75)–– Non-manual2125 (24.6)0.16 (−2.42, 3.14)–– Others1332 (15.4)−0.59 (−2.83, 2.01)–– P value<0.001–Personal annual income, CNY/year <10 0003121 (36.1)−1.08 (−3.34, 1.66)–– 10 000–14 9993675 (42.5)−0.53 (−2.80, 2.15)–– ≥15 0001443 (16.7)0.04 (−2.35, 3.05)–– Don’t know401 (4.6)−0.53 (−2.90, 2.54)–– P value<0.001––Poverty income ratio <1––569 (14.1)−1.70 (−3.13, 0.75) 1–3––1607 (39.9)−1.09 (−2.80, 1.13) >3––1853 (46.0)−0.11 (−1.95, 2.32) P value<0.001BMI, kg/m2 Underweight369 (4.3)−0.92 (−3.14, 1.26)31 (0.8)−1.70 (−3.16, −0.01) Normal weight5410 (62.6)−0.65 (−2.90, 2.11)891 (22.1)−0.56 (−2.52, 2.15) Overweight1814 (21.0)−0.63 (−2.99, 2.37)1446 (35.9)−0.52 (−2.42, 1.54) Obese1047 (12.1)−0.46 (−3.05, 2.40)1661 (41.2)−0.77 (−2.62, 1.58) P value0.0940.085Smoking status Never6939 (80.3)−0.73 (−3.01, 2.02)–– Former856 (9.9)0.08 (−2.40, 3.06)–– Current845 (9.8)−0.49 (−2.86, 2.18)–– P value<0.001Cotinine Unexposure (<0.015 ng/mL)––1082 (26.9)−0.51 (−2.23, 1.88) Exposure (≥0.015 ng/mL)––2947 (73.1)−0.75 (−2.69, 1.60) P value0.002Physical activity Inactive3635 (42.1)−0.01 (−2.45, 2.87)504 (12.5)−0.03 (−2.21, 2.37) Moderate4237 (49.0)−0.91 (−3.21, 1.81)1103 (27.4)−0.64 (−2.42, 1.75) Active768 (8.9)−1.78 (−3.57, 0.55)2422 (60.1)−0.76 (−2.72, 1.50) P value<0.001<0.001Self-rated health Poor1364 (15.8)−0.65 (−2.94, 2.14)933 (23.2)−1.29 (−3.08, 0.85) Good7276 (84.2)−0.59 (−2.94, 2.17)3096 (76.8)−0.49 (−2.39, 1.88) P value0.8270.058Self-reported hypertension No6092 (70.5)−0.48 (−2.85, 2.24)2050 (50.9)−0.54 (−2.45, 1.87) Yes2548 (29.5)−0.72 (−3.01, 2.12)1979 (49.1)−0.78 (−2.61, 1.48) P value0.0480.014Self-reported diabetes No7893 (91.4)−0.57 (−2.96, 2.38)3315 (82.3)−0.58 (−2.45, 1.81) Yes747 (8.6)−0.64 (−2.94, 2.13)714 (17.7)−1.03 (−2.94, 1.19) P value0.058<0.001BMI, body mass index; CDAI, Composite Dietary Antioxidant Index; GBCS, Guangzhou Biobank Cohort Study; NHANES, National Health and Nutrition Examination Survey. Characteristics of participants aged 50 years and above in the Guangzhou Biobank Cohort Study (N=8640) and NHANES (N=4029) BMI, body mass index; CDAI, Composite Dietary Antioxidant Index; GBCS, Guangzhou Biobank Cohort Study; NHANES, National Health and Nutrition Examination Survey. Association between CDAI and pulmonary function parameters After adjusting for age, sex, education, occupation, energy, personal annual income, BMI, smoking status, physical activity, self-rated health and self-reported hypertension, the CDAI was significantly associated with increased FEV1 (0.004 L, 95% CI 0.001 to 0.008) and FEV1pp (0.193%, 95% CI 0.045% to 0.342%) (both p<0.05) in the GBCS. After similar adjustment, a significant positive association was observed between CDAI and FEV1pp (0.353%, 95% CI 0.059% to 0.646%) in the NHANES (table 2). No significant interactions were found between CDAI and education, occupation, personal annual income, BMI, smoking status, physical activity, self-rated health, self-reported diabetes or self-reported hypertension regarding pulmonary function parameters (P for interaction ranging from 0.067 to 0.752), except for age and sex (P for interaction <0.05) in the GBCS. Stronger associations of the CDAI score with higher FEV1 (0.005 L, 95% CI 0.002 to 0.009), FEV1pp (0.246%, 95% CI 0.071% to 0.421%), FVC (0.006 L, 95% CI 0.002 to 0.010), FVCpp (0.193%, 95% CI 0.034% to 0.353%) were found in GBCS women. A similar positive association was observed in participants aged ≥65 years, with FEV1 (β=0.008 L, 95% CI 0.002 to 0.014) and FVC (β=0.009 L, 95% CI 0.002 to 0.016). Likewise, stronger association in women from the NHANES regarding the associations of CDAI with FEV1pp (0.687%, 95% CI 0.215% to 1.160%) and FVCpp (0.355%, 95% CI 0.031% to 0.678%) were identified (table 3). Furthermore, significant positive associations between CDAI and all five pulmonary function parameters in both GBCS and NHANES were observed (figure 1). Dose-response relationships between the CDAI and pulmonary function parameters. The associations were modelled using restricted cubic splines. (A–E) correspond to FEV1 (L), FEV1pp (%), FVC (L), FVCpp (%), and FEV1/FVC (%) in the GBCS, respectively. The solid lines represent the multivariable-adjusted beta coefficients (β), and the shaded areas indicate 95% CIs. (F–J) correspond to the same parameters in the NHANES. CDAI, Composite Dietary Antioxidant Index; FEV1, forced expiratory volume in 1 s; FVC, forced vital capacity; GBCS, Guangzhou Biobank Cohort Study; NHANES, National Health and Nutrition Examination Survey; pp, percentage predicted. Table 2Association of Composite Dietary Antioxidant Index and pulmonary function parameters in the Guangzhou Biobank Cohort Study (N=8640) and NHANES (N=4029)GBCSCrude modelModel 1*Model 2†β (95% CI)P valueβ (95% CI)P valueβ (95% CI)P valueFEV1 (L)0.011 (0.009 to 0.014)***<0.0010.001 (−0.0004 to 0.004)0.1090.004 (0.001 to 0.008)**0.009FEV1pp, %0.090 (−0.005 to 0.184)0.0620.070 (−0.024 to 0.163)0.1460.193 (0.045 to 0.342)*0.011FVC (L)0.015 (0.012 to 0.018)***<0.0010.001 (−0.001 to 0.004)0.3530.004 (−0.0003 to 0.008)0.052FVCpp, %0.068 (−0.017 to 0.154)0.1200.014 (−0.071 to 0.099)0.7470.122 (−0.012 to 0.256)0.075FEV1/FVC, %−0.001 (−0.041 to 0.040)0.9880.030 (−0.010 to 0.070)0.1360.003 (−0.036 to 0.043)0.882NHANESCrude modelModel 1‡Model 2§β (95% CI)P valueβ (95% CI)P valueβ (95% CI)P valueFEV1 (L)0.058 (0.047 to 0.069)***<0.0010.018 (0.010 to 0.026)***<0.0010.008 (−0.002 to 0.018)0.113FEV1pp, %0.625 (0.402 to 0.847)***<0.0010.384 (0.153 to 0.614)**0.0010.353 (0.059 to 0.646)*0.018FVC (L)0.081 (0.068 to 0.095)***<0.0010.026 (0.017 to 0.035)***<0.0010.009 (−0.003 to 0.019)0.135FVCpp, %0.394 (0.223 to 0.566)***<0.0010.318 (0.140 to 0.497)***<0.0010.209 (−0.013 to 0.431)0.065FEV1/FVC, %−0.063 (−0.173 to 0.047)0.262−0.026 (−0.145 to 0.092)0.6610.047 (−0.085 to 0.179)0.485*p<0.05, **p<0.01, ***p<0.001.*Model 1: adjusted for age and sex.†Model 2: additionally adjusted for education, occupation, energy, personal annual income, BMI, smoking status, physical activity, self-rated health and self-reported hypertension, self-reported diabetes.‡Model 1: adjusted for age and sex.§Model 2: additionally adjusted for education, race, energy, PIR, BMI, cotinine, physical activity, self-rated health and self-reported hypertension, self-reported diabetes.BMI, body mass index; FEV1, forced expiratory volume in 1 s; FEV1pp, percentage-predicted FEV1; FVC, forced vital capacity; NHANES, National Health and Nutrition Examination Survey; PIR, poverty income ratio. Association of Composite Dietary Antioxidant Index and pulmonary function parameters in the Guangzhou Biobank Cohort Study (N=8640) and NHANES (N=4029) p<0.05, **p<0.01, ***p<0.001. Model 1: adjusted for age and sex. Model 2: additionally adjusted for education, occupation, energy, personal annual income, BMI, smoking status, physical activity, self-rated health and self-reported hypertension, self-reported diabetes. Model 1: adjusted for age and sex. Model 2: additionally adjusted for education, race, energy, PIR, BMI, cotinine, physical activity, self-rated health and self-reported hypertension, self-reported diabetes. BMI, body mass index; FEV1, forced expiratory volume in 1 s; FEV1pp, percentage-predicted FEV1; FVC, forced vital capacity; NHANES, National Health and Nutrition Examination Survey; PIR, poverty income ratio. Table 3Association of Composite Dietary Antioxidant Index and pulmonary function parameters by sex and age in the GBCS (N=8640) and NHANES (N=4029)Crude modelModel 1*Model 2†β (95% CI)P valueβ (95% CI)P valueβ (95% CI)P valueGBCS¶ Men FEV1 (L)0.006 (0.001 to 0.011)*0.0330.004 (−0.001 to 0.009)0.0790.002 (−0.005 to 0.010)0.551 FEV1pp, %0.202 (0.022 to 0.382)*0.0280.164 (−0.013 to 0.342)0.0690.088 (−0.193 to 0.370)0.538 FVC (L)0.005 (−0.001 to 0.011)0.0970.003 (−0.002 to 0.009)0.236−0.001 (−0.009 to 0.009)0.963 FVCpp, %0.116 (−0.043 to 0.276)0.1520.084 (−0.072 to 0.240)0.293−0.030 (−0.279 to 0.219)0.812 FEV1/FVC, %0.077 (−0.008 to 0.161)0.0750.071 (−0.012 to 0.155)0.0950.124 (−0.008 to 0.256)0.065Women FEV1 (L)0.004 (0.002 to 0.007)***<0.0010.001 (−0.001 to 0.003)0.5050.005 (0.002 to 0.009)**0.002 FEV1pp, %0.099 (−0.013 to 0.212)0.0820.032 (−0.079 to 0.142)0.5760.246 (0.071 to 0.421)**0.006 FVC (L)0.004 (0.001 to 0.007)**0.0030.0003 (−0.002 to 0.003)0.7980.006 (0.002 to 0.010)**0.005 FVCpp, %0.072 (−0.032 to 0.175)0.175−0.016 (−0.117 to 0.085)0.7520.193 (0.034 to 0.353)*0.017 FEV1/FVC, %0.023 (−0.022 to 0.068)0.3110.015 (−0.030 to 0.059)0.5190.025 (−0.045 to 0.096)0.481Crude modelModel 1‡Model 2§β (95% CI)P valueβ (95% CI)P valueβ (95% CI)P valueNHANES** Men FEV1 (L)0.036 (0.024 to 0.048)***<0.0010.024 (0.012 to 0.035)***<0.0010.007 (−0.009 to 0.022)0.397 FEV1pp, %0.388 (0.130 to 0.645)**0.0030.319 (0.058 to 0.579)*0.0170.173 (−0.183 to 0.529)0.341 FVC (L)0.045 (0.030 to 0.059)***<0.0010.031 (0.017 to 0.045)***<0.0010.005 (−0.013 to 0.024)0.576 FVCpp, %0.360 (0.135 to 0.584)**0.0020.306 (0.076 to 0.535)**0.0090.075 (−0.226 to 0.375)0.627 FEV1/FVC, %0.083 (−0.042 to 0.207)0.1940.037 (−0.086 to 0.161)0.5540.077 (−0.096 to 0.250)0.384Women FEV1 (L)0.019 (0.007 to 0.031)**0.0020.011 (0.001 to 0.022)*0.0490.009 (−0.002 to 0.020)0.118 FEV1pp, %0.584 (0.176 to 0.992)**0.0050.474 (0.066 to 0.882)*0.0230.687 (0.215 to 1.160)**0.004 FVC (L)0.029 (0.017 to 0.041)***<0.0010.019 (0.009 to 0.030)***<0.0010.012 (−0.0001 to 0.023)0.052 FVCpp, %0.379 (0.091 to 0.667)**0.0090.331 (0.047 to 0.614)*0.0220.355 (0.031 to 0.678)*0.032 FEV1/FVC, %−0.071 (−0.285 to 0.144)0.519−0.109 (−0.321 to 0.104)0.3160.018 (−0.183 to 0.218)0.863Crude modelModel 1††Model 2‡‡β (95% CI)P valueβ (95% CI)P valueβ (95% CI)P valueGBCS§§ <65, years FEV1 (L)0.007 (0.004 to 0.010)***<0.001−0.001 (−0.004 to 0.002)0.4230.002 (−0.002 to 0.007)0.229 FEV1pp, %−0.006 (−0.115 to 0.102)0.909−0.014 (−0.123 to 0.095)0.7980.132 (−0.039 to 0.304)0.130 FVC (L)0.008 (0.004 to 0.012)***<0.001−0.003 (−0.006 to 0.001)0.0930.001 (−0.004 to 0.006)0.627 FVCpp, %−0.068 (−0.168 to 0.033)0.189−0.102 (−0.203 to −0.002)*0.0460.046 (−0.112 to 0.205)0.567 FEV1/FVC, %0.027 (−0.017 to 0.071)0.2330.051 (0.007 to 0.095)*0.0240.067 (−0.002 to 0.136)0.058≥65, years FEV1 (L)0.019 (0.015 to 0.023)***<0.0010.006 (0.003 to 0.009)***<0.0010.008 (0.002 to 0.014)**0.006 FEV1pp, %0.213 (0.044 to 0.381)*0.0130.207 (0.035 to 0.378)*0.0180.293 (0.018 to 0.567)*0.037 FVC (L)0.025 (0.020 to 0.030)***<0.0010.008 (0.003 to 0.012)***<0.0010.009 (0.002 to 0.016)**0.009 FVCpp, %0.248 (0.099 to 0.396)**0.0010.184 (0.033 to 0.334)*0.0170.245 (0.003 to 0.487)*0.047 FEV1/FVC, %−0.056 (−0.132 to 0.019)0.143−0.003 (−0.079 to 0.074)0.9450.034 (−0.087 to 0.156)0.579Crude modelModel 1***Model 2†††β (95% CI)P valueβ (95% CI)P valueβ (95% CI)P valueNHANES¶¶ <65, years FEV1 (L)0.052 (0.039 to 0.065)***<0.0010.017 (0.007 to 0.026)***<0.0010.007 (−0.004 to 0.019)0.214 FEV1pp, %0.604 (0.352 to 0.855)***<0.0010.378 (0.122 to 0.635)**0.0040.385 (0.044 to 0.726)*0.027 FVC (L)0.074 (0.058 to 0.090)***<0.0010.023 (0.013 to 0.034)***<0.0010.007 (−0.006 to 0.019)0.318 FVCpp, %0.378 (0.179 to 0.576)***<0.0010.279 (0.072 to 0.488)**0.0080.180 (−0.078 to 0.438)0.171 FEV1/FVC, %−0.095 (−0.223 to 0.033)0.146−0.021 (−0.156 to 0.115)0.7640.065 (−0.092 to 0.223)0.415≥65, years FEV1 (L)0.056 (0.036 to 0.077)***<0.0010.022 (0.006 to 0.038)**0.0080.009 (−0.009 to 0.028)0.319 FEV1pp, %0.680 (0.196 to 1.165)**0.0060.401 (−0.094 to 0.894)0.1120.229 (−0.337 to 0.794)0.428 FVC (L)0.082 (0.057 to 0.107)***<0.0010.033 (0.014 to 0.051)***<0.0010.013 (−0.009 to 0.035)0.244 FVCpp, %0.452 (0.103 to 0.801)*0.0110.409 (0.057 to 0.762)*0.0230.265 (−0.173 to 0.702)0.236 FEV1/FVC, %−0.088 (−0.304 to 0.127)0.421−0.039 (−0.281 to 0.204)0.755−0.006 (−0.244 to 0.231)0.960*p<0.05; **p<0.01; ***p<0.001.*Model 1: adjusted for age†Model 2: additionally adjusted for occupation, education, energy, personal annual income, BMI, smoking status, physical activity, self-rated health, and self-reported hypertension, self-reported diabetes.‡Model 1: adjusted for age§Model 2: additionally adjusted for race, education, energy, PIR, BMI, cotinine, physical activity, self-rated health and self-reported hypertension, self-reported diabetes¶Stratified by sex, FEV1 (L): P for interaction = 0.021; FEV1pp, %: 0.042; FVC (L): 0.166; FVC pp, %: 0.386; FEV1/FVC, %: 0.359.**Stratified by sex, FEV1 (L): P for interaction = 0.524; FEV1pp, %: 0.331; FVC (L): 0.788; FVC pp, %: 0.644; FEV1/FVC, % :0.241.††Model 1: adjusted for sex‡‡Model 2: additionally adjusted for occupation, education, energy, personal annual income, BMI, smoking status, physical activity, self-rated health, and self-reported hypertension, self-reported diabetes.§§Stratified by age, FEV1 (L): P for interaction <0.001; FEV1pp, %: 0.025; FVC (L): <0.001; FVC pp, %: <0.001; FEV1/FVC, %: 0.045.¶¶Stratified by age, FEV1 (L): P for interaction=0.342; FEV1pp, %: 0.069; FVC (L): 0.249; FVC pp, %: 0.497; FEV1/FVC, %: 0.145.***Model 1: adjusted for sex.†††additionally adjusted for race, education, energy, PIR, BMI, cotinine, physical activity, self-rated health, and self-reported hypertension, self-reported diabetesBMI, body mass index; FEV1, forced expiratory volume in 1 s; FEV1pp, percent-predicted FEV1; FVC, forced vital capacity; FVCpp, percent-predicted FVC; GBCS, Guangzhou Biobank Cohort Study; NHANES, National Health and Nutrition Examination Survey; PIR, poverty income ratio. Association of Composite Dietary Antioxidant Index and pulmonary function parameters by sex and age in the GBCS (N=8640) and NHANES (N=4029) *p<0.05; **p<0.01; ***p<0.001. Model 1: adjusted for age Model 2: additionally adjusted for occupation, education, energy, personal annual income, BMI, smoking status, physical activity, self-rated health, and self-reported hypertension, self-reported diabetes. Model 1: adjusted for age Model 2: additionally adjusted for race, education, energy, PIR, BMI, cotinine, physical activity, self-rated health and self-reported hypertension, self-reported diabetes Stratified by sex, FEV1 (L): P for interaction = 0.021; FEV1pp, %: 0.042; FVC (L): 0.166; FVC pp, %: 0.386; FEV1/FVC, %: 0.359. Stratified by sex, FEV1 (L): P for interaction = 0.524; FEV1pp, %: 0.331; FVC (L): 0.788; FVC pp, %: 0.644; FEV1/FVC, % :0.241. Model 1: adjusted for sex Model 2: additionally adjusted for occupation, education, energy, personal annual income, BMI, smoking status, physical activity, self-rated health, and self-reported hypertension, self-reported diabetes. Stratified by age, FEV1 (L): P for interaction <0.001; FEV1pp, %: 0.025; FVC (L): <0.001; FVC pp, %: <0.001; FEV1/FVC, %: 0.045. Stratified by age, FEV1 (L): P for interaction=0.342; FEV1pp, %: 0.069; FVC (L): 0.249; FVC pp, %: 0.497; FEV1/FVC, %: 0.145. Model 1: adjusted for sex. additionally adjusted for race, education, energy, PIR, BMI, cotinine, physical activity, self-rated health, and self-reported hypertension, self-reported diabetes BMI, body mass index; FEV1, forced expiratory volume in 1 s; FEV1pp, percent-predicted FEV1; FVC, forced vital capacity; FVCpp, percent-predicted FVC; GBCS, Guangzhou Biobank Cohort Study; NHANES, National Health and Nutrition Examination Survey; PIR, poverty income ratio. Based on the XGBoost algorithm model, higher CDAI was associated with a gradual rise in the Shapley value (scatter points), indicating an increase in pulmonary function indicators (solid line) in both GBCS and NHANES (figure 2). Additionally, CDAI appeared to be an important factor associated with FEV1 (ranked third), FEV1pp (ranked first), FVC (ranked fourth), FVCpp (ranked second), and FEV1/FVC (ranked fourth), respectively, in the GBCS (figure 3). Similarly, within the NHANES dataset, the CDAI ranked among the top five factors associated with pulmonary function indicators (figure 4). SHAP dependence plots illustrating the relationships between CDAI and pulmonary function parameters in the XGBoost model. (A–E) correspond to FEV1 (L), FEV1pp (%), FVC (L), FVCpp (%) and FEV1/FVC (%) in the GBCS, respectively. (F–J) correspond to the same parameters in the NHANES. The red lines represent the trend of SHAP values after multivariable adjustment. The colour of the scatter points corresponds to the magnitude of the pulmonary function parameter values, ranging from blue (lower values) to green (higher values). CDAI, Composite Dietary Antioxidant Index; FEV1, forced expiratory volume in 1 s; FVC, forced vital capacity; GBCS, Guangzhou Biobank Cohort Study; NHANES, National Health and Nutrition Examination Survey; pp, percentage predicted; SHAP, SHapley Additive exPlanations, XGBoost, extreme gradient boosting. Variable importance ranked by mean absolute SHAP values for pulmonary function parameters in the GBCS. (A–E) correspond to the XGBoost models for FEV1 (L), FEV1pp (%), FVC (L), FVCpp (%) and FEV1/FVC (%), respectively. The horizontal bar charts represent the mean absolute SHAP values, demonstrating the contribution of each fully adjusted variable to the prediction. BMI, body mass index; CDAI, Composite Dietary Antioxidant Index; FEV1, forced expiratory volume in 1 s; FVC, forced vital capacity; GBCS, Guangzhou Biobank Cohort Study; pp, percentage predicted; SHAP, SHapley Additive exPlanations, XGBoost, extreme gradient boosting. Variable importance ranked by mean absolute SHAP values for pulmonary function parameters in the NHANES. (A–E) correspond to the XGBoost models for FEV1 (L), FEV1pp (%), FVC (L), FVCpp (%) and FEV1/FVC (%), respectively. The horizontal bar charts represent the mean absolute SHAP values, demonstrating the contribution of each fully adjusted variable to the prediction. BMI, body mass index; CDAI, Composite Dietary Antioxidant Index; FEV1, forced expiratory volume in 1 s; FVC, forced vital capacity; GBCS, Guangzhou Biobank Cohort Study; NHANES, National Health and Nutrition Examination Survey; PIR, poverty income ratio; pp, percentage predicted; SHAP, SHapley Additive exPlanations, XGBoost, extreme gradient boosting. Association of antioxidants with pulmonary function parameters In the GBCS, selenium levels were significantly associated with increased FEV1 (0.054 L, 95% CI 0.023 to 0.085), FEV1pp (1.754%, 0.358% to 3.151%) and FVCpp (1.462%, 0.199% to 2.726%). Zinc levels were negatively associated with FEV1 (−0.099 L, 95% CI −0.175 to −0.022) and FEV1pp (−4.895%, −8.346% to −1.444%) (Supplementary Table 1). In the NHANES, vitamin E was positively associated with pulmonary function indicators, including the FEV1 (0.052 L, 95% CI 0.002 to 0.101), FEV1pp (1.784%, 0.203% to 3.366%), FVC (0.114 L, 95% CI 0.059 to 0.170) and FVCpp (1.706%, 0.490% to 2.922%) (online supplemental table 2). Sensitivity analyses Multiple sensitivity analyses were performed to assess the robustness of the main results. First, zinc showed a U-shaped association with pulmonary function indicators (online supplemental figure 2) in the GBCS, with a nadir at the level of 12.81 mg/day. Second, the partial dependence plot of the XGBoost model, assuming other features are fixed at their average levels, indicated positive associations between CDAI and pulmonary function indicators (online supplemental figure 3). Third, the CDAI scores, which were calculated using two 24-hour dietary recalls and included dietary supplements, showed consistent results: higher CDAI was significantly associated with increased FEV1pp (β=0.342%, 95% CI 0.078% to 0.605%) in the NHANES (online supplemental table 3). Fourth, we found consistent results with the primary analysis showing significant positive associations of harmonised CDAI with FEV1 and FEV1pp in GBCS, and with FEV1pp in NHANES, further supporting the robustness of our findings (online supplemental table 4). Fifth, we recalculated the CDAI excluding zinc and observed highly consistent results. The revised CDAI maintained significant positive associations with FEV1, FEV1pp, FVC and FVCpp in the GBCS, and with FEV1pp and FVCpp in the NHANES (online supplemental table 5). Discussion This study provides compelling evidence that a higher CDAI was positively associated with improved pulmonary function parameters in both the GBCS and the NHANES. The observed associations persisted after adjusting for a comprehensive set of potential confounders, including demographic, socioeconomic and lifestyle factors. Notably, the relationship between CDAI and pulmonary function was more pronounced among women and individuals over 65 years of age. These findings suggest that dietary antioxidants play a significant role in maintaining pulmonary health, potentially offering a protective effect against age-related decline in pulmonary function. As a comprehensive indicator of antioxidants intake through diet,34 higher CDAI was associated with delayed ageing,23 reduced risk of hyperlipidaemia,35 decreased incidence of chronic kidney disease,36 improved cardiovascular health and lowered risk of stroke.36 In this study, we observed a dose-response association between CDAI and better pulmonary function indicators. The exact aetiology and pathogenesis of declining pulmonary function remain unclear, but oxidative stress might play a crucial role.37 Oxidative stress disrupts the balance between antioxidants and prooxidants, leading to tissue and organ damage.38 Dietary intake of antioxidants helps regulate plasma redox status, protecting the body from reactive oxygen and nitrogen species.39 Exogenous antioxidants can prevent inflammation and oxidative stress in patients with lung disease by regulating inflammatory factors such as alkaline phosphatase and C reactive protein,40 thereby slowing the decline in pulmonary function and reducing the risk of lung disease.41 Systematic reviews and meta-analyses have demonstrated significant protective effects of vitamin C supplementation against asthma and exercise-induced bronchoconstriction.42–44 However, recent studies indicated that high-dose vitamin B6 and B12 supplementation was associated with an increased risk of lung cancer in men, particularly among smokers.45 46 Furthermore, for well-nourished elderly individuals, daily intake of physiological doses of multivitamin and mineral supplements or 200 mg of vitamin E did not reduce the incidence or severity of acute respiratory infections; in fact, vitamin E was shown to exacerbate disease severity.47 These findings underscore the complex impact of various vitamins and micronutrients on health, highlighting the necessity for cautious use of high-dose supplements.48 Our study results indicated that the potential protective effect of CDAI on pulmonary function was more pronounced in women and those aged 65 years and older. Considering the cyclic changes in female hormone levels and the hormonal regulation of the inflammatory process, this may be due to hormonal changes across sexes and ages. Oestrogen is known for its anti-inflammatory effects, which may enhance the anti-inflammatory effects of dietary antioxidants.49 In older adults, as age increases, physiological functions including pulmonary function naturally decline. A higher CDAI score indicates lower levels of inflammation and better antioxidant capacity, which may help slow the ageing process and reduce the risks of lung diseases.50 Our results indicate that, in older age groups, where inflammation and oxidative stress due to physiological ageing are more pronounced, maintaining a higher CDAI score may offer greater benefits. A study showed that higher serum levels of antioxidant vitamins (vitamin A, β-cryptoxanthin, vitamin C and vitamin E), selenium, standardised calcium, chloride and iron were independently associated with higher FEV1 levels.51 However, evidence on the effects of zinc on pulmonary function remains limited and controversial. For example, some studies showed a significant association between higher dietary zinc intake and a lower prevalence of pulmonary function disorders,52 while other studies indicated that increased dietary zinc intake reduced the risks of pulmonary function disorders.15 53 54 Interestingly, our study found a U-shape relationship between dietary zinc and FEV1 levels, indicating a potential threshold for dietary zinc intakes, after which the inverse associations with pulmonary function may be observed. One possible explanation is that dietary zinc has both antioxidant and pro-oxidative properties, which can elicit oxidative stress when levels exceed the physiological range.55 Our study had several strengths. To the best of our knowledge, this study for the first time reported the association between CDAI and pulmonary function parameters. Moreover, we rigorously examined two extensive datasets, GBCS and NHANES and showed consistent results. The association between CDAI and pulmonary function parameters was further explored through the application of different methods (ie, generalised linear regression, RCS analysis and an XGBoost model). Additionally, given the cohort nature of the GBCS, future studies could leverage its longitudinal data to evaluate the long-term effects of CDAI on pulmonary function. Such analyses would help elucidate whether sustained high dietary antioxidant intake confers long-term protective effects on pulmonary function trajectories. However, some limitations warrant consideration. First, due to the cross-sectional nature of this study, causal associations between CDAI and pulmonary function parameters cannot be established. Second, the CDAI was determined using varying combinations of nutrients in the NHANES and GBCS datasets, potentially introducing bias when evaluating the dietary antioxidant capacity in diverse study populations.11 26 However, our sensitivity analysis using a harmonised CDAI showed consistent results with our primary analysis. This consistency suggests that the dataset-specific discrepancy in CDAI components did not affect the main conclusions. Additionally, variations in the calculation formulas for FEV1pp and FVCpp between Chinese and American populations may limit the comparability of the results. In conclusion, our analysis of the GBCS and NHANES databases found significant positive associations of CDAI with FEV1 and FEV1pp, with a stronger association in women and those aged 65 and older. The findings underscore the potential of dietary antioxidants to mitigate age-related declines in pulmonary function and reduce the risk of lung diseases. Further interventional studies are needed to establish causality and to explore the optimal combinations and dosages of dietary antioxidants for maintaining pulmonary function. Supplementary material