One of the most important programs of the National Center for Health Statistics (NCHS) is the NHANES. Data that is indicative of the nation on the general health of the US population has been collected (25). With the use of a sophisticated multistage probability methodology, this program produces a nationally representative sample of non-institutionalized Americans (26). Since 1999, it has used a sophisticated, stratified, multi-stage probability sampling method to gather data from about five thousand people each year (27). IN addition, its survey cycle lasts 2 years. All procedures were approved by the NCHS Research Ethics Committee, and each participant gave their informed permission (28). Data from questionnaires, laboratory test indicators, physical examination characteristics, and sociodemographic status were also collected. There are extensive details about the program on the website.¹

This study analysed data from four NHANES cycles, spanning from 2011 to 2018. The survey was completed by 39,156 people in total (Representing an actual population of 103,877,855). Firstly, we excluded females (n = 19,848) and males under 20 years of age (n = 8,361). The subsequent exclusion criteria were as follows: (1) information about dietary (n = 2,815) is unknown; (2) interviewees who had not finished the UI survey (n = 304); (3) data about education status (n = 7) is unknown; (4) body mass index (n = 72) is unknown; (5) data about smoking status (n = 3) is unknown; (6) information about vigorous activities (n = 3) is unknown; (7) information about moderate activities (n = 5) is unknown; and (8) Total cholesterol (n = 3) is unknown. Finally, the study comprised 7,735 individuals in total after screening procedures, including 1,247 men with UUI and the remaining without UUI. Figure 1 shows the screening procedures.

The 24 h dietary recall interview is the current section of the NHANES nutritional evaluation. Dietary interviewers with training who are fluent in both Spanish and English perform dietary recall interviews in person. Through discontinuous two-day, 24 h dietary recall interviews, the NHANES gathered data on participants’ food intake. A mobile examination center (MEC) served as the venue for the initial interview. Every mobile examination center had a nutritional interview room with a common set of measurement guidelines. The second dietary recall interview took place over the phone three to ten days later (29, 30).

We used the average of two measurements to minimize bias and increase the reliability of the results. A modified version developed by previous researchers was utilized to calculate each participant’s CDAI (31). Zinc, selenium, carotenoids, and vitamins A, C, and E are the six dietary antioxidants that make up CDAI. We standardized each micronutrient by subtracting the mean of the six dietary vitamins and minerals and dividing by the total standard deviation to calculate a Z-score. Their Z-scores were then summed to obtain the composite value of CDAI. The following is the calculating formula:

In MECs (Mobile Examination Centres), only those who were ≥ 20 years old answered questions on urine incontinence. “During the past 12 months, have you leaked or lost control of even a small amount of urine with activity like coughing, lifting, or exercise?” asked participants, if they selected “yes.” grouped under “stress UI.” A positive response to the question, “During the past 12 months, have you leaked or lost control of even a small amount of urine with an urge or pressure to urinate and you could not get to the toilet fast enough?” formed the basis for the definition of “urgency UI.” If a participant answered “yes” to both the stress and urgency UI questions, they were categorized as mixed UI participants.

The following potential confounding variables were selected for this investigation based on prior research findings that may affect UI results. These variables included demographics, medical history, physical examination findings, and personal circumstances (6, 8, 32–35). Race, age, education, PIR, smoking status, alcohol use, diabetes, hypertension, total cholesterol levels, and exercise status were all considered categorical factors.

We then grouped these covariates. Two groups were formed from the participants according to their age, which was fifty. The age range of the first group was under 50, whereas the age range of the second group was over or equal to 50. Three categories—below high school diploma, high school diploma, and above high school diploma—were used to classify people’s educational positions. PIR < 2 was classified as low PIR, and PIR ≥ 2 as high PIR. We divided BMI into three groups: the first group was less than 25, the second group was greater than or equal to 25 and less than 30, and the third group was less than 30. When asked if they smoked, participants were divided as smokers if they replied “yes” (SMQ020). Answer the query “In the past 12 months, on those days that you drank alcoholic beverages, on average, how many drinks did you have?” to distinguish between drinkers and non-drinkers. Those who had less than 12 drinks were categorized as non-drinkers, while those who had more than or equal to 12 drinks were considered drinkers.

Those who responded “yes” to the inquiry “Have you been told you have hypertension?,” those who were taking antihypertensive medication, and those whose average of three measurements of systolic blood pressure (BP) was ≥140 mmHg and whose average of three measurements of diastolic blood pressure (BP) was ≥90 mmHg were all considered patients with hypertension. Participants were considered to have diabetes if they answered “yes” when asked if they had diabetes or if they used glucose-lowering drugs or insulin. Meanwhile, their fasting blood glucose (≥126 mg/dL) and glycosylated haemoglobin (≥6.5%) were used to diagnose diabetes. Total cholesterol <240 mg/dL was defined as a low cholesterol level and ≥ 240 mg/dL was defined as a high cholesterol level.

During the processing phase, NHANES sample weights were applied to guarantee the study population’s national representation. The baseline qualities were explained after the participants were classified according to the four CDAI categories. Categorical variables, expressed as weighted percentages (%), were compared using a chi-square test. Continuous variables were compared using weighted linear regression, and the results were shown as mean (± standard deviation).

To reduce confounding bias, we performed PSM processing based on UUI results. To ensure that the distribution of covariates was comparable between the UUI and non-UUI groups, we matched for the following factors: age, smoking, alcohol consumption, diabetes, hypertension, and total cholesterol (36–40). Propensity score matching (PSM) was carried out 1:1 using the R Software “MatchIt” program. The sample population was re-examined after PSM in order to validate the findings.

To determine whether there were any noteworthy trends or differences, a univariate analysis was performed. The study then developed three adjustment models in order to use multivariate regression analysis to examine the relationships between CDAI and UUI. Model 1 was the baseline model that did not alter any covariate variables. Race, age, education, and PIR were added to model 2. Model 3 included BMI, diabetes, hypertension, smoking, alcohol consumption, total cholesterol level, moderate activity, and vigorous activity on the basis of Model 2. A generalized additive model (GAM) was then developed to validate the dose–response relationship. A threshold effect analysis was first performed. Then, restricted cubic spline (RCS) and smooth curve fitting were used to describe the dose–response relationship between CDAI and UUI. The study used RCS to investigate whether there was a nonlinear association between CDAI and UUI, with the node set to 3. Smooth curve fitting was performed under the fully modified model. After obtaining the threshold value, the distribution of CDAI in the population was made into a violin plot for the next analysis.

We then categorized the population according to the thresholds and then regressed the CDAI and its components on the UUI separately to explore whether certain elements had a greater effect. Then, using subgroup analysis, the stratified association between CDAI and UUI was examined. Additionally, interaction tests were performed to assess the way in which relationships between subgroups interacted.

For all statistical studies, R (version 4.4.0) was utilized. When the two-sided p-value was less than 0.05, it was deemed statistically significant.

Based on the screening criteria, a total of 7,735 eligible participants from NHANES 2011–2018 were selected (Figure 1). Among them, 1,247 had UUI and 6,488 did not. The weighted estimates for the baseline characteristics of the study population are shown in Table 1. Study participants with a higher CDAI were found to be more heavily represented under the age of 50. In addition, those with higher levels of CDAI were more likely to be non-Hispanic white, more educated, have a higher PIR, lower smoking rates, lower alcohol consumption rates, lower body mass index, moderate exercise intensity, lower probability of having a history of diabetes and hypertension, and lower total cholesterol levels than those with lower levels of CDAI.

We found that differences in CDAI levels differed significantly in the presence or absence of UUI and that participants with UUI had lower CDAI levels. However, no significant CDAI level differences were observed in SUI and MUI.

Afterwards, the CDAI distribution was analysed and visualized as a violin plot (Figure 2). Many people have CDAI values clustered in the lower range, even less than 0 in a large proportion, while the number of people with higher antioxidant indices is smaller (median CDAI: 0.776).

We then balanced the effect of potential confounders associated with UUI through propensity score matching (PSM) analysis. In this investigation, a 1:1 PSM analysis was conducted. The standardized mean difference is visualized in Supplementary Figure S1, and the distributional balance plot of PSM is shown in Supplementary Figure S2. Following PSM, 1702 participants were enrolled in the study; 851 of them had UUI, while the remaining individuals did not (Figure 1). The weighted essential attributes of the research subjects were displayed in Supplementary Table S1 following PSM. Differences in variables between the two groups were managed to some degree. After PSM, age, race, education, PIR, hypertension, total cholesterol, vigorous activity, and UUI were found to be significantly associated with CDAI.

Preliminary exploration of pre- and post-PSM data with a univariate analysis of UUI. We found the following: prior to PSM, UUI was positively associated with age ≥ 50 years, body mass index ≥25, non-Hispanic black, hypertension, diabetes mellitus, and high cholesterol. In addition, UUI was negatively associated with other races, high school and higher education, high PIR, nonsmokers, vigorous activity, and Q2 and Q4 in the CDAI quartiles. After PSM, UUI was not associated with covariates such as age, high school education, 25 ≤ BMI < 30, smoking, hypertension, diabetes mellitus, level of exercise, and cholesterol level.

After PSM, only BMI ≥ 30 was positively associated with UUI. In contrast, other races, higher than high school education level, high PIR, and Q2 in CDAI were negatively associated with UUI (Table 2). We found that Q2 in CDAI was negatively correlated with UUI both before and after PSM (p = 0.007 and p = 0.004).

To further explore the relationship between CDAI and UUI, a weighted logistic regression analysis of the three models was conducted (Figure 3). The CDAI was converted (Q1-4; Q1 was used as a reference point) and examined by quaternity. In Model 1, no variables have been added. In Model 2, race, Age, education, and PIR adjustments were made. Model 3 was built using Model 2, modified to account for BMI, smoking, alcohol consumption, hypertensive disease, diabetes, total cholesterol, moderate activity, and vigorous activity.

In the pre-PSM analysis, only Q2 (p = 0.007) and Q4 (p < 0.001) were negatively correlated with UUI in the crude model, and this relationship was lost in the other models. In the analysis following the PSM, a negative correlation between Q2 and UUI was found in all three models (p = 0.004 in the crude model, p = 0.034 in model 2, and p = 0.028 in model 3).

In the threshold effect analysis, we found no linear relationship between CDAI and UUI either before or after PSM (p = 0.383 and p = 0.483). Moreover, the negative correlation between CDAI greater than or less than the K value and UUI was not significant before PSM (p = 0.086 and p = 0.341). After PSM, CDAI was negatively correlated with UUI when CDAI was <K value (p = 0.027), and the relationship was not significant when it was > K value (Table 3).

Figure 4 displays the results of the dose–response relationship for the Restricted Cubic Spline (RCS) (node setting of 4 before and after PSM). Under the fully adjusted model, CDAI and UUI prevalence showed a nonlinear relationship before and after PSM (nonlinear p of 0.010 before PSM, nonlinear p of 0.007 after PSM). At OR = 1, the reference values for CDAI before and after PSM were 0.776 and 0.523, respectively. The overall relationship between CDAI and UUI was significant (p for overall = 0.024 before PSM and p for overall = 0.018 after PSM).

The K value after PSM (K = 0.523) was used to distinguish the two groups of people and the multiple regression analysis of CDAI and UUI was performed before and after PSM, respectively. The nutrient elements of each component in the CDAI<0.523 group were subjected to sensitivity analysis (Supplementary Table S2) and visualized (Figure 5).

The negative correlation with UUI was not significant for CDAI ≥0.523 either before or after PSM (p = 0.323 before PSM, p = 0.278 after PSM). However, at a CDAI less than 0.523, the CDAI was negatively correlated with UUI incidence both before and after PSM (OR = 0.92, p = 0.03 before PSM and OR = 0.91, p = 0.033 after PSM). Moreover, among the components, only the z-score of zinc showed a relatively significant negative correlation with UUI (OR = 0.72, p = 0.002 before PSM and OR = 0.74, p = 0.02 after PSM), and no significant correlation was found between the rest of the components and the incidence of UUI.

To further explore the relationship between CDAI and zinc with UUI, we performed RCS plots for CDAI and zinc, respectively, under the fully adjusted model with the node setting of 3 before PSM (Figure 6). Neither the CDAI nor the z score of Zinc showed a nonlinear relationship with UUI (CDAI, nonlinear p of 0.754, Zinc, nonlinear p of 0.262). At OR = 1, the reference values were − 1.672 and − 0.287, respectively. The overall p-value for the RCS plotted for the relationship between CDAI and UUI was 0.042.

The study stratified the first subsection by race, age, education, PIR, body mass index, diabetes mellitus, hypertension, whether or not they smoked, whether or not they drank alcohol, total cholesterol, and moderate and vigorous activity (Table 4). Analyses were statistically significant regardless of pre- and post-PSM (p = 0.03 before PSM and p = 0.033 after PSM). Before and after PSM, CDAI was significantly negatively associated with UUI among those whose race was other Hispanic, higher than high school education, high PIR, non-drinkers, diabetics, and high cholesterol levels. In addition, after PSM, CDAI was negatively associated with UUI among those aged ≥50 years. Moreover, race played a significant moderating role in the effect after PSM, p for interaction = 0.043. It suggests a significant difference between racial subgroups. Apart from this, no significant interaction effects were found in other subgroups regardless of before or after PSM, suggesting that our results apply to almost all types of male populations.

Based on Figdraw, an intuitive diagram is presented in Figure 7. We discuss the mechanisms by which CDAI affects the prevalence of UUI in Part 4. In the figure, the green arrows represent typical foods rich in the Composite Dietary Antioxidant Index (CDAI), which are beneficial for reducing the risk of UUI. The red arrows indicate biological mechanisms discussed in the article that contribute to the occurrence of UUI. These include markers of inflammation and oxidative stress: CRP (C-reactive protein), IL-1β, IL-6, IL-8 (cytokines), ROS/RNS (reactive oxygen and nitrogen species), and MCP-1 (monocyte chemoattractant protein-1). Other involved factors include Rho (Ras homologous), parasympathetic nerves, detrusor overactivity (DO), and peptidergic/sensory nerves (For more details, please refer to the discussion section).