Machine Learning-Driven Investigation of Associations Between Phthalate Biomarkers and Glaucoma Using US NHANES Data (2011–2016)

Machine Learning-Driven Investigation of Associations Between Phthalate Biomarkers and Glaucoma Using US NHANES Data (2011–2016) | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Machine Learning-Driven Investigation of Associations Between Phthalate Biomarkers and Glaucoma Using US NHANES Data (2011–2016) Duncheng Xiao, Xiaoyan Liu, Yu Li, Mohan Li, Song Wang, Fan Yang, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7009815/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background As the leading cause of irreversible blindness globally, glaucoma involves a multifactorial etiology encompassing dysregulated intraocular pressure, optic neuropathy, and interactions between genetic predisposition and environmental determinants. Phthalates, ubiquitous endocrine-disrupting chemicals in plastics and personal care formulations, may adversely impact the central nervous and cardiovascular systems through hormonal interference, oxidative stress induction, and inflammatory pathway activation. Given their potential to target ocular structures including retinal neurofibers, microvasculature, and aqueous humor outflow pathways, research exploring the phthalate-glaucoma relationship remains nascent. Comprehensive analytical approaches are imperative for elucidating pathogenic mechanisms and characterizing associated risk profiles. Methods To evaluate associations between urinary phthalate metabolite concentrations and glaucoma susceptibility, this study integrated data from large prospective cohort studies or national health databases (2011–2016), incorporating phthalate biomarker measurements, glaucoma diagnostic status, and covariates (e.g., age, medical history). Eleven distinct machine learning algorithms were implemented for model development, with optimization conducted via stratified cross-validation. The optimal predictive model was selected guided by performance criteria such as the receiver operating characteristic curve's area under the curve (AUC). We implemented permutation feature importance analysis, evaluated accumulated local effects (ALE), and interpreted SHAP (SHapley Additive exPlanations) values to elucidate influential variables and their interaction patterns. Sensitivity analyses established the robustness of outcomes across subgroups. Conclusion This research applied explainable AI (XAI) frameworks for examining the relationship linking phthalate biomarkers to glaucoma risk. Among 11 evaluated models, the Gradient Boosting Machine (GBM) algorithm demonstrated superior predictive capability. Age constituted the most influential risk determinant. Several phthalate metabolites—specifically MEOP, MCNP, MEHP, and MECPP—were identified as significant contributors to glaucoma risk stratification. The results emphasize the vital role of integrating environmental exposure biomarkers in glaucoma prognostic models and highlight the necessity for mechanistic investigations into underlying biological pathways. Glaucoma Phthalates Explainable artificial intelligence Risk stratification NHANES Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction Glaucoma, representing the primary cause of irreversible vision loss worldwide, severely compromises global visual health. According to World Health Organization estimates, approximately 80 million individuals were affected globally in 2020, with projections indicating a rise to 112 million cases by 2040 1 . Although therapeutic interventions such as trabeculectomy, aqueous drainage device implantation, and pharmacologic intraocular pressure reduction can effectively modulate disease progression 2 , the insidious onset of early symptoms combined with healthcare access disparities frequently results in delayed diagnosis. Consequently, many patients present with irreversible optic nerve damage and visual field deficits at initial clinical assessment 3 , substantially elevating blindness risk 4 . The pathogenesis of glaucoma involves complex interplay among multiple factors, including elevated intraocular pressure, impaired optic nerve perfusion, genetic susceptibility, and oxidative stress mechanisms 4 , 5 . Therefore, identifying modifiable environmental risk factors holds substantial public health relevance for primary prevention and progression delay strategies 6 , 7 Phthalates constitute a class of synthetic compounds extensively utilized as plasticizers in consumer products, building materials, cosmetics, and food packaging systems 8 . Due to non-covalent binding within polymer matrices, these compounds readily migrate into environmental media through volatilization and leaching processes, subsequently enabling human exposure via inhalation, dietary intake, and dermal absorption routes 9 . Chronic systemic absorption of phthalates demonstrates established associations with multiple disease states, including diabetes mellitus, cardiovascular pathologies, and neurodevelopmental disorders 10 . Functioning as prototypical endocrine-disrupting chemicals, phthalates perturb endogenous hormonal signaling through receptor agonism/antagonism, disrupting cellular metabolic processes and physiologic functions 11 . Emerging toxicological evidence implicates phthalates in ocular pathophysiology. Preclinical studies indicate that phthalate exposure may induce retinal ganglion cell apoptosis and trabecular meshwork dysfunction via oxidative stress and pro-inflammatory cascades 8 , 12 . Nevertheless, epidemiological understanding regarding associations between specific phthalate biomarkers and glaucoma susceptibility remains notably underdeveloped, with mechanistic insights conspicuously limited. Machine learning represents a data-adaptive analytical paradigm capable of extracting latent association patterns from complex, high-dimensional datasets. This approach facilitates identification of risk determinants and interaction effects that often elude conventional statistical methodologies 13 . Within medical research domains, machine learning has demonstrated substantial utility in disease risk prognostication and diagnostic classification frameworks 14 . Although prior investigations have applied machine learning to elucidate ophthalmic disease mechanisms 15 , predictive modeling of glaucoma risk incorporating environmental determinants such as phthalate exposure remains scarce. Furthermore, comprehensive interpretability analyses of predictive outputs are frequently underutilized in existing literature 16 . This study therefore aims to leverage large-scale prospective cohort data and advanced machine learning techniques to interrogate associations between phthalate biomarker exposure profiles and glaucoma susceptibility. We implement explainable artificial intelligence methodologies—including permutation feature importance quantification, accumulated local effects analysis, and SHapley Additive exPlanations (SHAP)—to delineate predictor contributions and interaction networks. This integrative analytical strategy seeks to establish a robust scientific foundation for early glaucoma prevention initiatives and environmental risk factor mitigation protocols 17 . Methods Participants This study leveraged data from the National Health and Nutrition Examination Survey (NHANES)—a continuous cross-sectional surveillance system conducted by the National Center for Health Statistics (NCHS), Centers for Disease Control and Prevention (CDC), capturing nationally representative health metrics of non-institutionalized U.S. residents.This initiative systematically gathers health information from non-hospitalized U.S. residents. Study data were publicly accessed via the official CDC website ( https://www.cdc.gov/nchs/nhanes/ ). Ethical clearance was granted by the NCHS Ethics Review Board, and all participants provided written informed consent.This analysis incorporated data from three NHANES cycles (2011–2012, 2013–2014, and 2015–2016), comprising laboratory-measured urinary phthalate metabolites and paired sociodemographic records.Conducted by the National Center for Health Statistics (NCHS) at the Centers for Disease Control and Prevention (CDC), the National Health and Nutrition Examination Survey (NHANES) acquires nationally representative health metrics from non-institutionalized U.S. residents.Publicly accessible datasets were retrieved from the official CDC portal ( https://www.cdc.gov/nchs/nhanes/ ). The NCHS Ethics Review Board approved the study protocol, with written informed consent secured from all participants. Individuals lacking complete urinary phthalate or glaucoma status data were excluded, yielding a final analytical cohort of 2,113 subjects. Glaucoma Diagnosis and Phthalate Exposure Assessment For the diagnosis of glaucoma, participants were queried regarding prior glaucoma surgery using the survey item VIQ071: "Have you ever had glaucoma surgery?" Responses were dichotomized into "yes" or "no." Individuals affirming this query were categorized as having glaucoma. Urinary concentrations of eleven phthalate metabolites were quantified to assess exposure levels. The measured compounds comprised monobutyl phthalate (MnBP), mono(2-ethyl-5-oxohexyl) phthalate (MEOP), monoisobutyl phthalate (MiBP), monocarboxyloctyl phthalate (MCOP), monocarboxynonyl phthalate (MCNP), mono(3-carboxypropyl) phthalate (MCCP), monoethyl phthalate (MEP), monobenzyl phthalate (MBzP), mono(2-ethyl-5-carboxypentyl) phthalate (MECPP), mono(2-ethylhexyl) phthalate (MEHP), and mono(2-ethyl-5-hydroxyhexyl) phthalate (MEHHP). Quantification of urinary metabolites was performed using high-performance liquid chromatography (HPLC) with electrospray ionization tandem mass spectrometry (ESI-MS/MS) detection. Comprehensive methodological specifications and quality assurance procedures are documented in peer-reviewed publications and accessible through the CDC's online repository 18 . Sociodemographic and Clinical Covariates Sociodemographic information was obtained through self-administered questionnaires and included age, sex (male or female), race/ethnicity (non-Hispanic White, non-Hispanic Black, Mexican American, and other), marital status (married or living with a partner vs. unmarried or other), educational attainment (less than high school vs. high school or above), and poverty income ratio (PIR). A PIR < 1.00 indicated income below the federal poverty line, while a PIR ≥ 1.00 indicated income at or above this threshold. Body mass index (BMI) was computed as weight (kg) divided by height squared (m²). Participants were categorized into three groups: underweight (BMI < 18.5 kg/m²), normal-to-overweight (BMI 18.5–30.0 kg/m²), and obese (BMI ≥ 30.0 kg/m²). Lifestyle factors, also derived from questionnaire data, included alcohol consumption and smoking status. Alcohol consumption was classified into two groups: non-drinkers (defined as individuals who had never consumed alcohol or had consumed fewer than 12 drinks in their lifetime) and drinkers (defined as individuals who had consumed 12 or more drinks in their lifetime). Participants were stratified by smoking history into three groups: never smokers, former smokers, and current smokers. Clinical comorbidities considered in the analysis included hypertension, which was defined based on at least one of the following criteria: a physician-confirmed diagnosis via self-report, active use of antihypertensive medication, or elevated blood pressure measurements (systolic blood pressure ≥ 140 mmHg or diastolic blood pressure ≥ 90 mmHg). Additionally, urinary creatinine (Ucr), C-reactive protein (CRP), and hemoglobin A1c (HbA1c) were included as covariates, as these biomarkers have been identified in previous studies as potential confounders. Construction of the ML model The full dataset was randomly split into training (80%, N = 1692) and testing (20%, N = 421) subsets.In detail, we employed a five-fold cross-validation strategy, where each fold served as the test set once, with the other four folds constituting the training. The model underwent five rounds of training and validation for each iteration to obtain average performance metrics, facilitating a more accurate and thorough comparison of model performance. Subsequently, Eleven machine learning algorithms were applied to predict glaucoma using training data: Neural Network (NN), Support Vector Machine (SVM), Multi-Layer Perceptron (MLP), Gaussian Process (GP), Gradient Boosting Machine (GBM), Logistic Regression (LR), Naïve Bayes (NB), XGBoost (XGB), C5.0 Decision Trees (C5.0), k-Nearest Neighbors (KNN), and Random Forest (RF). Evaluation of ML model The predictive performance of all eleven machine learning models was assessed using diverse performance indicators,encompassing: the receiver operating characteristic (ROC) curve with its area under the curve (AUC), accuracy, sensitivity (recall), specificity, positive and negative predictive values (PPV). These metrics were calculated based on the testing dataset using R software on a Windows platform, employing packages such as 'caret', 'randomForest', 'pROC', 'stats', 'epiR', 'ggplot2', and 'dplyr'. Interpretable methods pipeline of prediction models Upon selection of the most optimal predictive model, we conducted permutation feature importance (PFI), accumulated local effect (ALE) analysis, and SHapley Additive exPlanations (SHAP) to explore key predictors and their interaction effects contributing to model predictions. Both PFI and SHAP were employed to quantify the relative importance of individual variables and to identify those with the greatest influence on model performance. These explainable machine learning approaches were implemented using R software on a Windows operating system, incorporating the following R packages: 'shapviz', 'xgboost', 'lime', 'caret', 'DMwR2', 'randomForest', and 'iml'. ALE analysis estimates the localized effect of each feature by partitioning the feature's range into intervals and evaluating the marginal change in model output across these intervals. This allows for assessment of how feature variation influences predictions within specific regions of the feature space. SHAP, grounded in cooperative game theory, quantifies each feature's marginal contribution to the model output by assigning SHAP values. To further interpret the results, SHAP dependence plots were generated to visualize the relationship between individual predictors—including urinary phthalate metabolites—and the predicted probability of glaucoma 19 . Statistical Analysis Continuous variables were expressed as mean ± standard deviation (SD) when normally distributed or median (interquartile range, IQR) otherwise. Categorical variables were presented as frequency counts with percentages. All analyses employed R statistical software (v4.1.2), with statistical significance indicated by two-tailed p-values < 0.05. Results Participant Characteristics The analytical cohort comprised 2,113 individuals: 1,974 unaffected controls and 139 glaucoma cases. Table 1 details their demographic, clinical, and biochemical profiles.The age distribution significantly differed between the two groups (P < 0.0001). In the affected group, 69.79% were aged ≥ 70 years, while younger age groups (40–59 years) were more prevalent in the unaffected group. Table 1 Baseline characteristics of participants. Characteristic Study Participants, N (%)/ Mean ± SD/ Median (Q1,Q3) Total (n = 2113 ) No (n = 1974) Yes (n = 139) P Age < 0.0001 40–49 541(32.82) 520(18.59) 2( 1.45) 50–59 502(30.16) 469(16.77) 10( 4.93) 60–69 507(19.02) 440( 9.56) 56(22.39) ≥ 70 563(18.01) 336( 6.22) 203(69.79) BMI 0.64 18.5–30.0 1276(61.38) 1192(61.44) 84(60.36) < 18.5 43( 2.08) 37(1.94) 6(4.27) ≥ 30.0 794(36.55) 745(36.62) 49(35.37) UCr, (mg/dL) 113.52(2.63) 113.38( 2.61) 115.73(10.46) < 0.0001 CRP, (mg/L) 0.43(0.03) 0.43(0.03) 0.39(0.06) 0.17 HbA1c, (%) 5.71(0.03) 5.70(0.03) 5.83(0.11) < 0.0001 MnBP 35.69(4.67) 36.38(5.00) 24.41(2.12) 0.43 MEOP 34.24(2.81) 34.11(3.02) 36.37(8.66) < 0.001 MiBP 9.11(0.60) 8.96(0.53) 11.67(3.58) 0.19 MCNP 6.38(0.98) 6.32(1.04) 7.48(3.37) 0.18 MCOP 14.09(1.47) 14.17(1.52) 12.81(3.04) < 0.001 MCCP 4.54(0.26) 4.61(0.28) 3.41(0.47) 0.49 MEP 375.62(30.88) 360.96( 26.89) 615.80(216.90) 0.15 MBzP 12.34(0.67) 12.20(0.68) 14.56(2.46) 0.1 MEHP 7.37(0.76) 7.34(0.83) 7.86(2.42) < 0.0001 MECPP 80.00(7.31) 80.09( 7.90) 78.44(19.93) < 0.0001 MEHHP 60.01(5.19) 59.91( 5.52) 61.76(15.52) < 0.0001 Gender < 0.001 Male 1055(47.91) 985(47.75) 70(50.54) Female 1058(52.09) 989(52.25) 69(49.46) Ethnic < 0.0001 Non-Hispanic White 1118(76.29) 1044(76.27) 74(76.53) Non-Hispanic Black 454(10.42) 419(10.32) 35(12.05) Mexican American 329( 5.62) 313(5.71) 16(4.11) Other 212( 7.67) 198(7.69) 14(7.31) Marital 0.03 Unmarried or other 782(31.09) 720(30.79) 62(36.03) Married or living with a partner 1331(68.91) 1254(69.21) 77(63.97) Education 0.99 Less than High school 635(18.92) 585(18.53) 50(25.33) High School or above 1478(81.08) 1389(81.47) 89(74.67) Poverty ratio 0.56 Below Poverty line (< 1.00) 359(10.89) 336(11.05) 23( 8.21) At or above Poverty line (≥ 1.00) 1754(89.11) 1638(88.95) 116(91.79) Smoking status < 0.0001 Never 1040(49.03) 969(49.38) 71(43.27) Former 664(30.93) 615(30.20) 49(43.00) Current 409(20.04) 390(20.42) 19(13.73) Alcohol drinking < 0.0001 No 301(11.50) 279(11.55) 22(10.68) Yes 1812(88.50) 1695(88.45) 117(89.32) Hypertension < 0.0001 No 920(48.48) 883(49.50) 37(31.83) Yes 1193(51.52) 1091(50.50) 102(68.17) UCr: urinary creatinine, BMI: body mass index (calculated as weight in kilograms divided by height in meters squared). Body mass index (BMI) distribution did not differ significantly between groups (P = 0.64), with the majority of participants having a BMI between 18.5–30.0 kg/m² in both groups.Biochemical markers showed notable differences. Affected individuals had higher serum creatinine (115.73 ± 10.46 vs. 113.38 ± 2.61 mg/dL; P < 0.0001) and HbA1c levels (5.83 ± 0.11% vs. 5.70 ± 0.03%; P < 0.0001), indicating renal and metabolic differences.Among phthalate metabolites, MEOP, MEHP, MEHHP, and MCOP showed statistically significant intergroup differences (all P values 0.05).Sex distribution was comparable (male: 50.54% vs. 47.75%; P < 0.001), while ethnicity distribution significantly varied (P < 0.0001), although Non-Hispanic Whites comprised the majority in both groups.Sociodemographic variables such as marital status showed a modest difference (P = 0.03), while education level and poverty status were comparable (P = 0.99 and P = 0.56, respectively). In lifestyle factors, former smoking was more common in the affected group (43.00% vs. 30.20%), whereas current smoking was less frequent (13.73% vs. 20.42%; P < 0.0001). Alcohol consumption did not significantly differ. Hypertension exhibited substantially higher prevalence among glaucoma cases (68.17% vs. 50.50%; P < 0.0001). Machine Learning Model Performance and Comparison Figure 1 and Supplemental Fig. 1 depict the receiver operating characteristic (ROC) curves, illustrating the predictive performance of all 11 machine learning models for glaucoma risk in both training and test sets. These curves assess the discriminative capacity of each algorithm in predicting glaucoma susceptibility. The AUC values on the test dataset indicated that the Gradient Boosting Machine (GBM) achieved the highest discriminative performance (AUC = 0.682), followed closely by XGBoost (AUC = 0.679) and Random Forest (RF, AUC = 0.678). Other models, including Gaussian Process (GP, AUC = 0.643), Naïve Bayes (NB, AUC = 0.637), Logistic Regression (LR, AUC = 0.615), Multilayer Perceptron (MLP, AUC = 0.611), Support Vector Machine (SVM, AUC = 0.601), and Neural Network (NN, AUC = 0.594) demonstrated moderate performance, while the K-Nearest Neighbor (KNN, AUC = 0.538) and C5.0 (AUC = 0.500) models showed limited predictive capability. These results suggest that ensemble-based algorithms such as GBM and RF outperform traditional classifiers in capturing the complex relationships between phthalate exposure and glaucoma risk, while simpler models like C5.0 and KNN may lack sufficient predictive power under current conditions 20 . A comprehensive summary of model performance metrics, including accuracy, precision, recall, and F1 score, is presented in Table 2 . Considering both predictive performance and interpretability, the GBM model was ultimately selected for downstream analysis. Table 2 Comparative Analysis of Discriminative Performance Across 11 Machine Learning Models Discriminative capabilities of the eleven machine learning algorithms were systematically compared within the testing cohort. All predictive frameworks were constructed without employing data augmentation methodologies. Standard abbreviations include: Area Under the Receiver Operator Curve (AUC), Machine Learning (ML), Supported Vector Machine (SVM), Neural Network (NN), Multi-Layer Perceptron (MLP), Gaussian Process (GP), Gradient Boosting Machine (GBM), Logistic Regression (LR), Naive Bayes (NB), C5.0 Decision Trees (C5.0), k-Nearest Neighbor (KNN), Random Forest (RF), Positive Predictive Value (PPV), Negative Predictive Value (NPV), Positive Likelihood Ratio (PLR), and Negative Likelihood Ratio (NLR). Interpretable methods pipeline Feature-Importance Analysis Based on LIME To enhance model interpretability, we applied the Local Interpretable Model-Agnostic Explanations (LIME) technique to generate local explanations and quantify the relative importance of predictive variables for glaucoma risk. As detailed in Supplemental Fig. 2 and Table 3 , the Local Interpretable Model-agnostic Explanations (LIME) technique quantified the relative contribution weights of urinary phthalate metabolites (e.g., MnBP, MEOP, MiBP, MCNP, MCOP) and baseline characteristics, encompassing age, sex, body mass index (BMI), race/ethnicity, education level, marital status, smoking history, and alcohol consumption. Table 3 LIME Feature Importance Table Feature MeanImportance LowerBound UpperBound1 UpperBound2 Age 0.045 0.049 0.051 0.049 HbA1c 0.032 0.034 0.038 0.034 C_reactiveprotein 0.032 0.034 0.038 0.034 Hypertension 0.026 0.031 0.033 0.031 BMI 0.024 0.029 0.033 0.029 MCNP_Q 0.009 0.011 0.011 0.011 MEP_Q 0.008 0.010 0.011 0.010 MEHP_Q 0.008 0.010 0.013 0.010 MnBP_Q 0.008 0.009 0.012 0.009 MEOP_Q 0.007 0.007 0.008 0.007 MBzP_Q 0.005 0.007 0.007 0.007 Smoke 0.004 0.006 0.006 0.006 MCCP_Q 0.005 0.006 0.007 0.006 Ethnic 0.003 0.004 0.005 0.004 MiBP_Q 0.004 0.004 0.006 0.004 MECPP_Q 0.003 0.004 0.005 0.004 MCOP_Q 0.001 0.003 0.004 0.003 MEHHP_Q 0.001 0.002 0.004 0.002 Alcohol.use 0.002 0.002 0.002 0.002 Education 0.001 0.001 0.001 0.001 Sex 0.000 0.000 0.000 0.000 Marital 0.000 0.000 0.000 0.000 Poverty_ratio 0.000 0.000 0.000 0.000 The results indicate that age is by far the most influential predictor, with a contribution score of approximately 0.05, markedly exceeding that of any other variable, underscoring its consistent and significant association with glaucoma classification. The next most important features were glycated haemoglobin (HbA1c) and C-reactive protein (CRP), each with an average importance of about 0.034, highlighting the roles of metabolic status and systemic inflammation in disease risk. Traditional cardiometabolic factors such as hypertension (0.031) and BMI (0.029) also ranked highly. Notably, several phthalate metabolites—MCNP, MEHP, MEP, and MnBP—demonstrated substantial importance scores, suggesting that environmental toxicants can rival classical clinical predictors in their contribution to glaucoma risk. The LIME analysis thus corroborates the biological plausibility of well-established risk factors while revealing the potential impact of environmental exposures on glaucoma susceptibility, offering important implications for public health interventions. SHAP-Based Global and Local Model Interpretation To augment model interpretability, SHapley Additive exPlanations (SHAP) elucidated individual predictor contributions to glaucoma risk within the random forest (RF) framework through quantitative and visual analytics. As shown in Fig. 2 A, the bar plot of mean absolute SHAP values ranks features by their average impact on the model's predictions. Age was identified as the most important contributor (mean |SHAP| ≈ 0.039), followed by body mass index (BMI), C-reactive protein (CRP), and HbA1c, highlighting the predictive value of metabolic and inflammatory indicators. Notably, several phthalate metabolites, including MEOP, MCNP, MEHP, MECPP, and MnBP, were also among the top-ranked variables, suggesting that environmental exposures play a non-negligible role in glaucoma risk stratification. The SHAP summary dot plot (Fig. 2 B) corroborated these findings and provided insight into the directionality of each feature's effect. Specifically, higher values of age, BMI, and HbA1c were associated with increased predicted risk, whereas lower levels of CRP tended to decrease the predicted probability of glaucoma. Among the phthalates, belonging to the upper quartile of exposure—particularly for MCNP, MEHP, and MECPP—was associated with a positive SHAP value, indicating an increased risk contribution. Individual-level waterfall plots (Supplemental Fig. 4E-H) further illustrated how feature attributions varied across participants. For example, in a 40-year-old individual (BMI = 29.2 kg/m², CRP = 0.351 mg/L; Fig. 4 E), younger age (Δlogit = − 0.0394) and low CRP (Δlogit = − 0.0277) significantly reduced the predicted risk, partially offset by high levels (Q4) of MEHHP and MEOP. Conversely, in a 66-year-old participant with obesity (BMI = 32.5 kg/m²; Fig. 4 F), high BMI substantially lowered the model prediction (Δlogit = − 0.0723), while advanced age and elevated MCNP/MECPP levels contributed to increased risk. Additional cases (Supplemental Fig. 4G-H) revealed variable importance patterns, where HbA1c or smoking status were dominant factors despite comparable environmental exposures. Collectively, SHAP-based interpretation validated the global ranking observed in LIME analysis: age remained the most influential predictor, while specific phthalate metabolites demonstrated measurable and sometimes comparable effects to classical clinical variables. The outcomes advocate incorporating environmental exposure biomarkers into glaucoma risk prediction frameworks and establish a foundation for individualized risk assessment. ALE-Based Analysis of Feature Effects on Glaucoma Risk As illustrated in Fig. 3 , accumulated local effects (ALE) analysis was conducted to visualize the marginal influence of individual features on the predicted risk of glaucoma. Among all variables, age exhibited the strongest and most consistent influence: the model predicted a steady increase in glaucoma probability with advancing age, especially beyond 60 years, indicating a clear age-related risk pattern. In contrast, variables such as sex, ethnicity, education, and marital status showed negligible effects on model predictions. BMI displayed a non-linear association, where both low and high values modestly elevated risk compared to mid-range levels. Similarly, higher HbA1c levels were positively associated with increased predicted glaucoma risk, especially in individuals with HbA1c exceeding 6.5%. For CRP, a slight U-shaped relationship was observed, indicating that both very low and very high CRP values may influence the model toward elevated risk. In terms of lifestyle factors, current smoking status had a mild positive effect on predicted risk, whereas alcohol use showed no meaningful contribution. Hypertension slightly reduced predicted risk, possibly reflecting medication use or diagnostic bias, while poverty ratio had minimal impact. Regarding phthalate metabolites, several compounds—most notably MnBP, MEHP, MCNP, and MECPP—demonstrated a trend toward increasing glaucoma risk in the highest exposure quartiles (Q1/Q2), with consistent positive ALE shifts. Conversely, MEHHP and MBzP exhibited weak or inconsistent effects, while MEP, MEOP, MiBP, MCOP, and MCCP appeared to have negligible or no association with predicted glaucoma probability across all quartiles. These findings highlight the non-linear and heterogeneous impact of phthalate exposures on glaucoma risk prediction. The ALE curves complement SHAP and LIME results, reinforcing age, HbA1c, and select phthalates (particularly MEHP and MCNP) as influential contributors, and suggest further investigation is warranted to uncover underlying biological mechanisms. Feature Interaction Analysis As illustrated in Fig. 4 , we evaluated the overall interaction strength of each variable to assess its synergistic influence with other predictors on glaucoma risk. The analysis was stratified by smoking status ("No" vs. "Yes") to investigate potential effect modification. In non-smokers, variables such as age, CRP, BMI, and hypertension showed moderate interaction strength (≥ 0.2), with age exhibiting the highest interaction effect among all predictors (interaction strength > 0.6). Among phthalate metabolites, only MnBP_Q and MECPP_Q demonstrated measurable interaction effects in this subgroup. In contrast, among smokers, more variables displayed higher interaction levels. Notably, age maintained the strongest overall interaction effect (interaction strength > 0.6), followed by CRP, HbA1c, and BMI, each with interaction strength > 0.4. Several phthalates, including MnBP_Q, MEP_Q, MECPP_Q, and MCNP_Q, also exhibited mild to moderate interaction effects, suggesting their influence may be context-dependent under smoking exposure. These findings suggest that age and CRP consistently interact with multiple variables, regardless of smoking status, and highlight the potential modulatory role of smoking in amplifying the interaction effects of metabolic and environmental risk factors on glaucoma prediction. Nonlinear Associations Between Phthalate Exposure and Glaucoma Risk Based on Restricted Cubic Spline Analysis To characterize potential nonlinear dose-response relationships, restricted cubic spline (RCS) regression with four knots was implemented to model associations between urinary phthalate metabolites and glaucoma risk. As illustrated in Fig. 5 , the exposure-response curves exhibited diverse patterns, including upward, downward, and U-shaped trends. Specifically, log-transformed concentrations of MBzP, MEHP, and MEP showed overall positive associations with the probability of glaucoma, while MCCP, MCNP, and MnBP demonstrated inverse associations. Notably, MCOP, MEHHP, MEOP, and MiBP displayed U-shaped curves, suggesting the existence of optimal exposure ranges where glaucoma risk is minimized—both excessively low and high levels may be associated with elevated risk. Sex-specific stratification revealed notable differences in exposure-response relationships. For MBzP, females exhibited a gradually decreasing trend in glaucoma risk with increasing exposure, whereas males showed a pronounced dose-dependent increase in risk. In the case of MEP, the curve in males indicated a threshold effect, with both low and high concentrations associated with higher risk, while females exhibited a monotonic increase in glaucoma risk with rising MEP levels. Interestingly, MEHP demonstrated a completely opposite trend between sexes: increasing exposure was linked to elevated glaucoma risk in males, but decreased risk in females. These observations indicate that biological sex may significantly modify the association between phthalate exposure and glaucoma development. Further mechanistic studies are warranted to elucidate the biological pathways underlying these sex-specific differences. Discussion This investigation applied explainable machine learning frameworks to elucidate the relationship between phthalate biomarkers and glaucoma susceptibility using NHANES 2011–2016 data. Among 11 evaluated models, the Gradient Boosting Machine (GBM) demonstrated superior predictive performance for glaucoma risk compared to conventional approaches like logistic regression 21 , 22 . Our interpretability analyses (SHAP, LIME, ALE) consistently identified age as the dominant predictor, aligning with glaucoma’s neurodegenerative nature and established association with aging 13 . Crucially, specific phthalate metabolites—MEHP, MCNP, and MECPP—emerged as key environmental determinants, where elevated exposure levels correlated positively with glaucoma risk 23 . This implies that these endocrine-disrupting compounds may contribute to optic nerve and trabecular meshwork pathology via oxidative stress induction and inflammatory cascade activation, corroborating animal studies linking phthalates to retinal ganglion cell injury. Methodologically, this work extends previous investigations by integrating multidimensional interpretability techniques. SHAP dependence plots confirmed age’s preeminent influence while revealing nonlinear risk factor effects, such as steepened glaucoma probability when HbA1c exceeds 6.5% 6,24 . Interaction analysis further demonstrated smoking’s capacity to amplify synergistic effects among age, CRP, and phthalates, with interaction strength exceeding 0.4 in smokers 25 . These insights overcome limitations of traditional epidemiological methods—often restricted to single-exposure assessments—by leveraging ML to disentangle complex interactions within high-dimensional environmental-metabolic-demographic datasets 26 . Our approach builds upon prior ML applications in disease prediction, demonstrating that advanced ensemble algorithms enhance predictive accuracy and robustness 13 , 27 . As an artificial intelligence component, ML extracts patterns from heterogeneous data through statistical algorithms, thereby refining decision-making and addressing methodological gaps in environmental exposure-health outcome modeling. The implementation of Permutation Feature Importance (PFI) analysis within the Random Forest framework enabled ranking of variable contributions to glaucoma risk prediction and identification of critical determinants 26 , 28 . Additionally, Accumulated Local Effects (ALE) analysis elucidated cumulative local impacts of individual predictors on glaucoma probability. Our results align with Zhang et al 26 , who established phthalate metabolite accumulation as a determinant of obesity risk using similar methodology. Within our ML-derived predictive model, we validated connections between phthalate exposure and glaucoma susceptibility. Specifically, mono-(2-ethylhexyl) phthalate (MEHP), mono-(3-carboxypropyl) phthalate (MCNP), and mono-(2-carboxymethyl) propyl phthalate (MECPP) were identified as pivotal components through PFI, ALE, and SHAP analyses. As plasticizers and solvents pervasive in consumer goods, phthalates readily enter humans via ingestion, inhalation, and dermal absorption 29 . Given their endocrine-disrupting properties, they represent significant public health concerns 30 . We observed marked positive correlations between elevated MEHP/MCNP concentrations and predicted glaucoma risk. While other environmental toxicants (e.g., heavy metals) have been extensively studied, phthalate-associated environmental contamination processes remain poorly characterized. Currently, population-level epidemiological investigations focusing on ocular disorders remain scarce 31 . The outcomes underscore the significance of phthalate metabolite concentrations in our ML model’s predictive capability and emphasize the need for heightened public awareness regarding phthalates’ chronic toxicity. Although this study provides compelling evidence linking phthalate biomarkers to heightened glaucoma susceptibility, underlying biological mechanisms require further elucidation. Due to their endocrine-disrupting properties, phthalates may influence glaucoma pathogenesis through multiple pathways. One plausible mechanism involves hormonal equilibrium disruption—particularly affecting estrogen and testosterone regulation—potentially altering optic nerve metabolism and cellular homeostasis. While epidemiological studies have primarily examined aggregate associations between environmental exposures and glaucoma risk, the role of specific metabolites identified herein constitutes an emerging research domain. The phthalate-glaucoma relationship is multifaceted and modulated by factors including genetic predisposition, age, and comorbidities like diabetes or hypertension—all established glaucoma risk enhancers 32 . Future research should clarify these interactions through longitudinal cohort studies, illuminating long-term effects of chronic phthalate exposure and cumulative ocular health impacts. Critically, phthalate exposure occurs through multiple vectors (diet, environmental pollution, consumer products). Therefore, quantifying aggregate phthalate burden from diverse sources and its cumulative effect on glaucoma pathogenesis should be prioritized. Interventions reducing exposure—such as stricter regulations on consumer product chemicals and promoting alternatives—could significantly mitigate glaucoma incidence, particularly in vulnerable populations. Several methodological constraints merit consideration. First, the cross-sectional nature of this study prevents definitive causal conclusions regarding the directionality of associations between phthalate exposure and glaucoma. Second, unmeasured confounders—including detailed dietary patterns, occupational exposures, and microenvironmental conditions—may influence results and were not fully captured in ML models 33 . Third, while GBM showed satisfactory performance, the range of evaluated ML algorithms might be limited, and model validation was not exhaustive. Additionally, moderate model performance (AUC < 0.7) necessitates larger cohorts and refined modeling for improved accuracy. Future prospective studies should validate biological mechanisms and explore clinical applicability, incorporating diverse population-based designs and mechanistic research on phthalate-glaucoma pathobiology. Concurrently, collecting multicenter clinical data could enhance predictive precision and model generalizability. Conclusion This study pioneered a data-driven machine learning framework for predicting phthalate-associated glaucoma risk. Through SHAP and allied interpretability techniques, we established the significance of specific phthalates (MEHP, MCNP) in glaucoma susceptibility stratification. The GBM model optimally characterized phthalate-glaucoma risk relationships, providing novel insights into this underexplored association. This work equips public health policymakers with evidence-based tools for early high-risk population identification and targeted interventions to reduce glaucoma incidence, advancing global efforts against this leading cause of irreversible blindness. In summary, our research introduces a novel paradigm for understanding phthalate biomarkers in glaucoma susceptibility. Findings emphasize integrating environmental exposure metrics with traditional risk factors for comprehensive glaucoma risk appraisal. Future investigations should validate these observations across diverse populations and elucidate biological mechanisms underpinning the identified associations, strengthening the evidence base for targeted prevention strategies and public health policy formulation. Declarations Acknowledgements We thank the Department of Ophthalmology of the Second Affiliated Hospital of Anhui Medical University for their collaborative and logistical work. Author contributions statement DCX conceived and designed the study. XYL was responsible for writing the original draft. YL and MHL analyzed and interpreted the patient data. SW and FY provided methodological support and software implementation. TCT, JG, ZXJ, and LMT critically reviewed and substantively revised the manuscript. All authors read and approved the final manuscript. Funding This work was supported by the National Natural Science Foundation of China (Grants 82371080, 82070986, 82471094, 82171043) and the Natural Science Foundation for Distinguished Young Scholars of Anhui Province (Grants 2308085J29, 2023AH020046). Data availability statement Publicly available datasets were analyzed in this study. This data can be found here: https://www.cdc.gov/nchs/nhanes/. Declaration of competing interest The authors declare no competing financial interests or personal relationships that could be perceived as influencing this work. Ethics approval and consent to participate declarations Ethics approval was waived as this study constituted a secondary analysis of de-identified, publicly available NHANES data. All primary study participants provided written informed consent under the original NHANES protocol approved by the NCHS Research Ethics Review Board. References Tham, Y.C. , et al. Global prevalence of glaucoma and projections of glaucoma burden through 2040: a systematic review and meta-analysis. in Ophthalmology , Vol. 121 2081-2090 (2014). Gedde, S.J. , et al. Treatment Outcomes in the Primary Tube Versus Trabeculectomy Study after 3 Years of Follow-up. in Ophthalmology , Vol. 127 333-345 (2020). Wang, X., Dai, W.W., Dang, Y.L., Hong, Y. & Zhang, C. Five Years' Outcomes of Trabeculectomy with Cross-linked Sodium Hyaluronate Gel Implantation for Chinese Glaucoma Patients. in Chin Med J (Engl) , Vol. 131 1562-1568 (2018). Quigley, H.A. The pathogenesis of optic nerve damage in glaucoma. in Trans New Orleans Acad Ophthalmol , Vol. 33 111-128 (1985). Virtanen, A., Haukka, J., Loukovaara, S. & Harju, M. Incidence of glaucoma filtration surgery from disease onset of open-angle glaucoma. in Acta Ophthalmol , Vol. 102 192-200 (2024). Almarzouki, N. Impact of Environmental Factors on Glaucoma Progression: A Systematic Review. in Clin Ophthalmol , Vol. 18 2705-2720 (2024). Boland, M.V. , et al. Comparative effectiveness of treatments for open-angle glaucoma: a systematic review for the U.S. Preventive Services Task Force. Ann Intern Med 158 , 271-279 (2013). Brassea-Perez, E. , et al. "Oxidative stress induced by phthalates in mammals: State of the art and potential biomarkers". Environ Res 206 , 112636 (2022). Serrano, S.E., Braun, J., Trasande, L., Dills, R. & Sathyanarayana, S. Phthalates and diet: a review of the food monitoring and epidemiology data. Environ Health 13 , 43 (2014). Benjamin, S. , et al. Phthalates impact human health: Epidemiological evidences and plausible mechanism of action. J Hazard Mater 340 , 360-383 (2017). Lee, S.S. & Mackey, D.A. Glaucoma - risk factors and current challenges in the diagnosis of a leading cause of visual impairment. Maturitas 163 , 15-22 (2022). Liu, X. , et al. Di-(2-ethyl hexyl) phthalate induced oxidative stress promotes microplastics mediated apoptosis and necroptosis in mice skeletal muscle by inhibiting PI3K/AKT/mTOR pathway. Toxicology 474 , 153226 (2022). Ling, X.C. , et al. Deep Learning in Glaucoma Detection and Progression Prediction: A Systematic Review and Meta-Analysis. Biomedicines 13 (2025). Deo, R.C. Machine Learning in Medicine. Circulation 132 , 1920-1930 (2015). Hood, D.C. & De Moraes, C.G. Efficacy of a Deep Learning System for Detecting Glaucomatous Optic Neuropathy Based on Color Fundus Photographs. Ophthalmology 125 , 1207-1208 (2018). Lee, E.J., Kim, T.W., Kim, J.A., Lee, S.H. & Kim, H. Predictive Modeling of Long-Term Glaucoma Progression Based on Initial Ophthalmic Data and Optic Nerve Head Characteristics. Transl Vis Sci Technol 11 , 24 (2022). Zhang, X., Li, F., Wang, D. & Lam, D.S.C. Visualization Techniques to Enhance the Explainability and Usability of Deep Learning Models in Glaucoma. Asia Pac J Ophthalmol (Phila) 12 , 347-348 (2023). Kato, K., Silva, M.J., Needham, L.L. & Calafat, A.M. Determination of 16 phthalate metabolites in urine using automated sample preparation and on-line preconcentration/high-performance liquid chromatography/tandem mass spectrometry. Anal Chem 77 , 2985-2991 (2005). Wojtuch, A., Jankowski, R. & Podlewska, S. How can SHAP values help to shape metabolic stability of chemical compounds? J Cheminform 13 , 74 (2021). Khalil, T., Khalid, S. & Syed, A.M. Review of Machine Learning techniques for glaucoma detection and prediction. in 2014 Science and Information Conference 438-442 (2014). Tao, S., Ravindranath, R. & Wang, S.Y. Predicting Glaucoma Progression to Surgery with Artificial Intelligence Survival Models. Ophthalmol Sci 3 , 100336 (2023). Chen, R. , et al. Predicting 24-hour intraocular pressure peaks and averages with machine learning. Front Med (Lausanne) 11 , 1459629 (2024). Casale, J. & Rice, A.S. Phthalates Toxicity. in StatPearls (StatPearls Publishing Copyright © 2025, StatPearls Publishing LLC., Treasure Island (FL), 2025). Chayan, T.I. , et al. Explainable AI Based Glaucoma Detection Using Transfer Learning and LIME. in 2022 IEEE Asia-Pacific Conference on Computer Science and Data Engineering (CSDE) 1-6 (2022). Jee, D., Huang, S., Kang, S. & Park, S. Polygenetic-Risk Scores for A Glaucoma Risk Interact with Blood Pressure, Glucose Control, and Carbohydrate Intake. Nutrients 12 (2020). Dong, R. , et al. Gender- and Age-Specific Relationships Between Phthalate Exposures and Obesity in Shanghai Adults. Arch Environ Contam Toxicol 73 , 431-441 (2017). Arnold, C. Looking Backward: Long-Term Lead Exposure and Risk of Glaucoma. Environ Health Perspect 127 , 54001 (2019). Strobl, C., Boulesteix, A.L., Kneib, T., Augustin, T. & Zeileis, A. Conditional variable importance for random forests. BMC Bioinformatics 9 , 307 (2008). Herr, C. , et al. Urinary di(2-ethylhexyl)phthalate (DEHP)—Metabolites and male human markers of reproductive function. International Journal of Hygiene and Environmental Health 212 , 648-653 (2009). Vandenberg, L.N. , et al. Hormones and endocrine-disrupting chemicals: low-dose effects and nonmonotonic dose responses. Endocr Rev 33 , 378-455 (2012). Wang, W. , et al. Bone Lead Levels and Risk of Incident Primary Open-Angle Glaucoma: The VA Normative Aging Study. Environ Health Perspect 126 , 087002 (2018). Tielsch, J.M. , et al. Racial variations in the prevalence of primary open-angle glaucoma. The Baltimore Eye Survey. JAMA 266 , 369-374 (1991). Vanderweele, T.J. & Arah, O.A. Bias formulas for sensitivity analysis of unmeasured confounding for general outcomes, treatments, and confounders. Epidemiology 22 , 42-52 (2011). Additional Declarations No competing interests reported. Supplementary Files Supplementarymaterial.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7009815","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":494404039,"identity":"1dfd54b3-913e-4e04-b8fa-6323e017cbfc","order_by":0,"name":"Duncheng Xiao","email":"","orcid":"","institution":"The Second Affiliated Hospital of Anhui Medical University","correspondingAuthor":false,"prefix":"","firstName":"Duncheng","middleName":"","lastName":"Xiao","suffix":""},{"id":494404041,"identity":"0080b640-8565-4a35-88a4-3e1ba72d3da0","order_by":1,"name":"Xiaoyan Liu","email":"","orcid":"","institution":"The Second Affiliated Hospital of Anhui Medical University","correspondingAuthor":false,"prefix":"","firstName":"Xiaoyan","middleName":"","lastName":"Liu","suffix":""},{"id":494404045,"identity":"03bdc731-7036-480c-8e17-e725a635de8c","order_by":2,"name":"Yu Li","email":"","orcid":"","institution":"The Second Affiliated Hospital of Anhui Medical University","correspondingAuthor":false,"prefix":"","firstName":"Yu","middleName":"","lastName":"Li","suffix":""},{"id":494404046,"identity":"a3aa44bf-2381-4b45-adfd-e59ae0ffa4c9","order_by":3,"name":"Mohan Li","email":"","orcid":"","institution":"The Second Affiliated Hospital of Anhui Medical University","correspondingAuthor":false,"prefix":"","firstName":"Mohan","middleName":"","lastName":"Li","suffix":""},{"id":494404047,"identity":"24dd8cd1-aac6-4277-a756-0c2b0ca7ff38","order_by":4,"name":"Song Wang","email":"","orcid":"","institution":"The Second Affiliated Hospital of Anhui Medical University","correspondingAuthor":false,"prefix":"","firstName":"Song","middleName":"","lastName":"Wang","suffix":""},{"id":494404048,"identity":"d24bda81-ddce-453c-a92a-dd07ba6cb792","order_by":5,"name":"Fan Yang","email":"","orcid":"","institution":"The First Affiliated Hospital of Anhui Medical University","correspondingAuthor":false,"prefix":"","firstName":"Fan","middleName":"","lastName":"Yang","suffix":""},{"id":494404049,"identity":"9a3bf8d6-7166-4519-8745-b518c81bcdcf","order_by":6,"name":"Tianchang Tao","email":"","orcid":"","institution":"The First Affiliated Hospital of Anhui Medical University","correspondingAuthor":false,"prefix":"","firstName":"Tianchang","middleName":"","lastName":"Tao","suffix":""},{"id":494404050,"identity":"52e18966-a824-41d3-8b59-d215771e1dd3","order_by":7,"name":"Jie Gao","email":"","orcid":"","institution":"The Second Affiliated Hospital of Anhui Medical University","correspondingAuthor":false,"prefix":"","firstName":"Jie","middleName":"","lastName":"Gao","suffix":""},{"id":494404051,"identity":"3cc20c0a-98c4-46b8-ba1b-c54182fb1f09","order_by":8,"name":"Zhengxuan Jiang","email":"","orcid":"","institution":"The Second Affiliated Hospital of Anhui Medical University","correspondingAuthor":false,"prefix":"","firstName":"Zhengxuan","middleName":"","lastName":"Jiang","suffix":""},{"id":494404052,"identity":"5df59705-3ff4-487e-bd88-e6fb9e92d73d","order_by":9,"name":"Liming Tao","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA3UlEQVRIiWNgGAWjYBAC+wMMjAcSGCQYGNgb2CBCBwhoASmDaOE5QIoWMEsigVgt7IcPHHiYY5EnH/n42aObbQxyfDcSGD8X4NPCk5ZwIHGbRLHh7TRz49w2BmPJGwnM0jPwOizHAKQlcePsBDNpoJbEDTcS2Jh58GnhfwPVMvP4N5CWesJaJKC2zJfgAduSYEBYyzOwXxI38OSUSeeckzCceeZhszR+hyUffPhzW13i/Pbj26Rzymzk+Y4nH/yMTwscGBwAU8A4ZWBsIEYDA4M8kepGwSgYBaNgBAIAZd1NDb9JL5EAAAAASUVORK5CYII=","orcid":"","institution":"The Second Affiliated Hospital of Anhui Medical University","correspondingAuthor":true,"prefix":"","firstName":"Liming","middleName":"","lastName":"Tao","suffix":""}],"badges":[],"createdAt":"2025-06-30 11:08:15","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7009815/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7009815/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":88260690,"identity":"d857da10-607b-4fee-a987-b197c27342db","added_by":"auto","created_at":"2025-08-04 15:17:02","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":62233,"visible":true,"origin":"","legend":"\u003cp\u003eThe ROC curves for 11 ML models on the test set.\u003c/p\u003e","description":"","filename":"1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7009815/v1/7e2deeda0f44cdb264348da8.jpg"},{"id":88260688,"identity":"eeda02a8-71d4-4946-8b86-24c6bc61e851","added_by":"auto","created_at":"2025-08-04 15:17:02","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":75222,"visible":true,"origin":"","legend":"\u003cp\u003eThe contribution of phthalate metabolites and baseline variables in the predictive model.\u003c/p\u003e","description":"","filename":"2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7009815/v1/a20bef82825c1d646a2121ac.jpg"},{"id":88260691,"identity":"392ac193-7fdb-4605-a12b-855d7858cb87","added_by":"auto","created_at":"2025-08-04 15:17:02","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":157115,"visible":true,"origin":"","legend":"\u003cp\u003eModel-derived associations between all input variables and the predicted risk of glaucoma.\u003c/p\u003e","description":"","filename":"3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7009815/v1/a0d531bd8b00a007817258b9.jpg"},{"id":88260696,"identity":"b9fae691-7ed1-4009-99f5-6a377b3607e8","added_by":"auto","created_at":"2025-08-04 15:17:02","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":83218,"visible":true,"origin":"","legend":"\u003cp\u003eInteraction effects between variables in relation to glaucoma risk.\u003c/p\u003e","description":"","filename":"4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7009815/v1/e04b82f6658904c8a054f693.jpg"},{"id":88260693,"identity":"ffaac938-0712-47a2-9568-b4dcec62896f","added_by":"auto","created_at":"2025-08-04 15:17:02","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":71523,"visible":true,"origin":"","legend":"\u003cp\u003eRestricted cubic spline analysis depicting the association between phthalate exposure and glaucoma risk.\u003c/p\u003e","description":"","filename":"5.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7009815/v1/554d93ca75fe5ff9ac6c4ac5.jpg"},{"id":94468036,"identity":"a50ac9c7-1f5d-4917-95aa-840d767104a4","added_by":"auto","created_at":"2025-10-27 15:23:53","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1802297,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7009815/v1/f47b4962-3df8-452f-ab18-c360ec1141a6.pdf"},{"id":88262076,"identity":"3c866e91-3bd2-4a23-82cb-5bf440c56df8","added_by":"auto","created_at":"2025-08-04 15:33:02","extension":"docx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":2855390,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarymaterial.docx","url":"https://assets-eu.researchsquare.com/files/rs-7009815/v1/504c33a8a54d8d59bf423131.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Machine Learning-Driven Investigation of Associations Between Phthalate Biomarkers and Glaucoma Using US NHANES Data (2011–2016)","fulltext":[{"header":"Introduction","content":"\u003cp\u003eGlaucoma, representing the primary cause of irreversible vision loss worldwide, severely compromises global visual health. According to World Health Organization estimates, approximately 80\u0026nbsp;million individuals were affected globally in 2020, with projections indicating a rise to 112\u0026nbsp;million cases by 2040\u003csup\u003e1\u003c/sup\u003e. Although therapeutic interventions such as trabeculectomy, aqueous drainage device implantation, and pharmacologic intraocular pressure reduction can effectively modulate disease progression\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e, the insidious onset of early symptoms combined with healthcare access disparities frequently results in delayed diagnosis. Consequently, many patients present with irreversible optic nerve damage and visual field deficits at initial clinical assessment\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e, substantially elevating blindness risk\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e. The pathogenesis of glaucoma involves complex interplay among multiple factors, including elevated intraocular pressure, impaired optic nerve perfusion, genetic susceptibility, and oxidative stress mechanisms\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e,\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e. Therefore, identifying modifiable environmental risk factors holds substantial public health relevance for primary prevention and progression delay strategies\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e,\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\u003cp\u003ePhthalates constitute a class of synthetic compounds extensively utilized as plasticizers in consumer products, building materials, cosmetics, and food packaging systems\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. Due to non-covalent binding within polymer matrices, these compounds readily migrate into environmental media through volatilization and leaching processes, subsequently enabling human exposure via inhalation, dietary intake, and dermal absorption routes\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e. Chronic systemic absorption of phthalates demonstrates established associations with multiple disease states, including diabetes mellitus, cardiovascular pathologies, and neurodevelopmental disorders\u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e. Functioning as prototypical endocrine-disrupting chemicals, phthalates perturb endogenous hormonal signaling through receptor agonism/antagonism, disrupting cellular metabolic processes and physiologic functions\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e. Emerging toxicological evidence implicates phthalates in ocular pathophysiology. Preclinical studies indicate that phthalate exposure may induce retinal ganglion cell apoptosis and trabecular meshwork dysfunction via oxidative stress and pro-inflammatory cascades\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e,\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e. Nevertheless, epidemiological understanding regarding associations between specific phthalate biomarkers and glaucoma susceptibility remains notably underdeveloped, with mechanistic insights conspicuously limited.\u003c/p\u003e\u003cp\u003eMachine learning represents a data-adaptive analytical paradigm capable of extracting latent association patterns from complex, high-dimensional datasets. This approach facilitates identification of risk determinants and interaction effects that often elude conventional statistical methodologies\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e. Within medical research domains, machine learning has demonstrated substantial utility in disease risk prognostication and diagnostic classification frameworks\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e. Although prior investigations have applied machine learning to elucidate ophthalmic disease mechanisms\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e, predictive modeling of glaucoma risk incorporating environmental determinants such as phthalate exposure remains scarce. Furthermore, comprehensive interpretability analyses of predictive outputs are frequently underutilized in existing literature\u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003eThis study therefore aims to leverage large-scale prospective cohort data and advanced machine learning techniques to interrogate associations between phthalate biomarker exposure profiles and glaucoma susceptibility. We implement explainable artificial intelligence methodologies\u0026mdash;including permutation feature importance quantification, accumulated local effects analysis, and SHapley Additive exPlanations (SHAP)\u0026mdash;to delineate predictor contributions and interaction networks. This integrative analytical strategy seeks to establish a robust scientific foundation for early glaucoma prevention initiatives and environmental risk factor mitigation protocols\u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003eParticipants\u003c/h2\u003e\u003cp\u003eThis study leveraged data from the National Health and Nutrition Examination Survey (NHANES)\u0026mdash;a continuous cross-sectional surveillance system conducted by the National Center for Health Statistics (NCHS), Centers for Disease Control and Prevention (CDC), capturing nationally representative health metrics of non-institutionalized U.S. residents.This initiative systematically gathers health information from non-hospitalized U.S. residents. Study data were publicly accessed via the official CDC website (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.cdc.gov/nchs/nhanes/\u003c/span\u003e\u003cspan address=\"https://www.cdc.gov/nchs/nhanes/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). Ethical clearance was granted by the NCHS Ethics Review Board, and all participants provided written informed consent.This analysis incorporated data from three NHANES cycles (2011\u0026ndash;2012, 2013\u0026ndash;2014, and 2015\u0026ndash;2016), comprising laboratory-measured urinary phthalate metabolites and paired sociodemographic records.Conducted by the National Center for Health Statistics (NCHS) at the Centers for Disease Control and Prevention (CDC), the National Health and Nutrition Examination Survey (NHANES) acquires nationally representative health metrics from non-institutionalized U.S. residents.Publicly accessible datasets were retrieved from the official CDC portal (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.cdc.gov/nchs/nhanes/\u003c/span\u003e\u003cspan address=\"https://www.cdc.gov/nchs/nhanes/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). The NCHS Ethics Review Board approved the study protocol, with written informed consent secured from all participants. Individuals lacking complete urinary phthalate or glaucoma status data were excluded, yielding a final analytical cohort of 2,113 subjects.\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eGlaucoma Diagnosis and Phthalate Exposure Assessment\u003c/h3\u003e\n\u003cp\u003eFor the diagnosis of glaucoma, participants were queried regarding prior glaucoma surgery using the survey item VIQ071: \"Have you ever had glaucoma surgery?\" Responses were dichotomized into \"yes\" or \"no.\" Individuals affirming this query were categorized as having glaucoma. Urinary concentrations of eleven phthalate metabolites were quantified to assess exposure levels. The measured compounds comprised monobutyl phthalate (MnBP), mono(2-ethyl-5-oxohexyl) phthalate (MEOP), monoisobutyl phthalate (MiBP), monocarboxyloctyl phthalate (MCOP), monocarboxynonyl phthalate (MCNP), mono(3-carboxypropyl) phthalate (MCCP), monoethyl phthalate (MEP), monobenzyl phthalate (MBzP), mono(2-ethyl-5-carboxypentyl) phthalate (MECPP), mono(2-ethylhexyl) phthalate (MEHP), and mono(2-ethyl-5-hydroxyhexyl) phthalate (MEHHP).\u003c/p\u003e\u003cp\u003eQuantification of urinary metabolites was performed using high-performance liquid chromatography (HPLC) with electrospray ionization tandem mass spectrometry (ESI-MS/MS) detection. Comprehensive methodological specifications and quality assurance procedures are documented in peer-reviewed publications and accessible through the CDC's online repository\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\n\u003ch3\u003eSociodemographic and Clinical Covariates\u003c/h3\u003e\n\u003cp\u003eSociodemographic information was obtained through self-administered questionnaires and included age, sex (male or female), race/ethnicity (non-Hispanic White, non-Hispanic Black, Mexican American, and other), marital status (married or living with a partner vs. unmarried or other), educational attainment (less than high school vs. high school or above), and poverty income ratio (PIR). A PIR\u0026thinsp;\u0026lt;\u0026thinsp;1.00 indicated income below the federal poverty line, while a PIR\u0026thinsp;\u0026ge;\u0026thinsp;1.00 indicated income at or above this threshold.\u003c/p\u003e\u003cp\u003eBody mass index (BMI) was computed as weight (kg) divided by height squared (m\u0026sup2;). Participants were categorized into three groups: underweight (BMI\u0026thinsp;\u0026lt;\u0026thinsp;18.5 kg/m\u0026sup2;), normal-to-overweight (BMI 18.5\u0026ndash;30.0 kg/m\u0026sup2;), and obese (BMI\u0026thinsp;\u0026ge;\u0026thinsp;30.0 kg/m\u0026sup2;).\u003c/p\u003e\u003cp\u003eLifestyle factors, also derived from questionnaire data, included alcohol consumption and smoking status. Alcohol consumption was classified into two groups: non-drinkers (defined as individuals who had never consumed alcohol or had consumed fewer than 12 drinks in their lifetime) and drinkers (defined as individuals who had consumed 12 or more drinks in their lifetime). Participants were stratified by smoking history into three groups: never smokers, former smokers, and current smokers.\u003c/p\u003e\u003cp\u003eClinical comorbidities considered in the analysis included hypertension, which was defined based on at least one of the following criteria: a physician-confirmed diagnosis via self-report, active use of antihypertensive medication, or elevated blood pressure measurements (systolic blood pressure\u0026thinsp;\u0026ge;\u0026thinsp;140 mmHg or diastolic blood pressure\u0026thinsp;\u0026ge;\u0026thinsp;90 mmHg). Additionally, urinary creatinine (Ucr), C-reactive protein (CRP), and hemoglobin A1c (HbA1c) were included as covariates, as these biomarkers have been identified in previous studies as potential confounders.\u003c/p\u003e\n\u003ch3\u003eConstruction of the ML model\u003c/h3\u003e\n\u003cp\u003eThe full dataset was randomly split into training (80%, N\u0026thinsp;=\u0026thinsp;1692) and testing (20%, N\u0026thinsp;=\u0026thinsp;421) subsets.In detail, we employed a five-fold cross-validation strategy, where each fold served as the test set once, with the other four folds constituting the training. The model underwent five rounds of training and validation for each iteration to obtain average performance metrics, facilitating a more accurate and thorough comparison of model performance. Subsequently, Eleven machine learning algorithms were applied to predict glaucoma using training data: Neural Network (NN), Support Vector Machine (SVM), Multi-Layer Perceptron (MLP), Gaussian Process (GP), Gradient Boosting Machine (GBM), Logistic Regression (LR), Na\u0026iuml;ve Bayes (NB), XGBoost (XGB), C5.0 Decision Trees (C5.0), k-Nearest Neighbors (KNN), and Random Forest (RF).\u003c/p\u003e\n\u003ch3\u003eEvaluation of ML model\u003c/h3\u003e\n\u003cp\u003eThe predictive performance of all eleven machine learning models was assessed using diverse performance indicators,encompassing: the receiver operating characteristic (ROC) curve with its area under the curve (AUC), accuracy, sensitivity (recall), specificity, positive and negative predictive values (PPV). These metrics were calculated based on the testing dataset using R software on a Windows platform, employing packages such as 'caret', 'randomForest', 'pROC', 'stats', 'epiR', 'ggplot2', and 'dplyr'.\u003c/p\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003eInterpretable methods pipeline of prediction models\u003c/h2\u003e\u003cp\u003eUpon selection of the most optimal predictive model, we conducted permutation feature importance (PFI), accumulated local effect (ALE) analysis, and SHapley Additive exPlanations (SHAP) to explore key predictors and their interaction effects contributing to model predictions. Both PFI and SHAP were employed to quantify the relative importance of individual variables and to identify those with the greatest influence on model performance. These explainable machine learning approaches were implemented using R software on a Windows operating system, incorporating the following R packages: 'shapviz', 'xgboost', 'lime', 'caret', 'DMwR2', 'randomForest', and 'iml'.\u003c/p\u003e\u003cp\u003eALE analysis estimates the localized effect of each feature by partitioning the feature's range into intervals and evaluating the marginal change in model output across these intervals. This allows for assessment of how feature variation influences predictions within specific regions of the feature space.\u003c/p\u003e\u003cp\u003eSHAP, grounded in cooperative game theory, quantifies each feature's marginal contribution to the model output by assigning SHAP values. To further interpret the results, SHAP dependence plots were generated to visualize the relationship between individual predictors\u0026mdash;including urinary phthalate metabolites\u0026mdash;and the predicted probability of glaucoma\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003eStatistical Analysis\u003c/h2\u003e\u003cp\u003eContinuous variables were expressed as mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation (SD) when normally distributed or median (interquartile range, IQR) otherwise. Categorical variables were presented as frequency counts with percentages. All analyses employed R statistical software (v4.1.2), with statistical significance indicated by two-tailed p-values\u0026thinsp;\u0026lt;\u0026thinsp;0.05.\u003c/p\u003e\u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\n \u003ch2\u003eParticipant Characteristics\u003c/h2\u003e\n \u003cp\u003eThe analytical cohort comprised 2,113 individuals: 1,974 unaffected controls and 139 glaucoma cases. Table \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e details their demographic, clinical, and biochemical profiles.The age distribution significantly differed between the two groups (P\u0026thinsp;\u0026lt;\u0026thinsp;0.0001). In the affected group, 69.79% were aged\u0026thinsp;\u0026ge;\u0026thinsp;70 years, while younger age groups (40\u0026ndash;59 years) were more prevalent in the unaffected group.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003ctable id=\"Tab1\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eBaseline characteristics of participants.\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003eCharacteristic\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colspan=\"4\"\u003e\n \u003cp\u003eStudy Participants, N (%)/ Mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD/ Median (Q1,Q3)\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eTotal (n\u0026thinsp;=\u0026thinsp;2113 )\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eNo (n\u0026thinsp;=\u0026thinsp;1974)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eYes (n\u0026thinsp;=\u0026thinsp;139)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eP\u003c/em\u003e\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eAge\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e40\u0026ndash;49\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e541(32.82)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e520(18.59)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e2( 1.45)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e50\u0026ndash;59\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e502(30.16)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e469(16.77)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e10( 4.93)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e60\u0026ndash;69\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e507(19.02)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e440( 9.56)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e56(22.39)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026ge;\u0026thinsp;70\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e563(18.01)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e336( 6.22)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e203(69.79)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eBMI\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.64\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e18.5\u0026ndash;30.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1276(61.38)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1192(61.44)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e84(60.36)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;18.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e43( 2.08)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e37(1.94)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e6(4.27)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026ge;\u0026thinsp;30.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e794(36.55)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e745(36.62)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e49(35.37)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eUCr, (mg/dL)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e113.52(2.63)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e113.38( 2.61)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e115.73(10.46)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eCRP, (mg/L)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.43(0.03)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.43(0.03)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.39(0.06)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.17\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eHbA1c, (%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e5.71(0.03)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e5.70(0.03)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e5.83(0.11)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eMnBP\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e35.69(4.67)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e36.38(5.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e24.41(2.12)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.43\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eMEOP\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e34.24(2.81)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e34.11(3.02)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e36.37(8.66)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eMiBP\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e9.11(0.60)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e8.96(0.53)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e11.67(3.58)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.19\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eMCNP\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e6.38(0.98)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e6.32(1.04)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e7.48(3.37)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.18\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eMCOP\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e14.09(1.47)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e14.17(1.52)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e12.81(3.04)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eMCCP\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4.54(0.26)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e4.61(0.28)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e3.41(0.47)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.49\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eMEP\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e375.62(30.88)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e360.96( 26.89)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e615.80(216.90)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.15\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eMBzP\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e12.34(0.67)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e12.20(0.68)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e14.56(2.46)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eMEHP\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e7.37(0.76)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e7.34(0.83)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e7.86(2.42)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eMECPP\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e80.00(7.31)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e80.09( 7.90)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e78.44(19.93)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eMEHHP\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e60.01(5.19)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e59.91( 5.52)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e61.76(15.52)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eGender\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1055(47.91)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e985(47.75)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e70(50.54)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFemale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1058(52.09)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e989(52.25)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e69(49.46)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eEthnic\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNon-Hispanic White\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1118(76.29)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1044(76.27)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e74(76.53)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNon-Hispanic Black\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e454(10.42)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e419(10.32)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e35(12.05)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMexican American\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e329( 5.62)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e313(5.71)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e16(4.11)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eOther\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e212( 7.67)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e198(7.69)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e14(7.31)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eMarital\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.03\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eUnmarried or other\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e782(31.09)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e720(30.79)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e62(36.03)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMarried or living with a partner\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1331(68.91)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1254(69.21)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e77(63.97)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eEducation\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.99\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eLess than High school\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e635(18.92)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e585(18.53)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e50(25.33)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHigh School or above\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1478(81.08)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1389(81.47)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e89(74.67)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003ePoverty ratio\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e0.56\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBelow Poverty line (\u0026lt;\u0026thinsp;1.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e359(10.89)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e336(11.05)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e23( 8.21)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAt or above Poverty line (\u0026ge;\u0026thinsp;1.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1754(89.11)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1638(88.95)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e116(91.79)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eSmoking status\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;\u003c/strong\u003e\u0026thinsp;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNever\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1040(49.03)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e969(49.38)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e71(43.27)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFormer\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e664(30.93)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e615(30.20)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e49(43.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCurrent\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e409(20.04)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e390(20.42)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e19(13.73)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eAlcohol drinking\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026lt;\u003c/strong\u003e\u0026thinsp;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e301(11.50)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e279(11.55)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e22(10.68)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1812(88.50)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1695(88.45)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e117(89.32)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eHypertension\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e\u0026lt;\u0026thinsp;0.0001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eNo\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e920(48.48)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e883(49.50)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e37(31.83)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eYes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1193(51.52)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e1091(50.50)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e102(68.17)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003ctfoot\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"5\"\u003eUCr: urinary creatinine, BMI: body mass index (calculated as weight in kilograms divided by height in meters squared).\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tfoot\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003eBody mass index (BMI) distribution did not differ significantly between groups (P\u0026thinsp;=\u0026thinsp;0.64), with the majority of participants having a BMI between 18.5\u0026ndash;30.0 kg/m\u0026sup2; in both groups.Biochemical markers showed notable differences. Affected individuals had higher serum creatinine (115.73\u0026thinsp;\u0026plusmn;\u0026thinsp;10.46 vs. 113.38\u0026thinsp;\u0026plusmn;\u0026thinsp;2.61 mg/dL; P\u0026thinsp;\u0026lt;\u0026thinsp;0.0001) and HbA1c levels (5.83\u0026thinsp;\u0026plusmn;\u0026thinsp;0.11% vs. 5.70\u0026thinsp;\u0026plusmn;\u0026thinsp;0.03%; P\u0026thinsp;\u0026lt;\u0026thinsp;0.0001), indicating renal and metabolic differences.Among phthalate metabolites, MEOP, MEHP, MEHHP, and MCOP showed statistically significant intergroup differences (all P values\u0026thinsp;\u0026lt;\u0026thinsp;0.001), with consistently elevated concentrations in glaucoma patients. Other metabolites such as MnBP, MEP, MBzP, and MCCP did not significantly differ (P\u0026thinsp;\u0026gt;\u0026thinsp;0.05).Sex distribution was comparable (male: 50.54% vs. 47.75%; P\u0026thinsp;\u0026lt;\u0026thinsp;0.001), while ethnicity distribution significantly varied (P\u0026thinsp;\u0026lt;\u0026thinsp;0.0001), although Non-Hispanic Whites comprised the majority in both groups.Sociodemographic variables such as marital status showed a modest difference (P\u0026thinsp;=\u0026thinsp;0.03), while education level and poverty status were comparable (P\u0026thinsp;=\u0026thinsp;0.99 and P\u0026thinsp;=\u0026thinsp;0.56, respectively).\u003c/p\u003e\n \u003cp\u003eIn lifestyle factors, former smoking was more common in the affected group (43.00% vs. 30.20%), whereas current smoking was less frequent (13.73% vs. 20.42%; P\u0026thinsp;\u0026lt;\u0026thinsp;0.0001). Alcohol consumption did not significantly differ. Hypertension exhibited substantially higher prevalence among glaucoma cases (68.17% vs. 50.50%; P\u0026thinsp;\u0026lt;\u0026thinsp;0.0001).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\n \u003ch2\u003eMachine Learning Model Performance and Comparison\u003c/h2\u003e\n \u003cp\u003eFigure \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e and Supplemental Fig. 1 depict the receiver operating characteristic (ROC) curves, illustrating the predictive performance of all 11 machine learning models for glaucoma risk in both training and test sets. These curves assess the discriminative capacity of each algorithm in predicting glaucoma susceptibility. The AUC values on the test dataset indicated that the Gradient Boosting Machine (GBM) achieved the highest discriminative performance (AUC\u0026thinsp;=\u0026thinsp;0.682), followed closely by XGBoost (AUC\u0026thinsp;=\u0026thinsp;0.679) and Random Forest (RF, AUC\u0026thinsp;=\u0026thinsp;0.678). Other models, including Gaussian Process (GP, AUC\u0026thinsp;=\u0026thinsp;0.643), Na\u0026iuml;ve Bayes (NB, AUC\u0026thinsp;=\u0026thinsp;0.637), Logistic Regression (LR, AUC\u0026thinsp;=\u0026thinsp;0.615), Multilayer Perceptron (MLP, AUC\u0026thinsp;=\u0026thinsp;0.611), Support Vector Machine (SVM, AUC\u0026thinsp;=\u0026thinsp;0.601), and Neural Network (NN, AUC\u0026thinsp;=\u0026thinsp;0.594) demonstrated moderate performance, while the K-Nearest Neighbor (KNN, AUC\u0026thinsp;=\u0026thinsp;0.538) and C5.0 (AUC\u0026thinsp;=\u0026thinsp;0.500) models showed limited predictive capability. These results suggest that ensemble-based algorithms such as GBM and RF outperform traditional classifiers in capturing the complex relationships between phthalate exposure and glaucoma risk, while simpler models like C5.0 and KNN may lack sufficient predictive power under current conditions\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e. A comprehensive summary of model performance metrics, including accuracy, precision, recall, and F1 score, is presented in Table \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e. Considering both predictive performance and interpretability, the GBM model was ultimately selected for downstream analysis.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eTable 2\u003c/strong\u003e Comparative Analysis of Discriminative Performance Across 11 Machine Learning Models\u003c/p\u003e\n \u003cp\u003e\u003cimg src=\"data:image/png;base64,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\" width=\"850\" height=\"313\"\u003e\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\n \u003cp\u003eDiscriminative capabilities of the eleven machine learning algorithms were systematically compared within the testing cohort. All predictive frameworks were constructed without employing data augmentation methodologies. Standard abbreviations include: Area Under the Receiver Operator Curve (AUC), Machine Learning (ML), Supported Vector Machine (SVM), Neural Network (NN), Multi-Layer Perceptron (MLP), Gaussian Process (GP), Gradient Boosting Machine (GBM), Logistic Regression (LR), Naive Bayes (NB), C5.0 Decision Trees (C5.0), k-Nearest Neighbor (KNN), Random Forest (RF), Positive Predictive Value (PPV), Negative Predictive Value (NPV), Positive Likelihood Ratio (PLR), and Negative Likelihood Ratio (NLR).\u003c/p\u003e\n \u003ch2\u003eInterpretable methods pipeline\u003c/h2\u003e\n \u003cdiv id=\"Sec14\" class=\"Section3\"\u003e\n \u003ch2\u003eFeature-Importance Analysis Based on LIME\u003c/h2\u003e\n \u003cp\u003eTo enhance model interpretability, we applied the Local Interpretable Model-Agnostic Explanations (LIME) technique to generate local explanations and quantify the relative importance of predictive variables for glaucoma risk. As detailed in Supplemental Fig. 2 and Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e, the Local Interpretable Model-agnostic Explanations (LIME) technique quantified the relative contribution weights of urinary phthalate metabolites (e.g., MnBP, MEOP, MiBP, MCNP, MCOP) and baseline characteristics, encompassing age, sex, body mass index (BMI), race/ethnicity, education level, marital status, smoking history, and alcohol consumption.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003ctable id=\"Tab3\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eLIME Feature Importance Table\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eFeature\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eMeanImportance\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eLowerBound\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eUpperBound1\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eUpperBound2\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAge\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.045\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.049\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.051\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.049\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHbA1c\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.032\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.034\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.038\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.034\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eC_reactiveprotein\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.032\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.034\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.038\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.034\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eHypertension\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.026\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.031\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.033\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.031\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBMI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.024\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.029\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.033\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.029\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMCNP_Q\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.009\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.011\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.011\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.011\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMEP_Q\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.008\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.010\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.011\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.010\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMEHP_Q\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.008\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.010\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.013\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.010\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMnBP_Q\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.008\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.009\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.012\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.009\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMEOP_Q\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.007\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.007\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.008\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.007\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMBzP_Q\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.005\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.007\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.007\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.007\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSmoke\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.004\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.006\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.006\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.006\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMCCP_Q\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.005\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.006\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.007\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.006\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eEthnic\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.003\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.004\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.005\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.004\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMiBP_Q\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.004\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.004\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.006\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.004\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMECPP_Q\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.003\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.004\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.005\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.004\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMCOP_Q\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.003\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.004\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.003\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMEHHP_Q\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.002\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.004\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.002\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAlcohol.use\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.002\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.002\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.002\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.002\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eEducation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSex\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.000\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMarital\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.000\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePoverty_ratio\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.000\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003eThe results indicate that age is by far the most influential predictor, with a contribution score of approximately 0.05, markedly exceeding that of any other variable, underscoring its consistent and significant association with glaucoma classification. The next most important features were glycated haemoglobin (HbA1c) and C-reactive protein (CRP), each with an average importance of about 0.034, highlighting the roles of metabolic status and systemic inflammation in disease risk. Traditional cardiometabolic factors such as hypertension (0.031) and BMI (0.029) also ranked highly.\u003c/p\u003e\n \u003cp\u003eNotably, several phthalate metabolites\u0026mdash;MCNP, MEHP, MEP, and MnBP\u0026mdash;demonstrated substantial importance scores, suggesting that environmental toxicants can rival classical clinical predictors in their contribution to glaucoma risk. The LIME analysis thus corroborates the biological plausibility of well-established risk factors while revealing the potential impact of environmental exposures on glaucoma susceptibility, offering important implications for public health interventions.\u003c/p\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\n \u003ch2\u003eSHAP-Based Global and Local Model Interpretation\u003c/h2\u003e\n \u003cp\u003eTo augment model interpretability, SHapley Additive exPlanations (SHAP) elucidated individual predictor contributions to glaucoma risk within the random forest (RF) framework through quantitative and visual analytics. As shown in Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003eA, the bar plot of mean absolute SHAP values ranks features by their average impact on the model\u0026apos;s predictions. Age was identified as the most important contributor (mean |SHAP| \u0026asymp; 0.039), followed by body mass index (BMI), C-reactive protein (CRP), and HbA1c, highlighting the predictive value of metabolic and inflammatory indicators. Notably, several phthalate metabolites, including MEOP, MCNP, MEHP, MECPP, and MnBP, were also among the top-ranked variables, suggesting that environmental exposures play a non-negligible role in glaucoma risk stratification.\u003c/p\u003e\n \u003cp\u003eThe SHAP summary dot plot (Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003eB) corroborated these findings and provided insight into the directionality of each feature\u0026apos;s effect. Specifically, higher values of age, BMI, and HbA1c were associated with increased predicted risk, whereas lower levels of CRP tended to decrease the predicted probability of glaucoma. Among the phthalates, belonging to the upper quartile of exposure\u0026mdash;particularly for MCNP, MEHP, and MECPP\u0026mdash;was associated with a positive SHAP value, indicating an increased risk contribution.\u003c/p\u003e\n \u003cp\u003eIndividual-level waterfall plots (Supplemental Fig. 4E-H) further illustrated how feature attributions varied across participants. For example, in a 40-year-old individual (BMI\u0026thinsp;=\u0026thinsp;29.2 kg/m\u0026sup2;, CRP\u0026thinsp;=\u0026thinsp;0.351 mg/L; Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003eE), younger age (\u0026Delta;logit\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.0394) and low CRP (\u0026Delta;logit\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.0277) significantly reduced the predicted risk, partially offset by high levels (Q4) of MEHHP and MEOP. Conversely, in a 66-year-old participant with obesity (BMI\u0026thinsp;=\u0026thinsp;32.5 kg/m\u0026sup2;; Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003eF), high BMI substantially lowered the model prediction (\u0026Delta;logit\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.0723), while advanced age and elevated MCNP/MECPP levels contributed to increased risk. Additional cases (Supplemental Fig. 4G-H) revealed variable importance patterns, where HbA1c or smoking status were dominant factors despite comparable environmental exposures.\u003c/p\u003e\n \u003cp\u003eCollectively, SHAP-based interpretation validated the global ranking observed in LIME analysis: age remained the most influential predictor, while specific phthalate metabolites demonstrated measurable and sometimes comparable effects to classical clinical variables. The outcomes advocate incorporating environmental exposure biomarkers into glaucoma risk prediction frameworks and establish a foundation for individualized risk assessment.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\n \u003ch2\u003eALE-Based Analysis of Feature Effects on Glaucoma Risk\u003c/h2\u003e\n \u003cp\u003eAs illustrated in Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e, accumulated local effects (ALE) analysis was conducted to visualize the marginal influence of individual features on the predicted risk of glaucoma. Among all variables, age exhibited the strongest and most consistent influence: the model predicted a steady increase in glaucoma probability with advancing age, especially beyond 60 years, indicating a clear age-related risk pattern.\u003c/p\u003e\n \u003cp\u003eIn contrast, variables such as sex, ethnicity, education, and marital status showed negligible effects on model predictions. BMI displayed a non-linear association, where both low and high values modestly elevated risk compared to mid-range levels. Similarly, higher HbA1c levels were positively associated with increased predicted glaucoma risk, especially in individuals with HbA1c exceeding 6.5%. For CRP, a slight U-shaped relationship was observed, indicating that both very low and very high CRP values may influence the model toward elevated risk.\u003c/p\u003e\n \u003cp\u003eIn terms of lifestyle factors, current smoking status had a mild positive effect on predicted risk, whereas alcohol use showed no meaningful contribution. Hypertension slightly reduced predicted risk, possibly reflecting medication use or diagnostic bias, while poverty ratio had minimal impact.\u003c/p\u003e\n \u003cp\u003eRegarding phthalate metabolites, several compounds\u0026mdash;most notably MnBP, MEHP, MCNP, and MECPP\u0026mdash;demonstrated a trend toward increasing glaucoma risk in the highest exposure quartiles (Q1/Q2), with consistent positive ALE shifts. Conversely, MEHHP and MBzP exhibited weak or inconsistent effects, while MEP, MEOP, MiBP, MCOP, and MCCP appeared to have negligible or no association with predicted glaucoma probability across all quartiles.\u003c/p\u003e\n \u003cp\u003eThese findings highlight the non-linear and heterogeneous impact of phthalate exposures on glaucoma risk prediction. The ALE curves complement SHAP and LIME results, reinforcing age, HbA1c, and select phthalates (particularly MEHP and MCNP) as influential contributors, and suggest further investigation is warranted to uncover underlying biological mechanisms.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\n \u003ch2\u003eFeature Interaction Analysis\u003c/h2\u003e\n \u003cp\u003eAs illustrated in Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e, we evaluated the overall interaction strength of each variable to assess its synergistic influence with other predictors on glaucoma risk. The analysis was stratified by smoking status (\u0026quot;No\u0026quot; vs. \u0026quot;Yes\u0026quot;) to investigate potential effect modification.\u003c/p\u003e\n \u003cp\u003eIn non-smokers, variables such as age, CRP, BMI, and hypertension showed moderate interaction strength (\u0026ge;\u0026thinsp;0.2), with age exhibiting the highest interaction effect among all predictors (interaction strength\u0026thinsp;\u0026gt;\u0026thinsp;0.6). Among phthalate metabolites, only MnBP_Q and MECPP_Q demonstrated measurable interaction effects in this subgroup.\u003c/p\u003e\n \u003cp\u003eIn contrast, among smokers, more variables displayed higher interaction levels. Notably, age maintained the strongest overall interaction effect (interaction strength\u0026thinsp;\u0026gt;\u0026thinsp;0.6), followed by CRP, HbA1c, and BMI, each with interaction strength\u0026thinsp;\u0026gt;\u0026thinsp;0.4. Several phthalates, including MnBP_Q, MEP_Q, MECPP_Q, and MCNP_Q, also exhibited mild to moderate interaction effects, suggesting their influence may be context-dependent under smoking exposure.\u003c/p\u003e\n \u003cp\u003eThese findings suggest that age and CRP consistently interact with multiple variables, regardless of smoking status, and highlight the potential modulatory role of smoking in amplifying the interaction effects of metabolic and environmental risk factors on glaucoma prediction.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\n \u003ch2\u003eNonlinear Associations Between Phthalate Exposure and Glaucoma Risk Based on Restricted Cubic Spline Analysis\u003c/h2\u003e\n \u003cp\u003eTo characterize potential nonlinear dose-response relationships, restricted cubic spline (RCS) regression with four knots was implemented to model associations between urinary phthalate metabolites and glaucoma risk. As illustrated in Fig. \u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003e, the exposure-response curves exhibited diverse patterns, including upward, downward, and U-shaped trends. Specifically, log-transformed concentrations of MBzP, MEHP, and MEP showed overall positive associations with the probability of glaucoma, while MCCP, MCNP, and MnBP demonstrated inverse associations. Notably, MCOP, MEHHP, MEOP, and MiBP displayed U-shaped curves, suggesting the existence of optimal exposure ranges where glaucoma risk is minimized\u0026mdash;both excessively low and high levels may be associated with elevated risk.\u003c/p\u003e\n \u003cp\u003eSex-specific stratification revealed notable differences in exposure-response relationships. For MBzP, females exhibited a gradually decreasing trend in glaucoma risk with increasing exposure, whereas males showed a pronounced dose-dependent increase in risk. In the case of MEP, the curve in males indicated a threshold effect, with both low and high concentrations associated with higher risk, while females exhibited a monotonic increase in glaucoma risk with rising MEP levels. Interestingly, MEHP demonstrated a completely opposite trend between sexes: increasing exposure was linked to elevated glaucoma risk in males, but decreased risk in females. These observations indicate that biological sex may significantly modify the association between phthalate exposure and glaucoma development. Further mechanistic studies are warranted to elucidate the biological pathways underlying these sex-specific differences.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis investigation applied explainable machine learning frameworks to elucidate the relationship between phthalate biomarkers and glaucoma susceptibility using NHANES 2011\u0026ndash;2016 data. Among 11 evaluated models, the Gradient Boosting Machine (GBM) demonstrated superior predictive performance for glaucoma risk compared to conventional approaches like logistic regression\u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e,\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e. Our interpretability analyses (SHAP, LIME, ALE) consistently identified age as the dominant predictor, aligning with glaucoma\u0026rsquo;s neurodegenerative nature and established association with aging\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e. Crucially, specific phthalate metabolites\u0026mdash;MEHP, MCNP, and MECPP\u0026mdash;emerged as key environmental determinants, where elevated exposure levels correlated positively with glaucoma risk\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e. This implies that these endocrine-disrupting compounds may contribute to optic nerve and trabecular meshwork pathology via oxidative stress induction and inflammatory cascade activation, corroborating animal studies linking phthalates to retinal ganglion cell injury.\u003c/p\u003e\u003cp\u003eMethodologically, this work extends previous investigations by integrating multidimensional interpretability techniques. SHAP dependence plots confirmed age\u0026rsquo;s preeminent influence while revealing nonlinear risk factor effects, such as steepened glaucoma probability when HbA1c exceeds 6.5%\u003csup\u003e6,24\u003c/sup\u003e. Interaction analysis further demonstrated smoking\u0026rsquo;s capacity to amplify synergistic effects among age, CRP, and phthalates, with interaction strength exceeding 0.4 in smokers\u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e. These insights overcome limitations of traditional epidemiological methods\u0026mdash;often restricted to single-exposure assessments\u0026mdash;by leveraging ML to disentangle complex interactions within high-dimensional environmental-metabolic-demographic datasets\u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e. Our approach builds upon prior ML applications in disease prediction, demonstrating that advanced ensemble algorithms enhance predictive accuracy and robustness\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e,\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e. As an artificial intelligence component, ML extracts patterns from heterogeneous data through statistical algorithms, thereby refining decision-making and addressing methodological gaps in environmental exposure-health outcome modeling.\u003c/p\u003e\u003cp\u003eThe implementation of Permutation Feature Importance (PFI) analysis within the Random Forest framework enabled ranking of variable contributions to glaucoma risk prediction and identification of critical determinants\u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e,\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e. Additionally, Accumulated Local Effects (ALE) analysis elucidated cumulative local impacts of individual predictors on glaucoma probability. Our results align with Zhang et al\u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e, who established phthalate metabolite accumulation as a determinant of obesity risk using similar methodology. Within our ML-derived predictive model, we validated connections between phthalate exposure and glaucoma susceptibility. Specifically, mono-(2-ethylhexyl) phthalate (MEHP), mono-(3-carboxypropyl) phthalate (MCNP), and mono-(2-carboxymethyl) propyl phthalate (MECPP) were identified as pivotal components through PFI, ALE, and SHAP analyses. As plasticizers and solvents pervasive in consumer goods, phthalates readily enter humans via ingestion, inhalation, and dermal absorption\u003csup\u003e\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e. Given their endocrine-disrupting properties, they represent significant public health concerns\u003csup\u003e\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e. We observed marked positive correlations between elevated MEHP/MCNP concentrations and predicted glaucoma risk. While other environmental toxicants (e.g., heavy metals) have been extensively studied, phthalate-associated environmental contamination processes remain poorly characterized. Currently, population-level epidemiological investigations focusing on ocular disorders remain scarce\u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e. The outcomes underscore the significance of phthalate metabolite concentrations in our ML model\u0026rsquo;s predictive capability and emphasize the need for heightened public awareness regarding phthalates\u0026rsquo; chronic toxicity.\u003c/p\u003e\u003cp\u003eAlthough this study provides compelling evidence linking phthalate biomarkers to heightened glaucoma susceptibility, underlying biological mechanisms require further elucidation. Due to their endocrine-disrupting properties, phthalates may influence glaucoma pathogenesis through multiple pathways. One plausible mechanism involves hormonal equilibrium disruption\u0026mdash;particularly affecting estrogen and testosterone regulation\u0026mdash;potentially altering optic nerve metabolism and cellular homeostasis. While epidemiological studies have primarily examined aggregate associations between environmental exposures and glaucoma risk, the role of specific metabolites identified herein constitutes an emerging research domain. The phthalate-glaucoma relationship is multifaceted and modulated by factors including genetic predisposition, age, and comorbidities like diabetes or hypertension\u0026mdash;all established glaucoma risk enhancers\u003csup\u003e\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e. Future research should clarify these interactions through longitudinal cohort studies, illuminating long-term effects of chronic phthalate exposure and cumulative ocular health impacts. Critically, phthalate exposure occurs through multiple vectors (diet, environmental pollution, consumer products). Therefore, quantifying aggregate phthalate burden from diverse sources and its cumulative effect on glaucoma pathogenesis should be prioritized. Interventions reducing exposure\u0026mdash;such as stricter regulations on consumer product chemicals and promoting alternatives\u0026mdash;could significantly mitigate glaucoma incidence, particularly in vulnerable populations.\u003c/p\u003e\u003cp\u003eSeveral methodological constraints merit consideration. First, the cross-sectional nature of this study prevents definitive causal conclusions regarding the directionality of associations between phthalate exposure and glaucoma. Second, unmeasured confounders\u0026mdash;including detailed dietary patterns, occupational exposures, and microenvironmental conditions\u0026mdash;may influence results and were not fully captured in ML models\u003csup\u003e\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u003c/sup\u003e. Third, while GBM showed satisfactory performance, the range of evaluated ML algorithms might be limited, and model validation was not exhaustive. Additionally, moderate model performance (AUC\u0026thinsp;\u0026lt;\u0026thinsp;0.7) necessitates larger cohorts and refined modeling for improved accuracy. Future prospective studies should validate biological mechanisms and explore clinical applicability, incorporating diverse population-based designs and mechanistic research on phthalate-glaucoma pathobiology. Concurrently, collecting multicenter clinical data could enhance predictive precision and model generalizability.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis study pioneered a data-driven machine learning framework for predicting phthalate-associated glaucoma risk. Through SHAP and allied interpretability techniques, we established the significance of specific phthalates (MEHP, MCNP) in glaucoma susceptibility stratification. The GBM model optimally characterized phthalate-glaucoma risk relationships, providing novel insights into this underexplored association. This work equips public health policymakers with evidence-based tools for early high-risk population identification and targeted interventions to reduce glaucoma incidence, advancing global efforts against this leading cause of irreversible blindness.\u003c/p\u003e\u003cp\u003eIn summary, our research introduces a novel paradigm for understanding phthalate biomarkers in glaucoma susceptibility. Findings emphasize integrating environmental exposure metrics with traditional risk factors for comprehensive glaucoma risk appraisal. Future investigations should validate these observations across diverse populations and elucidate biological mechanisms underpinning the identified associations, strengthening the evidence base for targeted prevention strategies and public health policy formulation.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe thank the Department of Ophthalmology of the Second Affiliated Hospital of Anhui Medical University for their collaborative and logistical work.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDCX conceived and designed the study. XYL was responsible for writing the original draft. YL and MHL analyzed and interpreted the patient data. SW and FY provided methodological support and software implementation. TCT, JG, ZXJ, and LMT critically reviewed and substantively revised the manuscript. All authors read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by the National Natural Science Foundation of China (Grants 82371080, 82070986, 82471094, 82171043) and the Natural Science Foundation for Distinguished Young Scholars of Anhui Province (Grants 2308085J29, 2023AH020046).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePublicly available datasets were analyzed in this study. This data can be found here: https://www.cdc.gov/nchs/nhanes/.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDeclaration of competing interest\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing financial interests or personal relationships that could be perceived as influencing this work.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate declarations\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eEthics approval was waived as this study constituted a secondary analysis of de-identified, publicly available NHANES data. All primary study participants provided written informed consent under the original NHANES protocol approved by the NCHS Research Ethics Review Board.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eTham, Y.C.\u003cem\u003e, et al.\u003c/em\u003e Global prevalence of glaucoma and projections of glaucoma burden through 2040: a systematic review and meta-analysis. in \u003cem\u003eOphthalmology\u003c/em\u003e, Vol. 121 2081-2090 (2014).\u003c/li\u003e\n\u003cli\u003eGedde, S.J.\u003cem\u003e, et al.\u003c/em\u003e Treatment Outcomes in the Primary Tube Versus Trabeculectomy Study after 3 Years of Follow-up. in \u003cem\u003eOphthalmology\u003c/em\u003e, Vol. 127 333-345 (2020).\u003c/li\u003e\n\u003cli\u003eWang, X., Dai, W.W., Dang, Y.L., Hong, Y. \u0026amp; Zhang, C. Five Years\u0026apos; Outcomes of Trabeculectomy with Cross-linked Sodium Hyaluronate Gel Implantation for Chinese Glaucoma Patients. in \u003cem\u003eChin Med J (Engl)\u003c/em\u003e, Vol. 131 1562-1568 (2018).\u003c/li\u003e\n\u003cli\u003eQuigley, H.A. The pathogenesis of optic nerve damage in glaucoma. in \u003cem\u003eTrans New Orleans Acad Ophthalmol\u003c/em\u003e, Vol. 33 111-128 (1985).\u003c/li\u003e\n\u003cli\u003eVirtanen, A., Haukka, J., Loukovaara, S. \u0026amp; Harju, M. Incidence of glaucoma filtration surgery from disease onset of open-angle glaucoma. in \u003cem\u003eActa Ophthalmol\u003c/em\u003e, Vol. 102 192-200 (2024).\u003c/li\u003e\n\u003cli\u003eAlmarzouki, N. Impact of Environmental Factors on Glaucoma Progression: A Systematic Review. in \u003cem\u003eClin Ophthalmol\u003c/em\u003e, Vol. 18 2705-2720 (2024).\u003c/li\u003e\n\u003cli\u003eBoland, M.V.\u003cem\u003e, et al.\u003c/em\u003e Comparative effectiveness of treatments for open-angle glaucoma: a systematic review for the U.S. Preventive Services Task Force. \u003cem\u003eAnn Intern Med\u003c/em\u003e \u003cstrong\u003e158\u003c/strong\u003e, 271-279 (2013).\u003c/li\u003e\n\u003cli\u003eBrassea-Perez, E.\u003cem\u003e, et al.\u003c/em\u003e \u0026quot;Oxidative stress induced by phthalates in mammals: State of the art and potential biomarkers\u0026quot;. \u003cem\u003eEnviron Res\u003c/em\u003e \u003cstrong\u003e206\u003c/strong\u003e, 112636 (2022).\u003c/li\u003e\n\u003cli\u003eSerrano, S.E., Braun, J., Trasande, L., Dills, R. \u0026amp; Sathyanarayana, S. Phthalates and diet: a review of the food monitoring and epidemiology data. \u003cem\u003eEnviron Health\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 43 (2014).\u003c/li\u003e\n\u003cli\u003eBenjamin, S.\u003cem\u003e, et al.\u003c/em\u003e Phthalates impact human health: Epidemiological evidences and plausible mechanism of action. \u003cem\u003eJ Hazard Mater\u003c/em\u003e \u003cstrong\u003e340\u003c/strong\u003e, 360-383 (2017).\u003c/li\u003e\n\u003cli\u003eLee, S.S. \u0026amp; Mackey, D.A. Glaucoma - risk factors and current challenges in the diagnosis of a leading cause of visual impairment. \u003cem\u003eMaturitas\u003c/em\u003e \u003cstrong\u003e163\u003c/strong\u003e, 15-22 (2022).\u003c/li\u003e\n\u003cli\u003eLiu, X.\u003cem\u003e, et al.\u003c/em\u003e Di-(2-ethyl hexyl) phthalate induced oxidative stress promotes microplastics mediated apoptosis and necroptosis in mice skeletal muscle by inhibiting PI3K/AKT/mTOR pathway. \u003cem\u003eToxicology\u003c/em\u003e \u003cstrong\u003e474\u003c/strong\u003e, 153226 (2022).\u003c/li\u003e\n\u003cli\u003eLing, X.C.\u003cem\u003e, et al.\u003c/em\u003e Deep Learning in Glaucoma Detection and Progression Prediction: A Systematic Review and Meta-Analysis. \u003cem\u003eBiomedicines\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e(2025).\u003c/li\u003e\n\u003cli\u003eDeo, R.C. Machine Learning in Medicine. \u003cem\u003eCirculation\u003c/em\u003e \u003cstrong\u003e132\u003c/strong\u003e, 1920-1930 (2015).\u003c/li\u003e\n\u003cli\u003eHood, D.C. \u0026amp; De Moraes, C.G. Efficacy of a Deep Learning System for Detecting Glaucomatous Optic Neuropathy Based on Color Fundus Photographs. \u003cem\u003eOphthalmology\u003c/em\u003e \u003cstrong\u003e125\u003c/strong\u003e, 1207-1208 (2018).\u003c/li\u003e\n\u003cli\u003eLee, E.J., Kim, T.W., Kim, J.A., Lee, S.H. \u0026amp; Kim, H. Predictive Modeling of Long-Term Glaucoma Progression Based on Initial Ophthalmic Data and Optic Nerve Head Characteristics. \u003cem\u003eTransl Vis Sci Technol\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 24 (2022).\u003c/li\u003e\n\u003cli\u003eZhang, X., Li, F., Wang, D. \u0026amp; Lam, D.S.C. Visualization Techniques to Enhance the Explainability and Usability of Deep Learning Models in Glaucoma. \u003cem\u003eAsia Pac J Ophthalmol (Phila)\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 347-348 (2023).\u003c/li\u003e\n\u003cli\u003eKato, K., Silva, M.J., Needham, L.L. \u0026amp; Calafat, A.M. Determination of 16 phthalate metabolites in urine using automated sample preparation and on-line preconcentration/high-performance liquid chromatography/tandem mass spectrometry. \u003cem\u003eAnal Chem\u003c/em\u003e \u003cstrong\u003e77\u003c/strong\u003e, 2985-2991 (2005).\u003c/li\u003e\n\u003cli\u003eWojtuch, A., Jankowski, R. \u0026amp; Podlewska, S. How can SHAP values help to shape metabolic stability of chemical compounds? \u003cem\u003eJ Cheminform\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 74 (2021).\u003c/li\u003e\n\u003cli\u003eKhalil, T., Khalid, S. \u0026amp; Syed, A.M. Review of Machine Learning techniques for glaucoma detection and prediction. in \u003cem\u003e2014 Science and Information Conference\u003c/em\u003e 438-442 (2014).\u003c/li\u003e\n\u003cli\u003eTao, S., Ravindranath, R. \u0026amp; Wang, S.Y. Predicting Glaucoma Progression to Surgery with Artificial Intelligence Survival Models. \u003cem\u003eOphthalmol Sci\u003c/em\u003e \u003cstrong\u003e3\u003c/strong\u003e, 100336 (2023).\u003c/li\u003e\n\u003cli\u003eChen, R.\u003cem\u003e, et al.\u003c/em\u003e Predicting 24-hour intraocular pressure peaks and averages with machine learning. \u003cem\u003eFront Med (Lausanne)\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 1459629 (2024).\u003c/li\u003e\n\u003cli\u003eCasale, J. \u0026amp; Rice, A.S. Phthalates Toxicity. in \u003cem\u003eStatPearls\u003c/em\u003e (StatPearls Publishing Copyright \u0026copy; 2025, StatPearls Publishing LLC., Treasure Island (FL), 2025).\u003c/li\u003e\n\u003cli\u003eChayan, T.I.\u003cem\u003e, et al.\u003c/em\u003e Explainable AI Based Glaucoma Detection Using Transfer Learning and LIME. in \u003cem\u003e2022 IEEE Asia-Pacific Conference on Computer Science and Data Engineering (CSDE)\u003c/em\u003e 1-6 (2022).\u003c/li\u003e\n\u003cli\u003eJee, D., Huang, S., Kang, S. \u0026amp; Park, S. Polygenetic-Risk Scores for A Glaucoma Risk Interact with Blood Pressure, Glucose Control, and Carbohydrate Intake. \u003cem\u003eNutrients\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e(2020).\u003c/li\u003e\n\u003cli\u003eDong, R.\u003cem\u003e, et al.\u003c/em\u003e Gender- and Age-Specific Relationships Between Phthalate Exposures and Obesity in Shanghai Adults. \u003cem\u003eArch Environ Contam Toxicol\u003c/em\u003e \u003cstrong\u003e73\u003c/strong\u003e, 431-441 (2017).\u003c/li\u003e\n\u003cli\u003eArnold, C. Looking Backward: Long-Term Lead Exposure and Risk of Glaucoma. \u003cem\u003eEnviron Health Perspect\u003c/em\u003e \u003cstrong\u003e127\u003c/strong\u003e, 54001 (2019).\u003c/li\u003e\n\u003cli\u003eStrobl, C., Boulesteix, A.L., Kneib, T., Augustin, T. \u0026amp; Zeileis, A. Conditional variable importance for random forests. \u003cem\u003eBMC Bioinformatics\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 307 (2008).\u003c/li\u003e\n\u003cli\u003eHerr, C.\u003cem\u003e, et al.\u003c/em\u003e Urinary di(2-ethylhexyl)phthalate (DEHP)\u0026mdash;Metabolites and male human markers of reproductive function. \u003cem\u003eInternational Journal of Hygiene and Environmental Health\u003c/em\u003e \u003cstrong\u003e212\u003c/strong\u003e, 648-653 (2009).\u003c/li\u003e\n\u003cli\u003eVandenberg, L.N.\u003cem\u003e, et al.\u003c/em\u003e Hormones and endocrine-disrupting chemicals: low-dose effects and nonmonotonic dose responses. \u003cem\u003eEndocr Rev\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, 378-455 (2012).\u003c/li\u003e\n\u003cli\u003eWang, W.\u003cem\u003e, et al.\u003c/em\u003e Bone Lead Levels and Risk of Incident Primary Open-Angle Glaucoma: The VA Normative Aging Study. \u003cem\u003eEnviron Health Perspect\u003c/em\u003e \u003cstrong\u003e126\u003c/strong\u003e, 087002 (2018).\u003c/li\u003e\n\u003cli\u003eTielsch, J.M.\u003cem\u003e, et al.\u003c/em\u003e Racial variations in the prevalence of primary open-angle glaucoma. The Baltimore Eye Survey. \u003cem\u003eJAMA\u003c/em\u003e \u003cstrong\u003e266\u003c/strong\u003e, 369-374 (1991).\u003c/li\u003e\n\u003cli\u003eVanderweele, T.J. \u0026amp; Arah, O.A. Bias formulas for sensitivity analysis of unmeasured confounding for general outcomes, treatments, and confounders. \u003cem\u003eEpidemiology\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 42-52 (2011).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Glaucoma, Phthalates, Explainable artificial intelligence, Risk stratification, NHANES","lastPublishedDoi":"10.21203/rs.3.rs-7009815/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7009815/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e\u003cp\u003eAs the leading cause of irreversible blindness globally, glaucoma involves a multifactorial etiology encompassing dysregulated intraocular pressure, optic neuropathy, and interactions between genetic predisposition and environmental determinants. Phthalates, ubiquitous endocrine-disrupting chemicals in plastics and personal care formulations, may adversely impact the central nervous and cardiovascular systems through hormonal interference, oxidative stress induction, and inflammatory pathway activation. Given their potential to target ocular structures including retinal neurofibers, microvasculature, and aqueous humor outflow pathways, research exploring the phthalate-glaucoma relationship remains nascent. Comprehensive analytical approaches are imperative for elucidating pathogenic mechanisms and characterizing associated risk profiles.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e\u003cp\u003eTo evaluate associations between urinary phthalate metabolite concentrations and glaucoma susceptibility, this study integrated data from large prospective cohort studies or national health databases (2011\u0026ndash;2016), incorporating phthalate biomarker measurements, glaucoma diagnostic status, and covariates (e.g., age, medical history). Eleven distinct machine learning algorithms were implemented for model development, with optimization conducted via stratified cross-validation. The optimal predictive model was selected guided by performance criteria such as the receiver operating characteristic curve's area under the curve (AUC). We implemented permutation feature importance analysis, evaluated accumulated local effects (ALE), and interpreted SHAP (SHapley Additive exPlanations) values to elucidate influential variables and their interaction patterns. Sensitivity analyses established the robustness of outcomes across subgroups.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e\u003cp\u003eThis research applied explainable AI (XAI) frameworks for examining the relationship linking phthalate biomarkers to glaucoma risk. Among 11 evaluated models, the Gradient Boosting Machine (GBM) algorithm demonstrated superior predictive capability. Age constituted the most influential risk determinant. Several phthalate metabolites\u0026mdash;specifically MEOP, MCNP, MEHP, and MECPP\u0026mdash;were identified as significant contributors to glaucoma risk stratification. The results emphasize the vital role of integrating environmental exposure biomarkers in glaucoma prognostic models and highlight the necessity for mechanistic investigations into underlying biological pathways.\u003c/p\u003e","manuscriptTitle":"Machine Learning-Driven Investigation of Associations Between Phthalate Biomarkers and Glaucoma Using US NHANES Data (2011–2016)","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-08-04 15:16:57","doi":"10.21203/rs.3.rs-7009815/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"00b8094a-8316-43ca-a415-a2f6bb8ad1f5","owner":[],"postedDate":"August 4th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-10-27T13:34:57+00:00","versionOfRecord":[],"versionCreatedAt":"2025-08-04 15:16:57","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7009815","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7009815","identity":"rs-7009815","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}