A Machine Learning Framework for Osteoarthritis Risk Prediction in Metabolic Syndrome: NHANES-Based Model Development and Clinical Tool Validation

A Machine Learning Framework for Osteoarthritis Risk Prediction in Metabolic Syndrome: NHANES-Based Model Development and Clinical Tool Validation | 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 A Machine Learning Framework for Osteoarthritis Risk Prediction in Metabolic Syndrome: NHANES-Based Model Development and Clinical Tool Validation Lintao Zhang, Xue Yun, Shangyi Geng, Jingge Wang, Zhaopeng Fan, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7261040/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 12 You are reading this latest preprint version Abstract Objective: The aim of this study was to develop a machine-learning-based predictive model for assessing osteoarthritis (OA) risk in patients with metabolic syndrome (MetS), to identify key predictors and develop a clinical risk assessment tool. Methods: Data from the National Health and Nutrition Examination Survey (NHANES, 1999-2023) were utilized to screen the core predictors in combination with LASSO(Least Absolute Shrinkage and Selection Operator) regression, and predictive models were constructed by machine learning algorithms such as XGBoost. The SHAP framework was introduced to parse variable contributions, and a column-line diagram tool was developed to enable individualized risk assessment. Results: The study included 13,250 patients with MetS and screened 14 core predictors including age, body fat percentage (BFP), and sleep disorders. The XGBoost model demonstrated the best predictive performance in the validation set (AUC=0.761), and the SHAP analysis showed that age (29.6% contribution) and BFP (14.5%) were the strongest risk drivers. Column line plots categorized risk into low, moderate, and high tertiles to guide targeted interventions. Conclusion: This study is the first to construct a dynamic prediction model of OA risk in patients with MetS, which highlights established metabolic factors contributing to OA risk and provides an evidence-based tool for the “metabolic-joint co-management” strategy, with significant potential for clinical translation. Metabolic syndrome Osteoarthritis NHANES Predictive modeling Machine learning Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 1. Background Osteoarthritis (OA) is one of the most disabling chronic musculoskeletal diseases in the world, with a prevalence rate of more than 50% in people over 60 years of age, and causing loss of joint function and pain, resulting in a direct healthcare expenditure and socio-economic burden of approximately US$150 billion annually. According to the World Health Organization, there are more than 500 million OA patients worldwide, and the prevalence rate will continue to rise with the aging of the population[1, 2].Metabolic Syndrome (MetS) is a metabolic disorder characterized by central obesity, insulin resistance, and abnormal blood pressure levels, with an increasing incidence in all regions of the world, affecting approximately one-quarter of the adult population, and is significantly associated with an increased risk of a number of diseases, including OA, which was found to be associated with increased risk of OA[3-5].Epidemiologic studies have confirmed that the risk of developing OA is 5.26 times higher in MetS patients with a mean age of 43.8 years compared to the general population[6]. Traditional knowledge of OA has focused on mechanical loading and joint wear and tear, however, new evidence suggests that its pathogenesis is closely related to metabolic imbalance[4]:①The biophysical and biochemical effects of hypertension on synovial membrane, subchondral bone and chondrocytes disrupt the homeostasis of the intra-articular environment, thereby contributing to the development and progression of OA. The endothelial-skeletal interactions involved in the pathogenesis of OA highlight the potential value of systemic modulators such as the renin-angiotensin system (RAS), endothelin, and Wnt signaling in the intervention of the disease[7]；②The body's oxidative stress caused by abnormal blood glucose levels produces excess advanced glycosylation products (AGEs) and insulin resistance, which cause damage to cellular metabolism, morphology, and cell membranes, thus accelerating OA[8]；③Lipid metabolism disorders through the adipose tissue will release lipocalin, leptin and other adipokines, triggering the abnormal elevation of TNF-α, IL-1β and IL-6 and other pro-inflammatory factors, which will destroy the cartilage extracellular matrix homeostasis through the NF-κB pathway, inhibit the cartilage repair and aggravate the inflammation of the synovium, and ultimately accelerate the progression of OA[9]；Although the MetS-OA association has been widely publicized, existing clinical prediction tools have significant limitations: previous cross-sectional and cohort studies have found bidirectional associations between the two, but metrics to assess the risk of dynamic monitoring of the two are lacking, making it difficult to capture the complex interaction effects of covariates[10]. Prospective cohort studies have also confirmed a positive risk relationship, but only one indicator (CRP) was found to be elevated to strengthen the association between the two diseases[11]. Despite the great efforts to discover biomarkers for MetS and OA, to date, predictive markers for differentiating and diagnosing early disease have yet to be elicited, and a system of diagnostically meaningful serum markers and dynamic risk assessment has yet to be developed[12, 13].This study is based on the NHANES database. The research results provide visualization tools and evidence-based evidence for the risk stratification of OA in patients with MetS[14]. 2. Materials and Methods 2.1 Study design The aim of this study is to analyze and screen the predictive risk factors for the prevalence of OA among patients with MetS through public databases, and to screen the relevant variables and construct objective diagnostic prediction models based on lasso regression and single-factor and multifactorial logistic regression, and then next to screen the best machine-learning algorithmic model of the comprehensive assessment indexes based on the included factors, and then further to visualize the importance of included various variables' importance, selecting the appropriate model to construct the column line graph, and finally selecting the recommended interventions according to the disease risk of different individuals to improve the progression of OA, slow down the change of the disease, and alleviate the burden on patients and healthcare. The data for this study were obtained from the publicly accessible NHANES database. Therefore, the requirement for a clinical trial number does not apply to this study. 2.2 Study population acquisition This retrospective study analyzed participants from 1999 to 2023 in the NHANES database, which operates in two-year cycles and collects detailed information on an average of approximately 5,000 individuals per year, with a total of 128,809 participants across all cycles. Participants were included by screening and cleaning the data through a series of conditions, and missing values were handled using a multiple checking method. The screening process is shown in Fig. 1 .The study's strict inclusion and exclusion criteria were designed to ensure the completeness and reliability of the information on the included cases, providing a reliable reference for assessing the strong associated risk factors for developing OA among Mets patients. 2.3 Study variables This study included 48 predictor variables, which we included by screening factors documented in the literature as well as by consulting specialized clinicians and searching past medical records (Supplementary Table 1 shows detailed information), involving demographic factors (gender, age, poverty level, education level), lifestyle habits (history of alcohol consumption, history of cigarette smoking, hours of sleep), medical history (history of chronic kidney disease, history of osteoporosis history of coronary artery disease, history of chronic obstructive pulmonary disease, history of sleep disorders), physical characteristics (BMI, body roundness index, percentage of body fat), common laboratory tests (HbA1c, ALT, AST, Bilirubin, Alkaline, Protein, Albumin, Globulin, Gamma, Cr, Uric-acid, Na, P acid, Na, P, Ca, K, Fe, Cl, Osmolality, H2CO3, TC, LDL-C, CRP, HS.CRP, T-fem, T-lum), novel composite indicators: inflammation indicators (ABSI, CALLY, DII, SII, TyG, WHR, WHtR, CDAI)[ 14 – 18 ]. 2.4 Diagnostic criteria for MetS and OA According to the current international diagnostic criteria for MetS, MetS can be diagnosed if three or more of the five items in the table are met[ 5 ]: (i) increased waist circumference (abdominal obesity), with waist circumference reference standards based on racial/regional differences; (ii) increased triglycerides (TG); (iii) decreased HDL cholesterol; (iv) increased blood pressure; and (v) increased fasting glucose; there are some differences in this diagnostic criterion compared to the earlier version in 2005, as shown in Table 1[ 19 ]. The diagnostic criteria are harmonized globally: different waist circumference thresholds are provided, taking into account ethnic and regional differences. In addition, abnormalities in indicators for which treatment has been received (e.g., lipid-lowering drugs, antihypertensive drugs, etc.) are also included in the diagnosis to prevent underdiagnosis. OA diagnosis was defined by self-reported physician diagnosis (MCQ160A + MCQ190/191/195). While radiographic confirmation was unavailable, this approach is widely accepted for large population studies[ 20 , 21 ]. 2.5 Model training, variable screening and model construction The original data set was randomly divided into the training set (70%) and the validation set (30%) to ensure that the subsets for constructing the model and cross-validation all came from the training set. Subsequently, the validation set was used to evaluate the model. LASSO regression is a process of compressing the regression coefficients by implementing a penalty function, imposing constraints on the absolute values of the regression coefficients to enhance the refinement of the model. Set the key parameter alpha to 1. In the Glmnet function, two lambda values were selected: min and lambda-1se. The former minimizes cross-validation errors to the greatest extent, while the latter provides a more concise model. Together, they help us balance the complexity of the model and the prediction accuracy. The LASSO variable screening is conducted on the training set. The regularization parameter λ is selected for 10-fold cross-validation in the training set, and the optimal λ value is selected by minimizing the cross-validation error. The screened variables were included in the multivariate logistic regression model to control the confounding effect among the variables, and the final variables were obtained based on the significance of the model to construct the regression model[ 22 ]. 2.6 Machine Learning and Visualizing Variables We used logistic regression modeling to predict the risk of OA in patients with MetS, defined as a binary classification problem for predicting the risk of developing the disease based on clinical characteristics. Also compare some common machine learning algorithms to construct the best prediction model[ 23 ]. Multivariate Logistic regression: logistic regression is a statistical model used for classification problems, while multivariate logistic regression deals with logistic regression problems with multiple independent variables.XGBoost (Extreme Gradient Boosting), which is an efficient machine learning algorithm based on the gradient boosting framework. XGBoost (Extreme Gradient Boosting) is a highly efficient machine learning algorithm based on a gradient boosting framework, which performs well in various data mining and machine learning competitions. To prevent overfitting, we implemented L2 regularization (lambda = 1) and set early stopping rounds = 10 based on validation loss. LightGBM (Light Gradient Boosting Machine) is a fast and efficient gradient boosting framework developed by Microsoft, which has the advantages of fast training speed, low memory consumption, etc.CatBoost (Categorical Boosting) is a machine learning algorithm based on gradient boosting. is a machine learning algorithm based on gradient boosting, which is particularly good at handling data with a large number of categorical features.RF (Random Forest), which is an integrated learning algorithm that improves the accuracy and stability of a model by constructing multiple decision trees and combining their predictions.KNN (K Nearest Neighbors), which is an instance-based learning algorithm that finds the nearest neighbor of each sample in the training set by calculating the distance between the samples to be classified and the samples in the training set, and finds the nearest neighbor of each sample in the training set. KNN (K Nearest Neighbor Algorithm), it is an instance-based learning algorithm that finds the nearest K neighbors by calculating the distance between the samples to be classified and the samples in the training set, and predicts the classes of the samples to be classified based on the classes of those neighbors. NN (Neural Network), which is a computational model that mimics the structure and function of biological neural networks and consists of a large number of interconnected neurons that are able to automatically learn features and patterns from data, and is used for various machine learning tasks, such as classification, regression, image recognition, and speech recognition[ 24 , 25 ]. The best performance of the model was evaluated by multiple indicators such as the area under the receiver operating characteristic (ROC) curve (AUC), Accuracy, Sensitivity, Specificity, Precision, and F1-Score, which reflects the true generalization ability of the model. shap (Shapley Additive Explanations) interpretability analysis is a technique used to explain the predictions generated by various ML models. This method is based on the output weights contributed by each feature to the model, allowing the behavior of the model to be interpreted at both global and local scales. This is achieved by developing an additive interpretation model that regards all features as contributors, thereby facilitating the calculation of the average incremental impact of each feature in all feasible feature combinations, and obtaining the SHAP value of each feature. It provides global and local interpretations, helping to understand the main influencing factors predicted by the model, as well as the prediction of individual sample factors. The variables in the best prediction model are ranked by importance and visualized using shape diagrams[ 26 ]. 2.7 Statistical analysis All data analyses in this study were performed using the R language (version 4.4.2), and the study analysis process was performed using a number of R software analysis packages, which are shown separately in Supplementary Table 2. Initial descriptive analysis of the raw data set was performed, and data conforming to normal distribution were expressed in the form of (mean ± standard deviation) and analyzed using the independent samples t-test; data not normally distributed were expressed in the form of (median, quartiles) and compared using the Mann-Whitney U test. Frequencies and percentages were used to describe count data, and analyses were performed using the chi-square test or Fisher's exact probability method. After identifying the relevant variables the model was trained using logistic regression and the remaining seven machine learning algorithms on the training subset of data, and during the model training process the model parameters were optimized using a 5-fold cross-validation method to balance the occurrence of overfitting and underfitting. During the study when the statistical value P < 0.05, it means that the difference is statistically significant. 3. Results 3.1 Baseline data of included participants After a series of exclusion screening processes, we included a total of 13,250 participants who could be diagnosed with MetS, of which 2,460 (18.57%) had an OA diagnosis. Firstly, we analyzed and found that 33 of the continuous type variables had a normal distribution, and the remaining 7 variables (CALLY, Alt, Ast, Gamma, CRP, HS_CRP, and SII) had a non-normally distributed, and a violin plot of the normality test is visible in Fig. 1 of the Supplementary Material. Comparisons between the patient and non-patient groups of the included participants are shown in Table 2. i) Demographic and metabolic indicators: the age of the patient group was significantly higher than that of the normal group (64.74 ± 12.03 vs. 55.67 ± 15.44 years, P < 0.001), and obesity-related indicators (BMI, body fat percentage BFP, waist-to-height ratio WHtR) and markers of insulin resistance (TyG) were significantly higher (all P < 0.001); laboratory biochemical parameters: albumin (40.83 ± 3.36 vs. 41.30 ± 3.34, P < 0.001) and globulin (29.92 ± 4.79 vs. 30.93 ± 4.83, P < 0.001) levels were significantly lower in the patient group, blood phosphorus (1.19 ± 0.19 vs. 1.17 ± 0.18), blood potassium (4.10 ± 0.42 vs. 4.05 ± 0.38) levels were significantly higher; in addition, dietary inflammation index (DII), glycosylated hemoglobin (HbA1c), and uric acid were the variables with no significant differences; ② Variables with significant differences in the non-normally distributed variables: C-reactive protein was higher in the patient group (CRP: 0.38 [0.17, 0.77] vs. 0.35 [ 0.16, 0.72], P < 0.001), and higher systemic immune-inflammatory index (SII: 518.55 [352.75, 735.54] vs 481.49 [347.22, 684.36], P < 0.001). Albumin transaminase and glutamate transaminase levels were significantly lower in the patient group. Non-significantly different variables: cally index (P = 0.086), high-sensitivity CRP (P = 0.194). (iii) Categorical variables (10): demographic variables with significant differences included: a higher proportion of males in the patient group (66.67% vs. 51.74%, P < 0.001), comorbidities and behaviors: a higher prevalence of coronary heart disease in the patient group (CHD: 14.02% vs. 7.11%), a higher prevalence of COPD in the patient group (COPD: 10.77% vs. 5.35%), chronic kidney disease (COPD: 10.77% vs. 5.35%), and a higher prevalence of chronic lung disease (COPD: 5.35%) in the patient group. ), chronic kidney disease (CKD: 38.17% vs. 30.19%), and sleep disorders (55.33% vs. 40.41%) were more common in the patient group (both P < 0.001); smoking (53.62% vs. 47.72%, P < 0.001), and there was also a significant difference between the educational levels, osteoporosis status, and only the alcohol consumption status did not have a significant differences. 3.2 Predictor screening We divided all MetS participants into the training set of 9275 cases and the test set of 3975 cases in a ratio of 7:3. The comparison between the two groups is shown in Supplementary Table 3, and the statistical analysis showed that the vast majority of the differences between the variables were not statistically significant. Firstly, the data in the training set were compared between groups, and the indicators whose differences were not statistically significant were removed (Supplementary Table 4), and the remaining variables were screened for predictors related to the outcome based on the LASSO regression method, and the screening process was demonstrated by the coefficient path diagram Fig. 2 A and the cross-validation error diagram Fig. 2 B, which were combined to efficiently screen the important in high-dimensional data by utilizing Lasso's regularization property. variables in high-dimensional data, and construct sparse regression models with excellent predictive performance, which are widely used in feature selection and dimensionality reduction analysis. After screening, it was shown that the outcome was closely related to 15 variables (Log Lambda ≈ -4.48). Finally, these variables were subjected to multifactorial logistic regression analysis, removing the variable with insignificant correlation with the outcome, “protein” (P = 0.683), and the remaining 14 variables: Smoke, Education, Age, BFP, Bilirubin, Globulin, P, Cl, T_lum, TyG, Sex, CHD, COPD, Sleep Disorder (Table 3 and Forest plot Fig. 3 ) 3.3 Machine learning model performance We trained machine learning models using 14 variables, the AUC values and 95% confidence intervals of other models were: LM = 0.744 (0.732–0.756), XGBoost = 0.791 (0.780–0.802), KNN = 0.661 (0.647–0.675), LightGBM = 0.820 (0.809–0.830), CatBoost = 0.774 (0.763–0.786), NB = 0.751 (0.739–0.763), NN = 0.750 (0.738–0.762) (Fig. 4 A). XGBoost showed the most balanced performance (Training AUC = 0.791, Test AUC = 0.761). Test set validation yielded AUC values: LM = 0.761, RF = 0.786, XGBoost = 0.761, KNN = 0.670, LightGBM = 0.760, CatBoost = 0.764, NB = 0.747, NN = 0.759 (Fig. 4 B). PR curve analysis (Fig. 4 C) showed XGBoost had stronger performance in balancing precision and recall. DCA curve (Fig. 4 D) indicated XGBoost had higher net benefit in some threshold intervals. Test set comparison of confusion matrix parameters (Accuracy, Sensitivity, Specificity, Precision, F1-Score; Table 4 ) confirmed XGBoost as the best-performing model in terms of stability and predictive performance. 3.5 SHAP-based model interpretability analysis In this study, we evaluated the relative significance of various risk factors for patients with MetS who also suffer from OA, and finally interpreted the machine learning prediction model using SHAP analysis, where the Shap value represents the degree to which each feature contributes to the prediction of the model (positive and negative values are taken as the absolute value), and the importance of the features is ranked according to the average absolute value of the Shap value. Based on the XGBoost model algorithm, we plot the bar chart of variable importance Shap values (Fig. 5 A), which is measured by the “Cover” metric (i.e., the proportion of samples covered by the feature in the node splits of the decision tree). The top three features in terms of importance are Age, BFP and Sleep Disorder, indicating that Age is the most critical predictor in the model, followed by BFP and Sleep Disorder. The remaining characteristics such as Globulin (Globulin) and Education (Education level) are of lesser importance. Figure 5 B demonstrates the overall direction and strength of the influence of each variable on the predicted outcomes of the model through the SHAP values. The horizontal axis shows the SHAP values (range − 1 to 3), with positive values indicating that the characteristics increase the risk of OA and negative values decreasing the risk. Characteristics were ranked in order of importance from top to bottom, Age (age) and BFP (body fat percentage) had the widest distribution of SHAP values and most of them were clustered in the positive region, indicating that advanced age and high body fat were associated with increased risk of OA. The SHAP values for Education were mostly distributed in the negative region, suggesting that high education level may be associated with a reduced risk. This figure comprehensively reflects the global influence pattern of the variables and provides a basis for subsequent individualized interpretation. Figure 5 C (Individual Sample SHAP Decomposition Plot) This figure demonstrates the contribution of each characteristic of a specific sample (BFP = 59.2, Age = 66, Sleep.Disorder = 0, P = 0.969, Education = 3, etc.) to the prediction of OA. The baseline value (E[f(x)] = 1.54) indicates the average predictive value of the model for all samples, while the predictive value for this individual (f(x) = 1.09) is lower than the baseline value, indicating a lower risk of OA. The specific characteristic contributions were as follows: BFP = 59.2 (SHAP = -0.431): a body fat percentage of 59.2 had a significant negative contribution to the predicted value; Age = 66 (SHAP = -0.263): an age of 66 years had a significant negative contribution to the outcome; Sleep Disorder = 0 (SHAP = + 0.244): the absence of sleep disorder had a positive contribution to the predicted value; P = 0.244. Disorder = 0 (SHAP = + 0.244): no sleep disorder contributes positively to the predicted value; P = 0.969 (SHAP = + 0.157): this eigenvalue contributes positively to the predicted value; Education = 3 (SHAP = -0.147): education level of 3 contributes negatively to the predicted value; TyG = 8.75 (SHAP = -0.0859): a TyG value of 8.75 contributes negatively to the predicted value; Smoke = 0 (SHAP = + 0.0726): non-smoking contributes positively to the predicted value; Sex = 1 (SHAP = -0.0549): gender 1 (assumed to be male) contributes negatively to the predicted value; COPD = 0 and the 5 other features contribute to the predicted value, which together affect the final prediction result. Figure 5 D (Individual Sample SHAP Aggregation Plot) This plot summarizes the feature contributions of the same sample in a more concise form. Key feature values (e.g. (P = 0.969), (Sleep.Disorder = 0), (BFP = 59.2), (Age = 66), (Education = 3), etc.) are shown at the top, and the direction and intensity of SHAP values for each feature are visualized in the bar chart below. The difference ((-0.45)) between the predicted value (f(x) = 1.09) and the baseline value (E[f(x)] = 1.54) is mainly dominated by the negative contributions of (BFP = 59.2) and (Age = 66), etc., while the positive contributions of (P = 0.969), (Sleep.Disorder = 0), etc., partially affect the difference. Overall value: Using SHAP values to analyze XGBoost from global to individual in an all-round way, cracking the problem of interpretability of black-box models and meeting the demand for transparent decision-making in the medical field.5A and 5B identify important features and influence directions, and 5C and 5D explain the prediction logics of specific samples to enhance the credibility and practicability of the model, and provide data support for personalized interventions; optimizing the feature engineering to improve the model efficiency. 3.6 Constructing Nomograms for Risk Prediction Nomogram, as a multivariate risk prediction tool, has the core advantage of transforming complex mathematical models into intuitive graphical interfaces to achieve individualized risk assessment and rapid clinical decision-making, which has more comprehensive value than traditional scoring systems or single biomarkers[ 27 , 28 ].To facilitate clinical application, we developed a nomogram (Fig. 6 ) based on the best model to visually predict the risk of OA in patients with MetS. the nomogram combines the top 8 of 14 key predictors identified in the study, including age, education, BFP, TyG, bilirubin, globulin, coronary artery disease, and sleep disorders. the nomogram is based on the best model for predicting the risk of OA in patients with MetS. the nomogram is based on the best model for predicting the risk of OA in patients with MetS. 3.7 Model validation and clinical utility Discriminant efficacy: validation set C-index was 0.815 (95% CI: 0.807–0.823), significantly better than the traditional model (AUC = 0.761), calibration: the Hosmer-Lemeshow test showed that the predicted probabilities were in high agreement with the observed probabilities (P = 0.24), and the calibration curves fit the ideal diagonal. Clinical applications: *Total Score and Risk Stratification Calculation: 0-350 points (total score after accumulating the scores of each variable), if the score is N, corresponding to the risk probability mapping: 0.2 + 0.0017 × Total Score. . Based on the column-line diagram predicting risk we classified individuals into three categories: ①Low risk (< 33.3%, total score < 78): joint protection education, as well as recommendations for healthy diet, weight control, and regular monitoring of metabolic indexes to prevent the progression of MetS. ②Medium risk (33.3%-66.6%, total score 78–272 points): weight management + sleep intervention, strengthening exercise, strict control of blood glucose, blood lipids, smoking cessation and restriction of alcohol, comprehensive improvement of MetS. ③High risk (> 66.6%, total score > 272 points): timely imaging assessment + metabolic regulation, seek medical treatment for drug and other interventions, strengthen lifestyle changes, strict management of MetS indicators to reduce the risk of OA. Example application : Patient A , CHD: 0 (no disease), score 0; Sleep.Disorder: 0 (no sleep disorder), score 0; Education: 1 (below high school), score 0; TyG (Triglyceride Glucose Index) = 9, corresponding to a score of 13; Bilirubin (Total Bilirubin) = 50, corresponding to a score of 20; BFP (Body Fat BFP (Body Fat Percentage) = 25%, corresponding to a score of 20; Age = 36 years, corresponding to a score of 20; Globulin = 80, corresponding to a score of 0; Total score: (0 + 0 + 0 + 0 + 13 + 20 + 20 + 20 + 20 + 0 = 73). According to the column chart, a total score of 73 corresponds to a Diagnostic Possibility ≈ 0.33) (low risk). Interventions Educate on joint protection and avoid strenuous exercise and joint injuries. Control body fat, maintain a healthy weight, eat a balanced diet and reduce high calorie intake. Regularly monitor metabolic indicators (e.g., TyG), ensure adequate sleep, and maintain good lifestyle habits. Patient B , CHD: 1 (diseased), score 11, Sleep Disorder: 1 (presence of sleep disorder), score 14.5, Education: 2 (high school), score 9, TyG = 11, corresponding to a score of 7.5; Bilirubin = 10, corresponding to a score of 46, BFP = 60%, score 67.5, Age = 82, score 80, Globulin = 40, score 50, total score: (11 + 14.5 + 9 + 7.5 + 46 + 67.5 + 80 + 50 = 285.5). According to the column line graph, the total score corresponds to Diagnostic Possibility ≈ 0.69) (high risk). Interventions Immediate imaging evaluation (e.g., X-ray, MRI) and thorough examination of joint structures. Strict regulation of metabolic parameters (glucose, lipids, blood pressure), combined with drug therapy if necessary. Intensive weight management and fat loss through low-calorie diet and joint-friendly exercise (e.g., swimming). Improve sleep disorders and seek guidance from sleep specialists; review joint and metabolic indexes every 1–2 months to adjust the treatment plan. 4. Discussion Taking the high prevalence and socioeconomic burden of OA and MetS as the research entry point, this paper systematically explored previous studies on the pathological association between the two on the metabolic-inflammatory axis, which is mainly mediated by metabolic disorders of the body's endo-environment (hypertension, hyperglycemia, and lipid abnormalities) and accelerates the progression of OA through inflammation, oxidative stress, and other mechanisms. In this study, 14 core predictors were screened by a joint LASSO regression-logistic regression-XGBoost model to construct the top 8 strong correlates, among which age (SHAP = 0.296) and percentage of body fat (SHAP = 0.145) were used as traditional risk factors, and their predictive efficacies were further validated in this study. The strong association between age and OA risk may stem from multiple mechanisms: on the one hand, increasing age implies aging, and the degeneration of the organism is accompanied by a decrease in the repair capacity of articular chondrocytes, the accumulation of oxidative stress, and the deterioration of the inflammatory microenvironment, which promote the progression of aging in the joints. The main relevant biomarkers are SA - β - gal (an enzyme exclusive to senescent cells), SAHF (packaging proliferation-promoting genes to trigger senescence), p53 (participating in cell-cycle regulation and stress responses), p21 (promoting cellular senescence), and p53 (promoting cellular senescence). (packaging proliferation - promoting genes to trigger senescence), p53 (participating in cell - cycle regulation and stress responses), p21 (restricting cell - cycle progression post DNA damage), and pRb (suppressing genes to halt the cell cycle). These markers collectively govern cellular senescence via enzymatic, gene - packaging, transcriptional, and cell - cycle inhibitory pathways, regulating processes from gene expression to cell - cycle inhibition. These markers collectively govern cellular senescence via enzymatic, gene - packaging, transcriptional, and cell - cycle inhibitory pathways, regulating processes from gene expression to cell - cycle arrest in senescent chondrocytes[ 29 ].On the other hand, although advanced age is the greatest risk factor for OA, OA is not an inevitable consequence of aging. Imaging changes such as bone encumbrances are common in the elderly population, but the degree of joint pain is often inconsistent with imaging severity. Musculoskeletal aging increases susceptibility to OA, but its severity is more related to joint injury, obesity, genetics, and anatomical factors of joint mechanics. Mechanisms of joint tissue aging include cellular senescence-induced secretory phenotypes and late glycosylation end-product formation in the matrix, which affect tissue mechanical properties. An in-depth understanding of the mechanisms of joint senescence may provide new targets for slowing the progression of OA, which is of public health significance in high-risk populations. age-related metabolic disorders (e.g., increased insulin resistance) may further accelerate cartilage degeneration in patients with MetS, and, although age-related progression is irreversible, measures to slow chondrocyte senescence and intervene with sensitive senescence markers could positively affect the management of this disease[ 30 ]. Previous studies have found that an imbalance of adipokines (leptin and lipocalin) induced by elevated body fat percentage regulates the pathologic process of OA. Leptin promotes synovial inflammation and abnormal remodeling of subchondral bone through the JAK/STAT and MAPK signaling pathways, while lipocalin inhibits inflammation and enhances cartilage matrix synthesis through activation of the PPAR-γ pathway. The study further showed that the association between body fat percentage and adipokines was particularly significant in knee OA, with leptin levels positively correlating with the degree of synovial inflammation and cartilage damage scores, and lipocalin levels negatively correlating with cartilage integrity[ 31 ]. In addition, a Mendelian randomization study demonstrated that sleep disorders are independent risk factors for OA[ 32 ], sleep deprivation induces the secretion of pro-inflammatory cytokines (e.g., IL-6, IL-1β, and TNF-α) [ 33 ], and excess IL-17 was found to accelerate the destruction of cartilage in a mouse model, and sleep disorders lead to abnormalities in melatonin production, which are likewise involved in the progression of OA [ 34 ]. A potential pathway between sleep disorders and OA could be metabolic disorders mediating this correlation, which would exacerbate joint degeneration and provide a new direction for multidimensional interventions. Experiments in rats have shown that globulin exhibits a unique predictive value as a recognition factor. Low globulin levels may reflect a chronic inflammatory state, which is associated with reduced complement system activity and impaired immune regulation, and supplementation of intra-articular α2macroglobulin may provide cartilage protection against arthritis[ 35 ]. Hyaluronic acid and γ-globulin have been found in human joints to protect joints in patients with advanced OA, and the protective effect is reduced at lower levels [ 36 ]. Plasma corticosteroid-binding globulin (CBG) was found to have an important role in regulating glucocorticoid bioavailability, and plasma levels of CBG were significantly reduced in rats that developed severe inflammation in a rat-induced arthritis model, which is similar to our findings. The results suggest that plasma CBG levels and integrity are important biomarkers mediating the link between inflammation onset and joint inflammation, and thus further explorations regarding the role of globin in arthritis are warranted[ 37 ]. A study preliminarily indicated that higher education level exerts a protective effect against OA occurrence, though this effect is partially counteracted by BMI and smoking (with a 35% mediating effect). In patients with MetS, education level may indirectly elevate OA risk by exacerbating metabolic disorders (e.g., insulin resistance)[ 38 ]. Notably, highly educated occupational groups (e.g., researchers, medical professionals) exhibit a higher incidence of joint pain than the general population due to long-term psychological stress; occupational stress may indirectly increase OA risk by activating the HPA axis, elevating cortisol levels, and exacerbating inflammatory responses and insulin resistance. Additionally, differences in health awareness may lead to detection bias: highly educated populations participate more actively in physical examinations and imaging screenings, resulting in significantly higher detection rates of early OA cases, and this association may be further amplified by occupational sedentariness and dietary deviations[ 39 ]. A prior study found a positive association between the TyG index and arthritis in U.S. adults under 60, but it was limited to normal-weight participants without diabetes [ 40 ]. The negative association between the TyG index and OA observed in the present study may stem from a "reverse causation" mechanism (pain-induced exercise limitation leading to metabolic suppression, i.e., passive reduction in glucose/triglycerides). Meanwhile, uncorrected metabolic heterogeneity (e.g., normal-weight metabolic abnormality subtypes) may cause misclassification bias, and chronic inflammation and energy metabolism disorders could exacerbate joint degeneration through non-insulin resistance pathways. Interestingly, there may be a threshold effect between the TyG index and OA risk: very low levels might reflect defective cartilage repair caused by energy metabolism imbalance. Although higher education showed protective effects in SHAP analysis (Fig. 5 B), its positive association in regression (Table 3 ) may reflect detection bias—educated individuals have better access to healthcare, increasing OA diagnosis rates. The inverse association between the TyG index and OA risk could also be attributed to reverse causality: joint pain limits physical activity, leading to improved glucose/lipid profiles in sedentary patients. A study developing bilirubin nanoparticle-targeted therapy for OA confirmed that bilirubin delays cartilage degeneration via antioxidant and anti-inflammatory mechanisms, suggesting that low serum total bilirubin may reduce the protective effect against OA[ 41 ]. A meta-analysis revealed a significant relationship between OA and the risk of developing CHD, with OA patients appearing to have higher cardiovascular risk. Thus, OA patients and physicians should recognize the importance of cardiovascular assessment in OA, and primary prevention and appropriate management of MetS may delay OA onset and slow its progression[ 42 ]. This paper develops the first dynamic risk prediction tool for OA in MetS patients based on a database using a predictive modeling approach with machine learning to fill a gap in clinical needs. Key Findings Interpretation: Enhanced validation of traditional factors and mechanism exploration of new factors can be used to identify high-risk MetS patients through early screening, optimize the timing of joint protection interventions, reduce the burden of further medical care after the onset of disease in patients with different risks, improve disease predictability, and mitigate the benefits of improved health economics. Complex algorithms are transformed into actionable decision-making tools for clinicians through visualized column-line diagrams, which are especially suitable for primary care scenarios. In particular, early identification of high-risk patients (e.g., total score > 272) can trigger multidisciplinary joint interventions (rheumatology-endocrinology-rehabilitation) to potentially delay OA progression through a triple strategy of metabolic modulation, anti-inflammatory therapy and joint protection. This study not only reveals the metabolic-inflammatory regulatory network of OA risk in MetS patients, but also creates a paradigm for the application of machine learning in the prediction of degenerative joint diseases. In the future, by integrating multi-omics data (e.g. metabolomics, epigenetic markers), it is expected to realize the leap from 'risk prediction' to 'precision prevention'. Limitations : Despite the significant results of this study, the following limitations need to be addressed: (1) retrospective study: the cross-sectional nature of the NHANES data limits causal inferences, e.g., the negative association between education level and OA risk may be influenced by unmeasured confounders. (ii) Population representativeness: the NHANES sample is dominated by the US population and lacks diversity data from Asia, Africa and other regions, which may affect the applicability of the model to non-Western populations. (iii) Indicator limitations: some metabolic markers may also have key roles in OA injury. Therefore, in order to further promote clinical translation, future studies should focus on the following directions: ① Prospective validation: validate the generalization ability of the model through multicenter cohorts (e.g., UK Biobank), and explore the relationship between longitudinal changes in metabolic markers (e.g., TyG index) and OA progression. (ii) Mechanistic studies: validate the molecular mechanisms of novel factors such as globin and sleep disorders using organoid or animal models, e.g. observe chondrocyte inflammatory pathway activation by lowering globin gene. (iii) Tool optimization: develop mobile risk assessment apps and integrate real-time biosensor data (e.g., sleep monitoring bracelets) to enhance the convenience and dynamic monitoring capability of the tools. Conclusion In this study, a dynamic prediction model of OA risk for MetS patients was constructed for the first time based on the NHANES database by integrating LASSO regression and XGBoost machine learning algorithm. The model demonstrated excellent discriminant efficacy in the validation set (AUC = 0.761) and identified eight core predictors, including age, body fat percentage (BFP), and sleep disorders, and its SHAP value revealed age (29.6% contribution) and body fat percentage (14.5%) as the strongest risk drivers. The column-line diagram tool enabled individualized risk stratification (low/medium/high risk), providing quantifiable clinical thresholds for intervention. The study confirms that disruption of the metabolic-inflammatory axis is a key mechanism for OA progression and validates the clinical value of novel predictors such as education level and globulin. Although limited by a retrospective design, this study provides the first evidence-based tool for a “metabolic-joint co-management” strategy, and the generalizability of the model needs to be further optimized through prospective cohort and mechanistic studies. Abbreviations OA：Osteoarthritis MetS：Metabolic syndrome CKD：Chronic Kidney Disease OP：Osteoporosis CHD：Coronary Heart Disease COPD ：Chronic Obstructive Pulmonary Disease BMI：Body Mass Index BRI：Body Roundness Index BFP：Body Fat Percentage ABSI：A Body Shape Index CALLY：CRP-albumin-lymphocyte index DII：Diet Inflammatory Index SII：Systemic Immune-Inflammation Index TyG：Triglyceride-Glucose Index WHR：Waist-Hip Ratio WHtR：Waist-to-Height Ratio Lasso：Least Absolute Shrinkage and Selection Operator LM： Logistic Multivariate Regression RF：Random Forest XGBoost：Extreme Gradient Boosting KNN：K-Nearest Neighbor LightGBM：Light Gradient Boosting Machine CatBoost：Categorical Boosting NB：Naive Bayes NN：Neural Network Shap：SHapley Additive exPlanations Declarations Authors’ contributions ： Lintao Zhang and Xue Yun contributed equally as co-first authors. Lintao Zhang and Xue Yun: conceptualization, methodology, formal analysis, investigation, writing original draft preparation; Shangyi Geng : validation, visualization, resources; Jingge Wang: methodology, software; Zhaopeng Fan ： formal analysis； Hua Guo: supervision, project administration, funding acquisition, writing review and editing; All authors agreed to submit the article to this journal, all involved made indispensable contributions to all aspects of this research work, and all agreed and approved the version submitted. Funding ： This work was supported by grants from: - The Scientific Research Program of the Health and Wellness Commission in Shaanxi Province, China (XAYC220010) - The Science and Technology Program in Xi'an, Shaanxi Province, China (22YXYJ0004) Ethics approval and consent to participate ： This study is not subject to ethical review and approval procedures because the data in the National Health and Nutrition Examination Survey (NHANES) database is publicly available and authorized by the National Center for Health Statistics (NCHS) of the United States. The research was conducted in accordance with the Declaration of Helsinki. All the data involved in the research have received ethical approval from the NCHS Research Ethics Review Committee, and all participants in the NHANES survey have provided informed consent. Clinical trial number: not applicable. Data availability ： The data used in this study are derived from the National Health and Nutrition Examination Survey (NHANES), a continuous program conducted by the U.S. Centers for Disease Control and Prevention (CDC) . Publicly available data from 1999 to 2023 are organized into 2-year cycles (e.g., 1999–2000, 2001–2002) and can be accessed through the CDC website after free registration . Data are available from the NHANES repository: (https://wwwn.cdc.gov/nchs/nhanes/). All analyses in this study were performed using publicly available data, and no restricted datasets were utilized. Competing interests ： The author declares that this research was conducted without any commercial or financial relationships and there are no potential conflicts of interest. References Tang, S., et al., Osteoarthritis. Nat Rev Dis Primers, 2025. 11 (1): p. 10. Global, regional, and national burden of osteoarthritis, 1990-2020 and projections to 2050: a systematic analysis for the Global Burden of Disease Study 2021. Lancet Rheumatol, 2023. 5 (9): p. e508-e522. Neeland, I.J., et al., Metabolic syndrome. Nat Rev Dis Primers, 2024. 10 (1): p. 77. Sampath, S.J.P., et al., Obesity, Metabolic Syndrome, and Osteoarthritis-An Updated Review. Curr Obes Rep, 2023. 12 (3): p. 308-331. Alberti, K.G., et al., Harmonizing the metabolic syndrome: a joint interim statement of the International Diabetes Federation Task Force on Epidemiology and Prevention; National Heart, Lung, and Blood Institute; American Heart Association; World Heart Federation; International Atherosclerosis Society; and International Association for the Study of Obesity. Circulation, 2009. 120 (16): p. 1640-5. Puenpatom, R.A. and T.W. Victor, Increased prevalence of metabolic syndrome in individuals with osteoarthritis: an analysis of NHANES III data. Postgrad Med, 2009. 121 (6): p. 9-20. Ching, K., et al., Hypertension meets osteoarthritis - revisiting the vascular aetiology hypothesis. Nat Rev Rheumatol, 2021. 17 (9): p. 533-549. Veronese, N., et al., Type 2 diabetes mellitus and osteoarthritis. Semin Arthritis Rheum, 2019. 49 (1): p. 9-19. Jiang, M., et al., Leptin Induced TLR4 Expression via the JAK2-STAT3 Pathway in Obesity-Related Osteoarthritis. Oxid Med Cell Longev, 2021. 2021 : p. 7385160. Liu, S.Y., et al., Bidirectional association between metabolic syndrome and osteoarthritis: a meta-analysis of observational studies. Diabetol Metab Syndr, 2020. 12 : p. 38. Zhang, S., et al., Metabolic syndrome increases osteoarthritis risk: findings from the UK Biobank prospective cohort study. BMC Public Health, 2024. 24 (1): p. 233. Lynskey, S.J., et al., Biomarkers of Osteoarthritis-A Narrative Review on Causal Links with Metabolic Syndrome. Life (Basel), 2023. 13 (3). Wang, H., et al., Metabolic Syndrome Increases the Risk for Knee Osteoarthritis: A Meta-Analysis. Evid Based Complement Alternat Med, 2016. 2016 : p. 7242478. Sharma, L., Osteoarthritis of the Knee. N Engl J Med, 2021. 384 (1): p. 51-59. Gibbs, A.J., et al., Recommendations for the management of hip and knee osteoarthritis: A systematic review of clinical practice guidelines. Osteoarthritis Cartilage, 2023. 31 (10): p. 1280-1292. Mahmoudian, A., et al., Early-stage symptomatic osteoarthritis of the knee - time for action. Nat Rev Rheumatol, 2021. 17 (10): p. 621-632. Ho, J., et al., Mendelian Randomization Studies of Lifestyle-Related Risk Factors for Osteoarthritis: A PRISMA Review and Meta-Analysis. Int J Mol Sci, 2022. 23 (19). He, Z.J., et al., Pain-Related Risk Factors Among Radiologic Stages of Knee Osteoarthritis: Data From the Osteoarthritis Initiative. Arthritis Care Res (Hoboken), 2023. 75 (6): p. 1333-1339. Alberti, K.G., P. Zimmet, and J. Shaw, The metabolic syndrome--a new worldwide definition. Lancet, 2005. 366 (9491): p. 1059-62. Zhu, S., et al., Association of smoking and osteoarthritis in US (NHANES 1999-2018). Sci Rep, 2023. 13 (1): p. 3911. Wang, J., et al., Vitamin D Status and Risk of All-Cause and Cause-Specific Mortality in Osteoarthritis Patients: Results from NHANES III and NHANES 2001-2018. Nutrients, 2022. 14 (21). Xi, L.J., et al., [Application of LASSO and its extended method in variable selection of regression analysis]. Zhonghua Yu Fang Yi Xue Za Zhi, 2023. 57 (1): p. 107-111. Lu, T., et al., Comparison of Machine Learning and Logic Regression Algorithms for Predicting Lymph Node Metastasis in Patients with Gastric Cancer: A two-Center Study. Technol Cancer Res Treat, 2024. 23 : p. 15330338231222331. Uddin, S., et al., Comparing different supervised machine learning algorithms for disease prediction. BMC Med Inform Decis Mak, 2019. 19 (1): p. 281. Spicer, J. and A.N. Sanborn, What does the mind learn? A comparison of human and machine learning representations. Curr Opin Neurobiol, 2019. 55 : p. 97-102. Ponce-Bobadilla, A.V., et al., Practical guide to SHAP analysis: Explaining supervised machine learning model predictions in drug development. Clin Transl Sci, 2024. 17 (11): p. e70056. Balachandran, V.P., et al., Nomograms in oncology: more than meets the eye. Lancet Oncol, 2015. 16 (4): p. e173-80. Huang, J., et al., Unveiling the ageing-related genes in diagnosing osteoarthritis with metabolic syndrome by integrated bioinformatics analysis and machine learning. Artif Cells Nanomed Biotechnol, 2025. 53 (1): p. 57-68. Musumeci, G., M.A. Szychlinska, and A. Mobasheri, Age-related degeneration of articular cartilage in the pathogenesis of osteoarthritis: molecular markers of senescent chondrocytes. Histol Histopathol, 2015. 30 (1): p. 1-12. Shane Anderson, A. and R.F. Loeser, Why is osteoarthritis an age-related disease? Best Pract Res Clin Rheumatol, 2010. 24 (1): p. 15-26. Zhang, C., et al., Adipokine Signaling Pathways in Osteoarthritis. Front Bioeng Biotechnol, 2022. 10 : p. 865370. Ni, J., et al., Evidence for causal effects of sleep disturbances on risk for osteoarthritis: a univariable and multivariable Mendelian randomization study. Osteoarthritis Cartilage, 2022. 30 (3): p. 443-450. Irwin, M.R., R. Olmstead, and J.E. Carroll, Sleep Disturbance, Sleep Duration, and Inflammation: A Systematic Review and Meta-Analysis of Cohort Studies and Experimental Sleep Deprivation. Biol Psychiatry, 2016. 80 (1): p. 40-52. Na, H.S., et al., Interleukin-1-Interleukin-17 Signaling Axis Induces Cartilage Destruction and Promotes Experimental Osteoarthritis. Front Immunol, 2020. 11 : p. 730. Wang, S., et al., Identification of α2-macroglobulin as a master inhibitor of cartilage-degrading factors that attenuates the progression of posttraumatic osteoarthritis. Arthritis Rheumatol, 2014. 66 (7): p. 1843-53. Park, J.Y., et al., Effects of hyaluronic acid and γ-globulin concentrations on the frictional response of human osteoarthritic articular cartilage. PLoS One, 2014. 9 (11): p. e112684. Hill, L.A., et al., Corticosteroid-binding globulin is a biomarker of inflammation onset and severity in female rats. J Endocrinol, 2016. 230 (2): p. 215-25. Gill, D., et al., Cardiometabolic traits mediating the effect of education on osteoarthritis risk: a Mendelian randomization study. Osteoarthritis Cartilage, 2021. 29 (3): p. 365-371. Tseng, P.C., et al., Additive Interaction of Work-Related Stress and Sleep Duration on Arthritis Among Middle-Aged Civil Servants. Psychol Res Behav Manag, 2021. 14 : p. 2093-2101. Yan, Y., et al., The association between triglyceride glucose index and arthritis: a population-based study. Lipids Health Dis, 2023. 22 (1): p. 132. Huang, H., et al., Opsonization Inveigles Macrophages Engulfing Carrier-Free Bilirubin/JPH203 Nanoparticles to Suppress Inflammation for Osteoarthritis Therapy. Adv Sci (Weinh), 2024. 11 (22): p. e2400713. Mathieu, S., et al., Cardiovascular profile in osteoarthritis: a meta-analysis of cardiovascular events and risk factors. Joint Bone Spine, 2019. 86 (6): p. 679-684. Tables Table 1. Diagnostic Criteria for Metabolic Syndrome (2009 Harmonized International Consensus) Component Diagnostic Criteria Units 1. Central Obesity Waist circumference ≥ specified thresholds by region/ethnicity: cm ① Europe, South America, Middle East: Men ≥94 cm;Women ≥80 cm ② South Asia, China, Japan: Men ≥90 cm; Women ≥80 cm ③ United States: Men ≥102 cm; Women ≥88 cm 2. Elevated Triglycerides (TG) ≥150 mg/dL (1.7 mmol/L) or under specific treatment for hypertriglyceridemia mg/dL (mmol/L) 3. Reduced HDL Cholesterol Men: <40 mg/dL (1.03 mmol/L); Women: <50 mg/dL (1.29 mmol/L) or under specific treatment for low HDL-C mg/dL (mmol/L) 4. Abnormal Blood Pressure Systolic ≥130 mmHg or diastolic ≥85 mmHg or diagnosed with hypertension and receiving treatment mmHg 5. Abnormal Fasting Glucose ≥100 mg/dL (5.6 mmol/L) or diagnosed with type 2 diabetes mellitus mg/dL (mmol/L) Table 2: Comparison of differences in research variables between all included patients and normal individuals (0= non-patients, 1=OA patients). Variables Total (n = 13250) 0 (n = 10790) 1 (n = 2460) P ABSI, Mean ± SD 0.08 ± 0.00 0.08 ± 0.00 0.08 ± 0.00 <.001 Age, Mean ± SD 57.35 ± 15.28 55.67 ± 15.44 64.74 ± 12.03 <.001 BMI, Mean ± SD 33.23 ± 6.89 33.11 ± 6.78 33.75 ± 7.32 <.001 BFP, Mean ± SD 41.12 ± 8.23 40.60 ± 8.21 43.40 ± 7.90 <.001 BRI, Mean ± SD 7.14 ± 2.27 7.04 ± 2.23 7.55 ± 2.39 <.001 Poverty, Mean ± SD 2.45 ± 1.57 2.43 ± 1.57 2.52 ± 1.54 0.009 DII, Mean ± SD 1.75 ± 1.77 1.74 ± 1.76 1.79 ± 1.80 0.181 HbA1c, Mean ± SD 6.38 ± 1.47 6.39 ± 1.52 6.33 ± 1.24 0.053 Bilirubin, Mean ± SD 10.32 ± 4.94 10.45 ± 4.98 9.77 ± 4.71 <.001 Alkaline, Mean ± SD 78.07 ± 28.78 77.96 ± 28.63 78.55 ± 29.43 0.365 Protein, Mean ± SD 71.96 ± 4.98 72.23 ± 4.96 70.75 ± 4.89 <.001 Albumin, Mean ± SD 41.22 ± 3.35 41.30 ± 3.34 40.83 ± 3.36 <.001 Globulin, Mean ± SD 30.75 ± 4.84 30.93 ± 4.83 29.92 ± 4.79 <.001 Cr, Mean ± SD 83.95 ± 48.51 83.28 ± 45.50 86.90 ± 59.90 <.001 Uric acid, Mean ± SD 352.02 ± 91.18 352.29 ± 91.08 350.86 ± 91.64 0.483 Na, Mean ± SD 139.44 ± 2.64 139.41 ± 2.58 139.56 ± 2.87 0.019 P, Mean ± SD 1.17 ± 0.18 1.17 ± 0.18 1.19 ± 0.19 <.001 Ca, Mean ± SD 2.34 ± 0.10 2.34 ± 0.10 2.35 ± 0.10 0.002 K, Mean ± SD 4.06 ± 0.39 4.05 ± 0.38 4.10 ± 0.42 <.001 Fe, Mean ± SD 14.52 ± 5.87 14.56 ± 5.98 14.36 ± 5.37 0.115 Cl, Mean ± SD 102.57 ± 3.38 102.63 ± 3.31 102.29 ± 3.63 <.001 Osmolality, Mean ± SD 280.63 ± 5.85 280.47 ± 5.74 281.37 ± 6.26 <.001 H2CO3, Mean ± SD 24.87 ± 2.50 24.81 ± 2.48 25.13 ± 2.56 <.001 TC, Mean ± SD 5.01 ± 1.19 5.02 ± 1.18 4.96 ± 1.24 0.036 Lldl, Mean ± SD 2.86 ± 1.06 2.87 ± 1.06 2.83 ± 1.08 0.132 T fem, Mean ± SD -0.57 ± 1.38 -0.52 ± 1.38 -0.79 ± 1.37 <.001 T lum, Mean ± SD -0.71 ± 1.76 -0.74 ± 1.75 -0.57 ± 1.80 <.001 Sleep hours, Mean ± SD 6.90 ± 1.51 6.89 ± 1.50 6.93 ± 1.54 0.175 TyG, Mean ± SD 9.18 ± 0.69 9.20 ± 0.69 9.11 ± 0.66 <.001 WHR, Mean ± SD 0.97 ± 0.07 0.97 ± 0.07 0.96 ± 0.08 <.001 WHtR, Mean ± SD 0.67 ± 0.08 0.66 ± 0.08 0.68 ± 0.09 <.001 CDAI, Mean ± SD 0.18 ± 3.73 0.18 ± 3.76 0.17 ± 3.57 0.921 CALLY, M (Q₁, Q₃) 299.13 (135.79, 665.94) 305.45 (136.67, 668.81) 284.76 (129.74, 639.17) 0.086 Alt, M (Q₁, Q₃) 22.00 (16.00, 31.00) 22.00 (17.00, 32.00) 20.00 (15.00, 27.00) <.001 Ast, M (Q₁, Q₃) 22.00 (18.00, 28.00) 22.00 (18.00, 28.00) 22.00 (18.00, 27.00) <.001 Gamma, M (Q₁, Q₃) 24.00 (17.00, 38.00) 25.00 (18.00, 38.00) 23.00 (17.00, 36.00) <.001 CRP, M (Q₁, Q₃) 0.35 (0.16, 0.73) 0.35 (0.16, 0.72) 0.38 (0.17, 0.77) <.001 HS CRP, M (Q₁, Q₃) 3.67 (1.70, 7.13) 3.66 (1.70, 7.09) 3.73 (1.75, 7.34) 0.194 SII, M (Q₁, Q₃) 486.58 (348.22, 695.25) 481.49 (347.22, 684.36) 518.55 (352.75, 735.54) <.001 Sex, n(%) <.001 Male 6027 (45.49) 5207 (48.26) 820 (33.33) Female 7223 (54.51) 5583 (51.74) 1640 (66.67) Alcohol, n(%) 0.196 NO 2228 (16.82) 1836 (17.02) 392 (15.93) YES 11022 (83.18) 8954 (82.98) 2068 (84.07) CHD, n(%) <.001 NO 12138 (91.61) 10023 (92.89) 2115 (85.98) YES 1112 (8.39) 767 (7.11) 345 (14.02) COPD, n(%) <.001 NO 12408 (93.65) 10213 (94.65) 2195 (89.23) YES 842 (6.35) 577 (5.35) 265 (10.77) CKD, n(%) <.001 NO 9054 (68.33) 7533 (69.81) 1521 (61.83) YES 4196 (31.67) 3257 (30.19) 939 (38.17) Sleep Disorder, n(%) <.001 NO 7529 (56.82) 6430 (59.59) 1099 (44.67) YES 5721 (43.18) 4360 (40.41) 1361 (55.33) Smoke, n(%) <.001 NO 6782 (51.18) 5641 (52.28) 1141 (46.38) YES 6468 (48.82) 5149 (47.72) 1319 (53.62) Education, n(%) <.001 High 6189 (46.71) 4912 (45.52) 1277 (51.91) OP, n(%) <.001 NO osteoporosis 6890 (52.00) 5677 (52.61) 1213 (49.31) Low bone mass 3634 (27.43) 2830 (26.23) 804 (32.68) Osteoporosis 2726 (20.57) 2283 (21.16) 443 (18.01) Table 3: Results of multivariate logistic regression analysis excluding confounding. characteristics OR OR（95%CI） P Age 1.051 1.047-1.055 ＜ 0.001 BFP 1.038 1.028-1.049 ＜ 0.001 Bilirubin 0.979 0.965-0.992 0.001 CHD 1.442 1.209-1.72 ＜ 0.001 Cl 0.973 0.956-0.99 0.001 COPD 1.435 1.18-1.746 ＜ 0.001 Education 1.29 1.185-1.403 ＜ 0.001 Globulin 0.955 0.936-0.974 ＜ 0.001 P 1.556 1.146-2.113 0.005 Protein 0.996 0.977-1.016 0.683 Sex 1.474 1.24-1.751 ＜ 0.001 Sleep Disorder 1.821 1.625-2.04 ＜ 0.001 Smoke 1.227 1.091-1.38 0.001 T_lum 1.092 1.057-1.129 ＜ 0.001 TyG 0.859 0.786-0.938 0.001 Table 4: Comparison of Performance between machine Learning Models and Logistic Regression Models. Model AUC (Train) AUC (Test) Accuracy Sensitivity Specificity Precision F1-Score Logistic Multivariate regression, LM 0.744 0.761 0.825 0.742 0.892 0.538 0.483 Random Forest, RF 1.000 0.786 0.814 0.871 0.908 0.482 0.431 XGBoost 0.791 0.761 0.802 0.787 0.806 0.484 0.599 K-Nearest Neighbor, KNN 0.661 0.670 0.690 0.761 0.674 0.351 0.480 LightGBM 0.820 0.760 0.722 0.773 0.710 0.382 0.511 CatBoost 0.774 0.764 0.656 0.775 0.628 0.325 0.458 Naive Bayes, NB 0.751 0.747 0.664 0.709 0.654 0.321 0.442 Neural Network, NN 0.750 0.759 0.662 0.722 0.648 0.322 0.445 Additional Declarations No competing interests reported. Supplementary Files SupplementaryMaterials.docx Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 18 Aug, 2025 Reviews received at journal 17 Aug, 2025 Reviews received at journal 15 Aug, 2025 Reviewers agreed at journal 13 Aug, 2025 Reviewers agreed at journal 12 Aug, 2025 Reviewers agreed at journal 07 Aug, 2025 Reviewers agreed at journal 04 Aug, 2025 Reviewers invited by journal 04 Aug, 2025 Editor invited by journal 01 Aug, 2025 Editor assigned by journal 01 Aug, 2025 Submission checks completed at journal 01 Aug, 2025 First submitted to journal 31 Jul, 2025 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-7261040","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":495680476,"identity":"f1133760-bf5d-43a8-a494-d0b1107d07be","order_by":0,"name":"Lintao Zhang","email":"","orcid":"","institution":"Xi 'an Fifth Hospital","correspondingAuthor":false,"prefix":"","firstName":"Lintao","middleName":"","lastName":"Zhang","suffix":""},{"id":495680478,"identity":"dd4fadaf-8c70-49db-ba66-b2fe8ad4c0fa","order_by":1,"name":"Xue Yun","email":"","orcid":"","institution":"Xi 'an Fifth Hospital","correspondingAuthor":false,"prefix":"","firstName":"Xue","middleName":"","lastName":"Yun","suffix":""},{"id":495680479,"identity":"6cff7706-f8f2-4c13-9451-89113ffb8271","order_by":2,"name":"Shangyi Geng","email":"","orcid":"","institution":"xi'an jiaotong university","correspondingAuthor":false,"prefix":"","firstName":"Shangyi","middleName":"","lastName":"Geng","suffix":""},{"id":495680481,"identity":"76a8f5e6-5364-4be3-b851-c4ef4fc280c4","order_by":3,"name":"Jingge Wang","email":"","orcid":"","institution":"Yan'an University","correspondingAuthor":false,"prefix":"","firstName":"Jingge","middleName":"","lastName":"Wang","suffix":""},{"id":495680483,"identity":"61727182-aa51-48a4-947c-3457a03ab918","order_by":4,"name":"Zhaopeng Fan","email":"","orcid":"","institution":"Xi 'an Fifth Hospital","correspondingAuthor":false,"prefix":"","firstName":"Zhaopeng","middleName":"","lastName":"Fan","suffix":""},{"id":495680485,"identity":"c71d0d43-1025-4aff-999b-23765f7c0898","order_by":5,"name":"Hua Guo","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA4ElEQVRIie2RsQrCMBCGUwqZglmvi75CcAgFu/kiV5S6VBBcOilSaAcVV32L+ghScIp7x/YR3Dqazoqpm0O++T7+++8IsVj+EMrTW40JbE+us6uxDczKAO4zUSvfueRpKZpDZFaGJJZekyVOoe6RV9Oyx2JESUAFrneOZYKsJDzf43fFzda+7kI5xLJCf0FAPQpDSnmtdArrUipkEyJgaVJQQJgBiCqWK6RuH2U+7hQhdH2CdNpD6Y6sF0NPHxnCQ8SMXUan9Na0yQa5fuWzbYMhz4/flTfYb+MWi8Vi+cgL925LewYx9tIAAAAASUVORK5CYII=","orcid":"","institution":"Xi 'an Fifth Hospital","correspondingAuthor":true,"prefix":"","firstName":"Hua","middleName":"","lastName":"Guo","suffix":""}],"badges":[],"createdAt":"2025-07-31 10:23:10","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7261040/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7261040/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":88506898,"identity":"16cfe7f3-44dc-4565-91db-a44001527d40","added_by":"auto","created_at":"2025-08-07 07:35:28","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":200002,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eshows the screening flowchart of the included studies from 1999 to 2023.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7261040/v1/7fbd71ddd263c731af04543b.png"},{"id":88506843,"identity":"06291870-1c83-4bc5-b175-9af2195cc58b","added_by":"auto","created_at":"2025-08-07 07:35:20","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":135882,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA: LASSO regression coefficient path diagram: X-axis: Log Lambda (the logarithm of the regularization parameter λ), the smaller the λ, the weaker the regularization strength, allowing more variables to enter the model. Y-axis: Regression coefficient. The colored curve represents the trajectory of the variable's coefficient varying with λ. As Log Lambda increases, the variables gradually decrease. The dashed line marks the optimal λ (determined through cross-validation), at which point the model optimally balances between bias and complexity, corresponding to the retained subset of variables. Figure 2 B: LASSO figure cross validation error: X axis with figure 2 A, Y: two deviation (goodness-of-fit measure model, the smaller deviation fitting, the better, the optimal lambda dotted line mark. Figure 2B selects the optimal λ through the deviation curve to avoid overfitting /underfitting.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7261040/v1/c09b7b08a258a4cc94359a28.png"},{"id":88506948,"identity":"5b1b6b9d-4d88-4e10-9018-ec470e767f1b","added_by":"auto","created_at":"2025-08-07 07:35:49","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":1736347,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eForest plot of multivariate logistic regression analysis, eliminating \"proteins\" that were not significantly correlated with the results (P=0.683) to reduce the influence of confounding factors among variables.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7261040/v1/73b9a2e7a82bb5019bd61273.png"},{"id":88506871,"identity":"53efab9f-debe-4211-9d56-640dfdc038c6","added_by":"auto","created_at":"2025-08-07 07:35:25","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":250985,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA: ROC curve of the training set (AUC performance), X-axis: 1-Specificity (false positive rate), Y-axis: Sensitivity (true positive rate). The larger the AUC, the better the classification ability of the model. Figure 4B: The ROC curve of the test set verifies the generalization ability of the model. Compared with Figure 4A, the AUC does not decrease significantly, indicating good generalization. Figure 4C: PR curve analysis, X-axis: Recall (Recall rate), Y-axis: Precision, evaluating the model's trade-off between recall (Recall) and Precision (Precision); Figure 4D: DCA curve analysis, X-axis: Threshold (probability threshold), Y-axis: Net Benefit, to evaluate clinical practicality and screen models with net benefits higher than \"total treatment\" and \"no treatment\". Overall value: 4 a, 4 b performance verification, PR curve from the classification performance, unbalanced data model fitness contrast, DCA is focused on the clinical utility, multidimensional evaluation model. (LM: Logistic Multivariate Regression, RF: Random Forest, XGBoost: Extreme Gradient Boosting, KNN: K-Nearest Neighbor, LightGBM: Light Gradient Boosting Machine, CatBoost: Categorical Boosting, NB: Naive Bayes, NN: \"Neural Network\". (\u003c/strong\u003eRF training AUC inflated due to overfitting）\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7261040/v1/6e1a4753166d633dbbfb051b.png"},{"id":88506848,"identity":"5f75e21b-553b-4913-85f8-4baa72da4a79","added_by":"auto","created_at":"2025-08-07 07:35:21","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":1731421,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eA: Variable importance (SHAP Value Bar Chart), X-axis: Variable Importance, Y-axis: Feature Name; Figure 5B: Global analysis graph of SHAP values, X-axis: Contribution of features to the prediction results, Y-axis: Feature names; Figure 5C: SHAP value individual decomposition graph (random sample), X-axis: The specific contribution of each feature to the prediction of this sample, accumulated as the difference between the predicted value and the baseline value, Y-axis: Feature name, the prediction logic of the decomposed individual, explaining how each feature affects the result; Figure 5D: Individual SHAP value summary graph, X-axis: Cumulative summary of feature contributions, decomposition from baseline values to final predicted values, Y-axis: None, concise summary of feature contributions of individual samples, highlighting the main influencing factors.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7261040/v1/2d6702b1fd3cbf11a072e56d.png"},{"id":88506787,"identity":"ffb28267-c724-4e0e-8c88-50d05b64f236","added_by":"auto","created_at":"2025-08-07 07:35:09","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":195133,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eNomogram construction based on the final screened and included variables (top 8),\" Nomogram for Predicting OA Risk in MetS Patients Based on NHANES: An 8-factor visualization tool integrating Age, body fat and Inflammatory Markers\".\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7261040/v1/d8dead21b180da9ffdafd98f.png"},{"id":88508371,"identity":"58c3d25a-fcaf-4bfa-acd3-b5681de74347","added_by":"auto","created_at":"2025-08-07 07:42:50","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":6948157,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7261040/v1/b8a5c718-aa36-45ac-bd16-a093e954c2a4.pdf"},{"id":88506854,"identity":"047fd349-2a8f-48b8-9b57-29d23aebc550","added_by":"auto","created_at":"2025-08-07 07:35:22","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":621734,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterials.docx","url":"https://assets-eu.researchsquare.com/files/rs-7261040/v1/4f9384a5011727aa8f9b6343.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"A Machine Learning Framework for Osteoarthritis Risk Prediction in Metabolic Syndrome: NHANES-Based Model Development and Clinical Tool Validation","fulltext":[{"header":"1. Background","content":"\u003cp\u003eOsteoarthritis (OA) is one of the most disabling chronic musculoskeletal diseases in the world, with a prevalence rate of more than 50% in people over 60 years of age, and causing loss of joint function and pain, resulting in a direct healthcare expenditure and socio-economic burden of approximately US$150 billion annually. According to the World Health Organization, there are more than 500 million OA patients worldwide, and the prevalence rate will continue to rise with the aging of the population[1, 2].Metabolic Syndrome (MetS) is a metabolic disorder characterized by central obesity, insulin resistance, and abnormal blood pressure levels, with an increasing incidence in all regions of the world, affecting approximately one-quarter of the adult population, and is significantly associated with an increased risk of a number of diseases, including OA, which was found to be associated with increased risk of OA[3-5].Epidemiologic studies have confirmed that the risk of developing OA is 5.26 times higher in MetS patients with a mean age of 43.8 years compared to the general population[6].\u003c/p\u003e\n\u003cp\u003eTraditional knowledge of OA has focused on mechanical loading and joint wear and tear, however, new evidence suggests that its pathogenesis is closely related to metabolic imbalance[4]:①The biophysical and biochemical effects of hypertension on synovial membrane, subchondral bone and chondrocytes disrupt the homeostasis of the intra-articular environment, thereby contributing to the development and progression of OA. The endothelial-skeletal interactions involved in the pathogenesis of OA highlight the potential value of systemic modulators such as the renin-angiotensin system (RAS), endothelin, and Wnt signaling in the intervention of the disease[7]；②The body\u0026apos;s oxidative stress caused by abnormal blood glucose levels produces excess advanced glycosylation products (AGEs) and insulin resistance, which cause damage to cellular metabolism, morphology, and cell membranes, thus accelerating OA[8]；③Lipid metabolism disorders through the adipose tissue will release lipocalin, leptin and other adipokines, triggering the abnormal elevation of TNF-\u0026alpha;, IL-1\u0026beta; and IL-6 and other pro-inflammatory factors, which will destroy the cartilage extracellular matrix homeostasis through the NF-\u0026kappa;B pathway, inhibit the cartilage repair and aggravate the inflammation of the synovium, and ultimately accelerate the progression of OA[9]；Although the MetS-OA association has been widely publicized, existing clinical prediction tools have significant limitations: previous cross-sectional and cohort studies have found bidirectional associations between the two, but metrics to assess the risk of dynamic monitoring of the two are lacking, making it difficult to capture the complex interaction effects of covariates[10]. Prospective cohort studies have also confirmed a positive risk relationship, but only one indicator (CRP) was found to be elevated to strengthen the association between the two diseases[11]. Despite the great efforts to discover biomarkers for MetS and OA, to date, predictive markers for differentiating and diagnosing early disease have yet to be elicited, and a system of diagnostically meaningful serum markers and dynamic risk assessment has yet to be developed[12, 13].This study is based on the NHANES database. The research results provide visualization tools and evidence-based evidence for the risk stratification of OA in patients with MetS[14].\u003c/p\u003e"},{"header":"2. Materials and Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1 Study design\u003c/h2\u003e\u003cp\u003eThe aim of this study is to analyze and screen the predictive risk factors for the prevalence of OA among patients with MetS through public databases, and to screen the relevant variables and construct objective diagnostic prediction models based on lasso regression and single-factor and multifactorial logistic regression, and then next to screen the best machine-learning algorithmic model of the comprehensive assessment indexes based on the included factors, and then further to visualize the importance of included various variables' importance, selecting the appropriate model to construct the column line graph, and finally selecting the recommended interventions according to the disease risk of different individuals to improve the progression of OA, slow down the change of the disease, and alleviate the burden on patients and healthcare. The data for this study were obtained from the publicly accessible NHANES database. Therefore, the requirement for a clinical trial number does not apply to this study.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e2.2 Study population acquisition\u003c/h2\u003e\u003cp\u003eThis retrospective study analyzed participants from 1999 to 2023 in the NHANES database, which operates in two-year cycles and collects detailed information on an average of approximately 5,000 individuals per year, with a total of 128,809 participants across all cycles. Participants were included by screening and cleaning the data through a series of conditions, and missing values were handled using a multiple checking method. The screening process is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.The study's strict inclusion and exclusion criteria were designed to ensure the completeness and reliability of the information on the included cases, providing a reliable reference for assessing the strong associated risk factors for developing OA among Mets patients.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003e2.3 Study variables\u003c/h2\u003e\u003cp\u003eThis study included 48 predictor variables, which we included by screening factors documented in the literature as well as by consulting specialized clinicians and searching past medical records (Supplementary Table\u0026nbsp;1 shows detailed information), involving demographic factors (gender, age, poverty level, education level), lifestyle habits (history of alcohol consumption, history of cigarette smoking, hours of sleep), medical history (history of chronic kidney disease, history of osteoporosis history of coronary artery disease, history of chronic obstructive pulmonary disease, history of sleep disorders), physical characteristics (BMI, body roundness index, percentage of body fat), common laboratory tests (HbA1c, ALT, AST, Bilirubin, Alkaline, Protein, Albumin, Globulin, Gamma, Cr, Uric-acid, Na, P acid, Na, P, Ca, K, Fe, Cl, Osmolality, H2CO3, TC, LDL-C, CRP, HS.CRP, T-fem, T-lum), novel composite indicators: inflammation indicators (ABSI, CALLY, DII, SII, TyG, WHR, WHtR, CDAI)[\u003cspan additionalcitationids=\"CR15 CR16 CR17\" citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e].\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\u003ch2\u003e2.4 Diagnostic criteria for MetS and OA\u003c/h2\u003e\u003cp\u003eAccording to the current international diagnostic criteria for MetS, MetS can be diagnosed if three or more of the five items in the table are met[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]:\u003c/p\u003e\u003cp\u003e(i) increased waist circumference (abdominal obesity), with waist circumference reference standards based on racial/regional differences; (ii) increased triglycerides (TG); (iii) decreased HDL cholesterol; (iv) increased blood pressure; and (v) increased fasting glucose; there are some differences in this diagnostic criterion compared to the earlier version in 2005, as shown in Table\u0026nbsp;1[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. The diagnostic criteria are harmonized globally: different waist circumference thresholds are provided, taking into account ethnic and regional differences. In addition, abnormalities in indicators for which treatment has been received (e.g., lipid-lowering drugs, antihypertensive drugs, etc.) are also included in the diagnosis to prevent underdiagnosis.\u003c/p\u003e\u003cp\u003eOA diagnosis was defined by self-reported physician diagnosis (MCQ160A\u0026thinsp;+\u0026thinsp;MCQ190/191/195). While radiographic confirmation was unavailable, this approach is widely accepted for large population studies[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e].\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\u003ch2\u003e2.5 Model training, variable screening and model construction\u003c/h2\u003e\u003cp\u003eThe original data set was randomly divided into the training set (70%) and the validation set (30%) to ensure that the subsets for constructing the model and cross-validation all came from the training set. Subsequently, the validation set was used to evaluate the model. LASSO regression is a process of compressing the regression coefficients by implementing a penalty function, imposing constraints on the absolute values of the regression coefficients to enhance the refinement of the model. Set the key parameter alpha to 1. In the Glmnet function, two lambda values were selected: min and lambda-1se. The former minimizes cross-validation errors to the greatest extent, while the latter provides a more concise model. Together, they help us balance the complexity of the model and the prediction accuracy. The LASSO variable screening is conducted on the training set. The regularization parameter λ is selected for 10-fold cross-validation in the training set, and the optimal λ value is selected by minimizing the cross-validation error. The screened variables were included in the multivariate logistic regression model to control the confounding effect among the variables, and the final variables were obtained based on the significance of the model to construct the regression model[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e].\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003e2.6 Machine Learning and Visualizing Variables\u003c/h2\u003e\u003cp\u003eWe used logistic regression modeling to predict the risk of OA in patients with MetS, defined as a binary classification problem for predicting the risk of developing the disease based on clinical characteristics. Also compare some common machine learning algorithms to construct the best prediction model[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eMultivariate Logistic regression: logistic regression is a statistical model used for classification problems, while multivariate logistic regression deals with logistic regression problems with multiple independent variables.XGBoost (Extreme Gradient Boosting), which is an efficient machine learning algorithm based on the gradient boosting framework. XGBoost (Extreme Gradient Boosting) is a highly efficient machine learning algorithm based on a gradient boosting framework, which performs well in various data mining and machine learning competitions. To prevent overfitting, we implemented L2 regularization (lambda\u0026thinsp;=\u0026thinsp;1) and set early stopping rounds\u0026thinsp;=\u0026thinsp;10 based on validation loss. LightGBM (Light Gradient Boosting Machine) is a fast and efficient gradient boosting framework developed by Microsoft, which has the advantages of fast training speed, low memory consumption, etc.CatBoost (Categorical Boosting) is a machine learning algorithm based on gradient boosting. is a machine learning algorithm based on gradient boosting, which is particularly good at handling data with a large number of categorical features.RF (Random Forest), which is an integrated learning algorithm that improves the accuracy and stability of a model by constructing multiple decision trees and combining their predictions.KNN (K Nearest Neighbors), which is an instance-based learning algorithm that finds the nearest neighbor of each sample in the training set by calculating the distance between the samples to be classified and the samples in the training set, and finds the nearest neighbor of each sample in the training set. KNN (K Nearest Neighbor Algorithm), it is an instance-based learning algorithm that finds the nearest K neighbors by calculating the distance between the samples to be classified and the samples in the training set, and predicts the classes of the samples to be classified based on the classes of those neighbors. NN (Neural Network), which is a computational model that mimics the structure and function of biological neural networks and consists of a large number of interconnected neurons that are able to automatically learn features and patterns from data, and is used for various machine learning tasks, such as classification, regression, image recognition, and speech recognition[\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eThe best performance of the model was evaluated by multiple indicators such as the area under the receiver operating characteristic (ROC) curve (AUC), Accuracy, Sensitivity, Specificity, Precision, and F1-Score, which reflects the true generalization ability of the model. shap (Shapley Additive Explanations) interpretability analysis is a technique used to explain the predictions generated by various ML models. This method is based on the output weights contributed by each feature to the model, allowing the behavior of the model to be interpreted at both global and local scales. This is achieved by developing an additive interpretation model that regards all features as contributors, thereby facilitating the calculation of the average incremental impact of each feature in all feasible feature combinations, and obtaining the SHAP value of each feature. It provides global and local interpretations, helping to understand the main influencing factors predicted by the model, as well as the prediction of individual sample factors. The variables in the best prediction model are ranked by importance and visualized using shape diagrams[\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e].\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003e2.7 Statistical analysis\u003c/h2\u003e\u003cp\u003eAll data analyses in this study were performed using the R language (version 4.4.2), and the study analysis process was performed using a number of R software analysis packages, which are shown separately in Supplementary Table\u0026nbsp;2. Initial descriptive analysis of the raw data set was performed, and data conforming to normal distribution were expressed in the form of (mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation) and analyzed using the independent samples t-test; data not normally distributed were expressed in the form of (median, quartiles) and compared using the Mann-Whitney U test. Frequencies and percentages were used to describe count data, and analyses were performed using the chi-square test or Fisher's exact probability method. After identifying the relevant variables the model was trained using logistic regression and the remaining seven machine learning algorithms on the training subset of data, and during the model training process the model parameters were optimized using a 5-fold cross-validation method to balance the occurrence of overfitting and underfitting. During the study when the statistical value P\u0026thinsp;\u0026lt;\u0026thinsp;0.05, it means that the difference is statistically significant.\u003c/p\u003e\u003c/div\u003e"},{"header":"3. Results ","content":"\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003e3.1 Baseline data of included participants\u003c/h2\u003e\u003cp\u003eAfter a series of exclusion screening processes, we included a total of 13,250 participants who could be diagnosed with MetS, of which 2,460 (18.57%) had an OA diagnosis. Firstly, we analyzed and found that 33 of the continuous type variables had a normal distribution, and the remaining 7 variables (CALLY, Alt, Ast, Gamma, CRP, HS_CRP, and SII) had a non-normally distributed, and a violin plot of the normality test is visible in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e of the Supplementary Material. Comparisons between the patient and non-patient groups of the included participants are shown in Table\u0026nbsp;2. i) Demographic and metabolic indicators: the age of the patient group was significantly higher than that of the normal group (64.74\u0026thinsp;\u0026plusmn;\u0026thinsp;12.03 vs. 55.67\u0026thinsp;\u0026plusmn;\u0026thinsp;15.44 years, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001), and obesity-related indicators (BMI, body fat percentage BFP, waist-to-height ratio WHtR) and markers of insulin resistance (TyG) were significantly higher (all P\u0026thinsp;\u0026lt;\u0026thinsp;0.001); laboratory biochemical parameters: albumin (40.83\u0026thinsp;\u0026plusmn;\u0026thinsp;3.36 vs. 41.30\u0026thinsp;\u0026plusmn;\u0026thinsp;3.34, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and globulin (29.92\u0026thinsp;\u0026plusmn;\u0026thinsp;4.79 vs. 30.93\u0026thinsp;\u0026plusmn;\u0026thinsp;4.83, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001) levels were significantly lower in the patient group, blood phosphorus (1.19\u0026thinsp;\u0026plusmn;\u0026thinsp;0.19 vs. 1.17\u0026thinsp;\u0026plusmn;\u0026thinsp;0.18), blood potassium (4.10\u0026thinsp;\u0026plusmn;\u0026thinsp;0.42 vs. 4.05\u0026thinsp;\u0026plusmn;\u0026thinsp;0.38) levels were significantly higher; in addition, dietary inflammation index (DII), glycosylated hemoglobin (HbA1c), and uric acid were the variables with no significant differences; ② Variables with significant differences in the non-normally distributed variables: C-reactive protein was higher in the patient group (CRP: 0.38 [0.17, 0.77] vs. 0.35 [ 0.16, 0.72], P\u0026thinsp;\u0026lt;\u0026thinsp;0.001), and higher systemic immune-inflammatory index (SII: 518.55 [352.75, 735.54] vs 481.49 [347.22, 684.36], P\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Albumin transaminase and glutamate transaminase levels were significantly lower in the patient group. Non-significantly different variables: cally index (P\u0026thinsp;=\u0026thinsp;0.086), high-sensitivity CRP (P\u0026thinsp;=\u0026thinsp;0.194). (iii) Categorical variables (10): demographic variables with significant differences included: a higher proportion of males in the patient group (66.67% vs. 51.74%, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001), comorbidities and behaviors: a higher prevalence of coronary heart disease in the patient group (CHD: 14.02% vs. 7.11%), a higher prevalence of COPD in the patient group (COPD: 10.77% vs. 5.35%), chronic kidney disease (COPD: 10.77% vs. 5.35%), and a higher prevalence of chronic lung disease (COPD: 5.35%) in the patient group. ), chronic kidney disease (CKD: 38.17% vs. 30.19%), and sleep disorders (55.33% vs. 40.41%) were more common in the patient group (both P\u0026thinsp;\u0026lt;\u0026thinsp;0.001); smoking (53.62% vs. 47.72%, P\u0026thinsp;\u0026lt;\u0026thinsp;0.001), and there was also a significant difference between the educational levels, osteoporosis status, and only the alcohol consumption status did not have a significant differences.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003e3.2 Predictor screening\u003c/h2\u003e\u003cp\u003eWe divided all MetS participants into the training set of 9275 cases and the test set of 3975 cases in a ratio of 7:3. The comparison between the two groups is shown in Supplementary Table\u0026nbsp;3, and the statistical analysis showed that the vast majority of the differences between the variables were not statistically significant. Firstly, the data in the training set were compared between groups, and the indicators whose differences were not statistically significant were removed (Supplementary Table\u0026nbsp;4), and the remaining variables were screened for predictors related to the outcome based on the LASSO regression method, and the screening process was demonstrated by the coefficient path diagram Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA and the cross-validation error diagram Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB, which were combined to efficiently screen the important in high-dimensional data by utilizing Lasso's regularization property. variables in high-dimensional data, and construct sparse regression models with excellent predictive performance, which are widely used in feature selection and dimensionality reduction analysis. After screening, it was shown that the outcome was closely related to 15 variables (Log Lambda \u0026asymp; -4.48). Finally, these variables were subjected to multifactorial logistic regression analysis, removing the variable with insignificant correlation with the outcome, \u0026ldquo;protein\u0026rdquo; (P\u0026thinsp;=\u0026thinsp;0.683), and the remaining 14 variables: Smoke, Education, Age, BFP, Bilirubin, Globulin, P, Cl, T_lum, TyG, Sex, CHD, COPD, Sleep Disorder (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e3\u003c/span\u003e and Forest plot Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e)\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003e3.3 Machine learning model performance\u003c/h2\u003e\u003cp\u003eWe trained machine learning models using 14 variables, the AUC values and 95% confidence intervals of other models were: LM\u0026thinsp;=\u0026thinsp;0.744 (0.732\u0026ndash;0.756), XGBoost\u0026thinsp;=\u0026thinsp;0.791 (0.780\u0026ndash;0.802), KNN\u0026thinsp;=\u0026thinsp;0.661 (0.647\u0026ndash;0.675), LightGBM\u0026thinsp;=\u0026thinsp;0.820 (0.809\u0026ndash;0.830), CatBoost\u0026thinsp;=\u0026thinsp;0.774 (0.763\u0026ndash;0.786), NB\u0026thinsp;=\u0026thinsp;0.751 (0.739\u0026ndash;0.763), NN\u0026thinsp;=\u0026thinsp;0.750 (0.738\u0026ndash;0.762) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eA). XGBoost showed the most balanced performance (Training AUC\u0026thinsp;=\u0026thinsp;0.791, Test AUC\u0026thinsp;=\u0026thinsp;0.761). Test set validation yielded AUC values: LM\u0026thinsp;=\u0026thinsp;0.761, RF\u0026thinsp;=\u0026thinsp;0.786, XGBoost\u0026thinsp;=\u0026thinsp;0.761, KNN\u0026thinsp;=\u0026thinsp;0.670, LightGBM\u0026thinsp;=\u0026thinsp;0.760, CatBoost\u0026thinsp;=\u0026thinsp;0.764, NB\u0026thinsp;=\u0026thinsp;0.747, NN\u0026thinsp;=\u0026thinsp;0.759 (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eB). PR curve analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eC) showed XGBoost had stronger performance in balancing precision and recall. DCA curve (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eD) indicated XGBoost had higher net benefit in some threshold intervals. Test set comparison of confusion matrix parameters (Accuracy, Sensitivity, Specificity, Precision, F1-Score; Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e4\u003c/span\u003e) confirmed XGBoost as the best-performing model in terms of stability and predictive performance.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003e3.5 SHAP-based model interpretability analysis\u003c/h2\u003e\u003cp\u003eIn this study, we evaluated the relative significance of various risk factors for patients with MetS who also suffer from OA, and finally interpreted the machine learning prediction model using SHAP analysis, where the Shap value represents the degree to which each feature contributes to the prediction of the model (positive and negative values are taken as the absolute value), and the importance of the features is ranked according to the average absolute value of the Shap value. Based on the XGBoost model algorithm, we plot the bar chart of variable importance Shap values (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eA), which is measured by the \u0026ldquo;Cover\u0026rdquo; metric (i.e., the proportion of samples covered by the feature in the node splits of the decision tree). The top three features in terms of importance are Age, BFP and Sleep Disorder, indicating that Age is the most critical predictor in the model, followed by BFP and Sleep Disorder. The remaining characteristics such as Globulin (Globulin) and Education (Education level) are of lesser importance. Figure\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eB demonstrates the overall direction and strength of the influence of each variable on the predicted outcomes of the model through the SHAP values. The horizontal axis shows the SHAP values (range \u0026minus;\u0026thinsp;1 to 3), with positive values indicating that the characteristics increase the risk of OA and negative values decreasing the risk. Characteristics were ranked in order of importance from top to bottom, Age (age) and BFP (body fat percentage) had the widest distribution of SHAP values and most of them were clustered in the positive region, indicating that advanced age and high body fat were associated with increased risk of OA. The SHAP values for Education were mostly distributed in the negative region, suggesting that high education level may be associated with a reduced risk. This figure comprehensively reflects the global influence pattern of the variables and provides a basis for subsequent individualized interpretation. Figure\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eC (Individual Sample SHAP Decomposition Plot) This figure demonstrates the contribution of each characteristic of a specific sample (BFP\u0026thinsp;=\u0026thinsp;59.2, Age\u0026thinsp;=\u0026thinsp;66, Sleep.Disorder\u0026thinsp;=\u0026thinsp;0, P\u0026thinsp;=\u0026thinsp;0.969, Education\u0026thinsp;=\u0026thinsp;3, etc.) to the prediction of OA. The baseline value (E[f(x)]\u0026thinsp;=\u0026thinsp;1.54) indicates the average predictive value of the model for all samples, while the predictive value for this individual (f(x)\u0026thinsp;=\u0026thinsp;1.09) is lower than the baseline value, indicating a lower risk of OA. The specific characteristic contributions were as follows: BFP\u0026thinsp;=\u0026thinsp;59.2 (SHAP = -0.431): a body fat percentage of 59.2 had a significant negative contribution to the predicted value; Age\u0026thinsp;=\u0026thinsp;66 (SHAP = -0.263): an age of 66 years had a significant negative contribution to the outcome; Sleep Disorder\u0026thinsp;=\u0026thinsp;0 (SHAP\u0026thinsp;=\u0026thinsp;+\u0026thinsp;0.244): the absence of sleep disorder had a positive contribution to the predicted value; P\u0026thinsp;=\u0026thinsp;0.244. Disorder\u0026thinsp;=\u0026thinsp;0 (SHAP\u0026thinsp;=\u0026thinsp;+\u0026thinsp;0.244): no sleep disorder contributes positively to the predicted value; P\u0026thinsp;=\u0026thinsp;0.969 (SHAP\u0026thinsp;=\u0026thinsp;+\u0026thinsp;0.157): this eigenvalue contributes positively to the predicted value; Education\u0026thinsp;=\u0026thinsp;3 (SHAP = -0.147): education level of 3 contributes negatively to the predicted value; TyG\u0026thinsp;=\u0026thinsp;8.75 (SHAP = -0.0859): a TyG value of 8.75 contributes negatively to the predicted value; Smoke\u0026thinsp;=\u0026thinsp;0 (SHAP\u0026thinsp;=\u0026thinsp;+\u0026thinsp;0.0726): non-smoking contributes positively to the predicted value; Sex\u0026thinsp;=\u0026thinsp;1 (SHAP = -0.0549): gender 1 (assumed to be male) contributes negatively to the predicted value; COPD\u0026thinsp;=\u0026thinsp;0 and the 5 other features contribute to the predicted value, which together affect the final prediction result. Figure\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eD (Individual Sample SHAP Aggregation Plot) This plot summarizes the feature contributions of the same sample in a more concise form. Key feature values (e.g. (P\u0026thinsp;=\u0026thinsp;0.969), (Sleep.Disorder\u0026thinsp;=\u0026thinsp;0), (BFP\u0026thinsp;=\u0026thinsp;59.2), (Age\u0026thinsp;=\u0026thinsp;66), (Education\u0026thinsp;=\u0026thinsp;3), etc.) are shown at the top, and the direction and intensity of SHAP values for each feature are visualized in the bar chart below. The difference ((-0.45)) between the predicted value (f(x)\u0026thinsp;=\u0026thinsp;1.09) and the baseline value (E[f(x)]\u0026thinsp;=\u0026thinsp;1.54) is mainly dominated by the negative contributions of (BFP\u0026thinsp;=\u0026thinsp;59.2) and (Age\u0026thinsp;=\u0026thinsp;66), etc., while the positive contributions of (P\u0026thinsp;=\u0026thinsp;0.969), (Sleep.Disorder\u0026thinsp;=\u0026thinsp;0), etc., partially affect the difference. Overall value: Using SHAP values to analyze XGBoost from global to individual in an all-round way, cracking the problem of interpretability of black-box models and meeting the demand for transparent decision-making in the medical field.5A and 5B identify important features and influence directions, and 5C and 5D explain the prediction logics of specific samples to enhance the credibility and practicability of the model, and provide data support for personalized interventions; optimizing the feature engineering to improve the model efficiency.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003e3.6 Constructing Nomograms for Risk Prediction\u003c/h2\u003e\u003cp\u003eNomogram, as a multivariate risk prediction tool, has the core advantage of transforming complex mathematical models into intuitive graphical interfaces to achieve individualized risk assessment and rapid clinical decision-making, which has more comprehensive value than traditional scoring systems or single biomarkers[\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e].To facilitate clinical application, we developed a nomogram (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e) based on the best model to visually predict the risk of OA in patients with MetS. the nomogram combines the top 8 of 14 key predictors identified in the study, including age, education, BFP, TyG, bilirubin, globulin, coronary artery disease, and sleep disorders. the nomogram is based on the best model for predicting the risk of OA in patients with MetS. the nomogram is based on the best model for predicting the risk of OA in patients with MetS.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\u003ch2\u003e3.7 Model validation and clinical utility\u003c/h2\u003e\u003cp\u003e Discriminant efficacy: validation set C-index was 0.815 (95% CI: 0.807\u0026ndash;0.823), significantly better than the traditional model (AUC\u0026thinsp;=\u0026thinsp;0.761), calibration: the Hosmer-Lemeshow test showed that the predicted probabilities were in high agreement with the observed probabilities (P\u0026thinsp;=\u0026thinsp;0.24), and the calibration curves fit the ideal diagonal. Clinical applications:\u003c/p\u003e\u003cp\u003e*Total Score and Risk Stratification Calculation: 0-350 points (total score after accumulating the scores of each variable), if the score is N, corresponding to the risk probability mapping: \u003cb\u003e0.2\u0026thinsp;+\u0026thinsp;0.0017 \u0026times; Total Score.\u003c/b\u003e\u003c/p\u003e\u003cp\u003e.\u003c/p\u003e\u003cp\u003eBased on the column-line diagram predicting risk we classified individuals into three categories: ①Low risk (\u0026lt;\u0026thinsp;33.3%, total score\u0026thinsp;\u0026lt;\u0026thinsp;78): joint protection education, as well as recommendations for healthy diet, weight control, and regular monitoring of metabolic indexes to prevent the progression of MetS. ②Medium risk (33.3%-66.6%, total score 78\u0026ndash;272 points): weight management\u0026thinsp;+\u0026thinsp;sleep intervention, strengthening exercise, strict control of blood glucose, blood lipids, smoking cessation and restriction of alcohol, comprehensive improvement of MetS. ③High risk (\u0026gt;\u0026thinsp;66.6%, total score\u0026thinsp;\u0026gt;\u0026thinsp;272 points): timely imaging assessment\u0026thinsp;+\u0026thinsp;metabolic regulation, seek medical treatment for drug and other interventions, strengthen lifestyle changes, strict management of MetS indicators to reduce the risk of OA.\u003c/p\u003e\u003cp\u003e\u003cb\u003eExample application\u003c/b\u003e:\u003c/p\u003e\u003cp\u003e\u003cb\u003ePatient A\u003c/b\u003e, CHD: 0 (no disease), score 0; Sleep.Disorder: 0 (no sleep disorder), score 0; Education: 1 (below high school), score 0; TyG (Triglyceride Glucose Index)\u0026thinsp;=\u0026thinsp;9, corresponding to a score of 13; Bilirubin (Total Bilirubin)\u0026thinsp;=\u0026thinsp;50, corresponding to a score of 20; BFP (Body Fat BFP (Body Fat Percentage)\u0026thinsp;=\u0026thinsp;25%, corresponding to a score of 20; Age\u0026thinsp;=\u0026thinsp;36 years, corresponding to a score of 20; Globulin\u0026thinsp;=\u0026thinsp;80, corresponding to a score of 0; Total score: (0\u0026thinsp;+\u0026thinsp;0\u0026thinsp;+\u0026thinsp;0\u0026thinsp;+\u0026thinsp;0\u0026thinsp;+\u0026thinsp;13\u0026thinsp;+\u0026thinsp;20\u0026thinsp;+\u0026thinsp;20\u0026thinsp;+\u0026thinsp;20\u0026thinsp;+\u0026thinsp;20\u0026thinsp;+\u0026thinsp;0\u0026thinsp;=\u0026thinsp;73).\u003c/p\u003e\u003cp\u003eAccording to the column chart, a total score of 73 corresponds to a Diagnostic Possibility\u0026thinsp;\u0026asymp;\u0026thinsp;0.33) (low risk).\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eInterventions\u003c/strong\u003e\u003cp\u003eEducate on joint protection and avoid strenuous exercise and joint injuries. Control body fat, maintain a healthy weight, eat a balanced diet and reduce high calorie intake. Regularly monitor metabolic indicators (e.g., TyG), ensure adequate sleep, and maintain good lifestyle habits.\u003c/p\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003ePatient B\u003c/b\u003e, CHD: 1 (diseased), score 11, Sleep Disorder: 1 (presence of sleep disorder), score 14.5, Education: 2 (high school), score 9, TyG\u0026thinsp;=\u0026thinsp;11, corresponding to a score of 7.5; Bilirubin\u0026thinsp;=\u0026thinsp;10, corresponding to a score of 46, BFP\u0026thinsp;=\u0026thinsp;60%, score 67.5, Age\u0026thinsp;=\u0026thinsp;82, score 80, Globulin\u0026thinsp;=\u0026thinsp;40, score 50, total score: (11\u0026thinsp;+\u0026thinsp;14.5\u0026thinsp;+\u0026thinsp;9\u0026thinsp;+\u0026thinsp;7.5\u0026thinsp;+\u0026thinsp;46\u0026thinsp;+\u0026thinsp;67.5\u0026thinsp;+\u0026thinsp;80\u0026thinsp;+\u0026thinsp;50\u0026thinsp;=\u0026thinsp;285.5).\u003c/p\u003e\u003cp\u003eAccording to the column line graph, the total score corresponds to Diagnostic Possibility\u0026thinsp;\u0026asymp;\u0026thinsp;0.69) (high risk).\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eInterventions\u003c/strong\u003e\u003cp\u003eImmediate imaging evaluation (e.g., X-ray, MRI) and thorough examination of joint structures. Strict regulation of metabolic parameters (glucose, lipids, blood pressure), combined with drug therapy if necessary. Intensive weight management and fat loss through low-calorie diet and joint-friendly exercise (e.g., swimming). Improve sleep disorders and seek guidance from sleep specialists; review joint and metabolic indexes every 1\u0026ndash;2 months to adjust the treatment plan.\u003c/p\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eTaking the high prevalence and socioeconomic burden of OA and MetS as the research entry point, this paper systematically explored previous studies on the pathological association between the two on the metabolic-inflammatory axis, which is mainly mediated by metabolic disorders of the body's endo-environment (hypertension, hyperglycemia, and lipid abnormalities) and accelerates the progression of OA through inflammation, oxidative stress, and other mechanisms. In this study, 14 core predictors were screened by a joint LASSO regression-logistic regression-XGBoost model to construct the top 8 strong correlates, among which age (SHAP = 0.296) and percentage of body fat (SHAP = 0.145) were used as traditional risk factors, and their predictive efficacies were further validated in this study. The strong association between age and OA risk may stem from multiple mechanisms: on the one hand, increasing age implies aging, and the degeneration of the organism is accompanied by a decrease in the repair capacity of articular chondrocytes, the accumulation of oxidative stress, and the deterioration of the inflammatory microenvironment, which promote the progression of aging in the joints. The main relevant biomarkers are SA - β - gal (an enzyme exclusive to senescent cells), SAHF (packaging proliferation-promoting genes to trigger senescence), p53 (participating in cell-cycle regulation and stress responses), p21 (promoting cellular senescence), and p53 (promoting cellular senescence). (packaging proliferation - promoting genes to trigger senescence), p53 (participating in cell - cycle regulation and stress responses), p21 (restricting cell - cycle progression post DNA damage), and pRb (suppressing genes to halt the cell cycle). These markers collectively govern cellular senescence via enzymatic, gene - packaging, transcriptional, and cell - cycle inhibitory pathways, regulating processes from gene expression to cell - cycle inhibition. These markers collectively govern cellular senescence via enzymatic, gene - packaging, transcriptional, and cell - cycle inhibitory pathways, regulating processes from gene expression to cell - cycle arrest in senescent chondrocytes[\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e].On the other hand, although advanced age is the greatest risk factor for OA, OA is not an inevitable consequence of aging. Imaging changes such as bone encumbrances are common in the elderly population, but the degree of joint pain is often inconsistent with imaging severity. Musculoskeletal aging increases susceptibility to OA, but its severity is more related to joint injury, obesity, genetics, and anatomical factors of joint mechanics. Mechanisms of joint tissue aging include cellular senescence-induced secretory phenotypes and late glycosylation end-product formation in the matrix, which affect tissue mechanical properties. An in-depth understanding of the mechanisms of joint senescence may provide new targets for slowing the progression of OA, which is of public health significance in high-risk populations. age-related metabolic disorders (e.g., increased insulin resistance) may further accelerate cartilage degeneration in patients with MetS, and, although age-related progression is irreversible, measures to slow chondrocyte senescence and intervene with sensitive senescence markers could positively affect the management of this disease[\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]. Previous studies have found that an imbalance of adipokines (leptin and lipocalin) induced by elevated body fat percentage regulates the pathologic process of OA. Leptin promotes synovial inflammation and abnormal remodeling of subchondral bone through the JAK/STAT and MAPK signaling pathways, while lipocalin inhibits inflammation and enhances cartilage matrix synthesis through activation of the PPAR-γ pathway. The study further showed that the association between body fat percentage and adipokines was particularly significant in knee OA, with leptin levels positively correlating with the degree of synovial inflammation and cartilage damage scores, and lipocalin levels negatively correlating with cartilage integrity[\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]. In addition, a Mendelian randomization study demonstrated that sleep disorders are independent risk factors for OA[\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e], sleep deprivation induces the secretion of pro-inflammatory cytokines (e.g., IL-6, IL-1β, and TNF-α) [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e], and excess IL-17 was found to accelerate the destruction of cartilage in a mouse model, and sleep disorders lead to abnormalities in melatonin production, which are likewise involved in the progression of OA [\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]. A potential pathway between sleep disorders and OA could be metabolic disorders mediating this correlation, which would exacerbate joint degeneration and provide a new direction for multidimensional interventions.\u003c/p\u003e\u003cp\u003eExperiments in rats have shown that globulin exhibits a unique predictive value as a recognition factor. Low globulin levels may reflect a chronic inflammatory state, which is associated with reduced complement system activity and impaired immune regulation, and supplementation of intra-articular α2macroglobulin may provide cartilage protection against arthritis[\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]. Hyaluronic acid and γ-globulin have been found in human joints to protect joints in patients with advanced OA, and the protective effect is reduced at lower levels [\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e]. Plasma corticosteroid-binding globulin (CBG) was found to have an important role in regulating glucocorticoid bioavailability, and plasma levels of CBG were significantly reduced in rats that developed severe inflammation in a rat-induced arthritis model, which is similar to our findings. The results suggest that plasma CBG levels and integrity are important biomarkers mediating the link between inflammation onset and joint inflammation, and thus further explorations regarding the role of globin in arthritis are warranted[\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eA study preliminarily indicated that higher education level exerts a protective effect against OA occurrence, though this effect is partially counteracted by BMI and smoking (with a 35% mediating effect). In patients with MetS, education level may indirectly elevate OA risk by exacerbating metabolic disorders (e.g., insulin resistance)[\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e]. Notably, highly educated occupational groups (e.g., researchers, medical professionals) exhibit a higher incidence of joint pain than the general population due to long-term psychological stress; occupational stress may indirectly increase OA risk by activating the HPA axis, elevating cortisol levels, and exacerbating inflammatory responses and insulin resistance. Additionally, differences in health awareness may lead to detection bias: highly educated populations participate more actively in physical examinations and imaging screenings, resulting in significantly higher detection rates of early OA cases, and this association may be further amplified by occupational sedentariness and dietary deviations[\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]. A prior study found a positive association between the TyG index and arthritis in U.S. adults under 60, but it was limited to normal-weight participants without diabetes [\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e]. The negative association between the TyG index and OA observed in the present study may stem from a \"reverse causation\" mechanism (pain-induced exercise limitation leading to metabolic suppression, i.e., passive reduction in glucose/triglycerides). Meanwhile, uncorrected metabolic heterogeneity (e.g., normal-weight metabolic abnormality subtypes) may cause misclassification bias, and chronic inflammation and energy metabolism disorders could exacerbate joint degeneration through non-insulin resistance pathways.\u003c/p\u003e\u003cp\u003eInterestingly, there may be a threshold effect between the TyG index and OA risk: very low levels might reflect defective cartilage repair caused by energy metabolism imbalance. Although higher education showed protective effects in SHAP analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eB), its positive association in regression (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e3\u003c/span\u003e) may reflect detection bias—educated individuals have better access to healthcare, increasing OA diagnosis rates. The inverse association between the TyG index and OA risk could also be attributed to reverse causality: joint pain limits physical activity, leading to improved glucose/lipid profiles in sedentary patients. A study developing bilirubin nanoparticle-targeted therapy for OA confirmed that bilirubin delays cartilage degeneration via antioxidant and anti-inflammatory mechanisms, suggesting that low serum total bilirubin may reduce the protective effect against OA[\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e]. A meta-analysis revealed a significant relationship between OA and the risk of developing CHD, with OA patients appearing to have higher cardiovascular risk. Thus, OA patients and physicians should recognize the importance of cardiovascular assessment in OA, and primary prevention and appropriate management of MetS may delay OA onset and slow its progression[\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eThis paper develops the first dynamic risk prediction tool for OA in MetS patients based on a database using a predictive modeling approach with machine learning to fill a gap in clinical needs. Key Findings Interpretation: Enhanced validation of traditional factors and mechanism exploration of new factors can be used to identify high-risk MetS patients through early screening, optimize the timing of joint protection interventions, reduce the burden of further medical care after the onset of disease in patients with different risks, improve disease predictability, and mitigate the benefits of improved health economics. Complex algorithms are transformed into actionable decision-making tools for clinicians through visualized column-line diagrams, which are especially suitable for primary care scenarios. In particular, early identification of high-risk patients (e.g., total score \u0026gt; 272) can trigger multidisciplinary joint interventions (rheumatology-endocrinology-rehabilitation) to potentially delay OA progression through a triple strategy of metabolic modulation, anti-inflammatory therapy and joint protection.\u003c/p\u003e\u003cp\u003eThis study not only reveals the metabolic-inflammatory regulatory network of OA risk in MetS patients, but also creates a paradigm for the application of machine learning in the prediction of degenerative joint diseases. In the future, by integrating multi-omics data (e.g. metabolomics, epigenetic markers), it is expected to realize the leap from 'risk prediction' to 'precision prevention'.\u003c/p\u003e\u003cp\u003e\u003cb\u003eLimitations\u003c/b\u003e: Despite the significant results of this study, the following limitations need to be addressed: (1) retrospective study: the cross-sectional nature of the NHANES data limits causal inferences, e.g., the negative association between education level and OA risk may be influenced by unmeasured confounders. (ii) Population representativeness: the NHANES sample is dominated by the US population and lacks diversity data from Asia, Africa and other regions, which may affect the applicability of the model to non-Western populations. (iii) Indicator limitations: some metabolic markers may also have key roles in OA injury. Therefore, in order to further promote clinical translation, future studies should focus on the following directions: ① Prospective validation: validate the generalization ability of the model through multicenter cohorts (e.g., UK Biobank), and explore the relationship between longitudinal changes in metabolic markers (e.g., TyG index) and OA progression. (ii) Mechanistic studies: validate the molecular mechanisms of novel factors such as globin and sleep disorders using organoid or animal models, e.g. observe chondrocyte inflammatory pathway activation by lowering globin gene. (iii) Tool optimization: develop mobile risk assessment apps and integrate real-time biosensor data (e.g., sleep monitoring bracelets) to enhance the convenience and dynamic monitoring capability of the tools.\u003c/p\u003e"},{"header":"Conclusion","content":"In this study, a dynamic prediction model of OA risk for MetS patients was constructed for the first time based on the NHANES database by integrating LASSO regression and XGBoost machine learning algorithm. The model demonstrated excellent discriminant efficacy in the validation set (AUC = 0.761) and identified eight core predictors, including age, body fat percentage (BFP), and sleep disorders, and its SHAP value revealed age (29.6% contribution) and body fat percentage (14.5%) as the strongest risk drivers. The column-line diagram tool enabled individualized risk stratification (low/medium/high risk), providing quantifiable clinical thresholds for intervention. The study confirms that disruption of the metabolic-inflammatory axis is a key mechanism for OA progression and validates the clinical value of novel predictors such as education level and globulin. Although limited by a retrospective design, this study provides the first evidence-based tool for a “metabolic-joint co-management” strategy, and the generalizability of the model needs to be further optimized through prospective cohort and mechanistic studies.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eOA：Osteoarthritis\u003c/p\u003e\n\u003cp\u003eMetS：Metabolic syndrome\u003c/p\u003e\n\u003cp\u003eCKD：Chronic Kidney Disease\u003c/p\u003e\n\u003cp\u003eOP：Osteoporosis\u003c/p\u003e\n\u003cp\u003eCHD：Coronary Heart Disease\u003c/p\u003e\n\u003cp\u003eCOPD ：Chronic Obstructive Pulmonary Disease\u003c/p\u003e\n\u003cp\u003eBMI：Body Mass Index\u003c/p\u003e\n\u003cp\u003eBRI：Body Roundness Index\u003c/p\u003e\n\u003cp\u003eBFP：Body Fat Percentage\u003c/p\u003e\n\u003cp\u003eABSI：A Body Shape Index\u003c/p\u003e\n\u003cp\u003eCALLY：CRP-albumin-lymphocyte index\u003c/p\u003e\n\u003cp\u003eDII：Diet Inflammatory Index\u003c/p\u003e\n\u003cp\u003eSII：Systemic Immune-Inflammation Index\u003c/p\u003e\n\u003cp\u003eTyG：Triglyceride-Glucose Index\u003c/p\u003e\n\u003cp\u003eWHR：Waist-Hip Ratio\u003c/p\u003e\n\u003cp\u003eWHtR：Waist-to-Height Ratio\u003c/p\u003e\n\u003cp\u003eLasso：Least Absolute Shrinkage and Selection Operator\u003c/p\u003e\n\u003cp\u003eLM：\u0026nbsp;Logistic Multivariate Regression\u003c/p\u003e\n\u003cp\u003eRF：Random Forest\u003c/p\u003e\n\u003cp\u003eXGBoost：Extreme Gradient Boosting\u003c/p\u003e\n\u003cp\u003eKNN：K-Nearest Neighbor\u003c/p\u003e\n\u003cp\u003eLightGBM：Light Gradient Boosting Machine\u003c/p\u003e\n\u003cp\u003eCatBoost：Categorical Boosting\u003c/p\u003e\n\u003cp\u003eNB：Naive Bayes\u003c/p\u003e\n\u003cp\u003eNN：Neural Network\u003c/p\u003e\n\u003cp\u003eShap：SHapley Additive exPlanations\u003c/p\u003e\n"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAuthors\u0026rsquo; contributions\u003c/strong\u003e\u003cstrong\u003e：\u003c/strong\u003eLintao Zhang and Xue Yun contributed equally as co-first authors. \u003cstrong\u003eLintao Zhang and Xue Yun:\u0026nbsp;\u003c/strong\u003econceptualization, methodology, formal analysis, investigation, writing original draft preparation;\u003cstrong\u003e\u0026nbsp;Shangyi Geng :\u0026nbsp;\u003c/strong\u003evalidation, visualization, resources; \u003cstrong\u003eJingge Wang:\u0026nbsp;\u003c/strong\u003emethodology, software; \u003cstrong\u003eZhaopeng Fan\u003c/strong\u003e\u003cstrong\u003e：\u003c/strong\u003eformal analysis； \u003cstrong\u003eHua Guo:\u003c/strong\u003e supervision, project administration, funding acquisition, writing review and editing; All authors agreed to submit the article to this journal, all involved made indispensable contributions to all aspects of this research work, and all agreed and approved the version submitted.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003cstrong\u003e：\u003c/strong\u003eThis work was supported by grants from:\u003c/p\u003e\n\u003cp\u003e- The Scientific Research Program of the Health and Wellness Commission in Shaanxi Province, China (XAYC220010)\u003c/p\u003e\n\u003cp\u003e- The Science and Technology Program in Xi\u0026apos;an, Shaanxi Province, China (22YXYJ0004)\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003cstrong\u003e：\u003c/strong\u003eThis study is not subject to ethical review and approval procedures because the data in the National Health and Nutrition Examination Survey (NHANES) database is publicly available and authorized by the National Center for Health Statistics (NCHS) of the United States. The research was conducted in accordance with the Declaration of Helsinki. All the data involved in the research have received ethical approval from the NCHS Research Ethics Review Committee, and all participants in the NHANES survey have provided informed consent.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eClinical trial number:\u003c/strong\u003e not applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003cstrong\u003e：\u003c/strong\u003eThe data used in this study are derived from the National Health and Nutrition Examination Survey (NHANES), a continuous program conducted by the U.S. Centers for Disease Control and Prevention (CDC) . Publicly available data from 1999 to 2023 are organized into 2-year cycles (e.g., 1999\u0026ndash;2000, 2001\u0026ndash;2002) and can be accessed through the CDC website after free registration . Data are available from the NHANES repository: (https://wwwn.cdc.gov/nchs/nhanes/). All analyses in this study were performed using publicly available data, and no restricted datasets were utilized.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003cstrong\u003e：\u003c/strong\u003eThe author declares that this research was conducted without any commercial or financial relationships and there are no potential conflicts of interest.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eTang, S., et al., \u003cem\u003eOsteoarthritis.\u003c/em\u003e Nat Rev Dis Primers, 2025. \u003cstrong\u003e11\u003c/strong\u003e(1): p. 10.\u003c/li\u003e\n\u003cli\u003e\u003cem\u003eGlobal, regional, and national burden of osteoarthritis, 1990-2020 and projections to 2050: a systematic analysis for the Global Burden of Disease Study 2021.\u003c/em\u003e Lancet Rheumatol, 2023. \u003cstrong\u003e5\u003c/strong\u003e(9): p. e508-e522.\u003c/li\u003e\n\u003cli\u003eNeeland, I.J., et al., \u003cem\u003eMetabolic syndrome.\u003c/em\u003e Nat Rev Dis Primers, 2024. \u003cstrong\u003e10\u003c/strong\u003e(1): p. 77.\u003c/li\u003e\n\u003cli\u003eSampath, S.J.P., et al., \u003cem\u003eObesity, Metabolic Syndrome, and Osteoarthritis-An Updated Review.\u003c/em\u003e Curr Obes Rep, 2023. \u003cstrong\u003e12\u003c/strong\u003e(3): p. 308-331.\u003c/li\u003e\n\u003cli\u003eAlberti, K.G., et al., \u003cem\u003eHarmonizing the metabolic syndrome: a joint interim statement of the International Diabetes Federation Task Force on Epidemiology and Prevention; National Heart, Lung, and Blood Institute; American Heart Association; World Heart Federation; International Atherosclerosis Society; and International Association for the Study of Obesity.\u003c/em\u003e Circulation, 2009. \u003cstrong\u003e120\u003c/strong\u003e(16): p. 1640-5.\u003c/li\u003e\n\u003cli\u003ePuenpatom, R.A. and T.W. Victor, \u003cem\u003eIncreased prevalence of metabolic syndrome in individuals with osteoarthritis: an analysis of NHANES III data.\u003c/em\u003e Postgrad Med, 2009. \u003cstrong\u003e121\u003c/strong\u003e(6): p. 9-20.\u003c/li\u003e\n\u003cli\u003eChing, K., et al., \u003cem\u003eHypertension meets osteoarthritis - revisiting the vascular aetiology hypothesis.\u003c/em\u003e Nat Rev Rheumatol, 2021. \u003cstrong\u003e17\u003c/strong\u003e(9): p. 533-549.\u003c/li\u003e\n\u003cli\u003eVeronese, N., et al., \u003cem\u003eType 2 diabetes mellitus and osteoarthritis.\u003c/em\u003e Semin Arthritis Rheum, 2019. \u003cstrong\u003e49\u003c/strong\u003e(1): p. 9-19.\u003c/li\u003e\n\u003cli\u003eJiang, M., et al., \u003cem\u003eLeptin Induced TLR4 Expression via the JAK2-STAT3 Pathway in Obesity-Related Osteoarthritis.\u003c/em\u003e Oxid Med Cell Longev, 2021. \u003cstrong\u003e2021\u003c/strong\u003e: p. 7385160.\u003c/li\u003e\n\u003cli\u003eLiu, S.Y., et al., \u003cem\u003eBidirectional association between metabolic syndrome and osteoarthritis: a meta-analysis of observational studies.\u003c/em\u003e Diabetol Metab Syndr, 2020. \u003cstrong\u003e12\u003c/strong\u003e: p. 38.\u003c/li\u003e\n\u003cli\u003eZhang, S., et al., \u003cem\u003eMetabolic syndrome increases osteoarthritis risk: findings from the UK Biobank prospective cohort study.\u003c/em\u003e BMC Public Health, 2024. \u003cstrong\u003e24\u003c/strong\u003e(1): p. 233.\u003c/li\u003e\n\u003cli\u003eLynskey, S.J., et al., \u003cem\u003eBiomarkers of Osteoarthritis-A Narrative Review on Causal Links with Metabolic Syndrome.\u003c/em\u003e Life (Basel), 2023. \u003cstrong\u003e13\u003c/strong\u003e(3).\u003c/li\u003e\n\u003cli\u003eWang, H., et al., \u003cem\u003eMetabolic Syndrome Increases the Risk for Knee Osteoarthritis: A Meta-Analysis.\u003c/em\u003e Evid Based Complement Alternat Med, 2016. \u003cstrong\u003e2016\u003c/strong\u003e: p. 7242478.\u003c/li\u003e\n\u003cli\u003eSharma, L., \u003cem\u003eOsteoarthritis of the Knee.\u003c/em\u003e N Engl J Med, 2021. \u003cstrong\u003e384\u003c/strong\u003e(1): p. 51-59.\u003c/li\u003e\n\u003cli\u003eGibbs, A.J., et al., \u003cem\u003eRecommendations for the management of hip and knee osteoarthritis: A systematic review of clinical practice guidelines.\u003c/em\u003e Osteoarthritis Cartilage, 2023. \u003cstrong\u003e31\u003c/strong\u003e(10): p. 1280-1292.\u003c/li\u003e\n\u003cli\u003eMahmoudian, A., et al., \u003cem\u003eEarly-stage symptomatic osteoarthritis of the knee - time for action.\u003c/em\u003e Nat Rev Rheumatol, 2021. \u003cstrong\u003e17\u003c/strong\u003e(10): p. 621-632.\u003c/li\u003e\n\u003cli\u003eHo, J., et al., \u003cem\u003eMendelian Randomization Studies of Lifestyle-Related Risk Factors for Osteoarthritis: A PRISMA Review and Meta-Analysis.\u003c/em\u003e Int J Mol Sci, 2022. \u003cstrong\u003e23\u003c/strong\u003e(19).\u003c/li\u003e\n\u003cli\u003eHe, Z.J., et al., \u003cem\u003ePain-Related Risk Factors Among Radiologic Stages of Knee Osteoarthritis: Data From the Osteoarthritis Initiative.\u003c/em\u003e Arthritis Care Res (Hoboken), 2023. \u003cstrong\u003e75\u003c/strong\u003e(6): p. 1333-1339.\u003c/li\u003e\n\u003cli\u003eAlberti, K.G., P. Zimmet, and J. Shaw, \u003cem\u003eThe metabolic syndrome--a new worldwide definition.\u003c/em\u003e Lancet, 2005. \u003cstrong\u003e366\u003c/strong\u003e(9491): p. 1059-62.\u003c/li\u003e\n\u003cli\u003eZhu, S., et al., \u003cem\u003eAssociation of smoking and osteoarthritis in US (NHANES 1999-2018).\u003c/em\u003e Sci Rep, 2023. \u003cstrong\u003e13\u003c/strong\u003e(1): p. 3911.\u003c/li\u003e\n\u003cli\u003eWang, J., et al., \u003cem\u003eVitamin D Status and Risk of All-Cause and Cause-Specific Mortality in Osteoarthritis Patients: Results from NHANES III and NHANES 2001-2018.\u003c/em\u003e Nutrients, 2022. \u003cstrong\u003e14\u003c/strong\u003e(21).\u003c/li\u003e\n\u003cli\u003eXi, L.J., et al., \u003cem\u003e[Application of LASSO and its extended method in variable selection of regression analysis].\u003c/em\u003e Zhonghua Yu Fang Yi Xue Za Zhi, 2023. \u003cstrong\u003e57\u003c/strong\u003e(1): p. 107-111.\u003c/li\u003e\n\u003cli\u003eLu, T., et al., \u003cem\u003eComparison of Machine Learning and Logic Regression Algorithms for Predicting Lymph Node Metastasis in Patients with Gastric Cancer: A two-Center Study.\u003c/em\u003e Technol Cancer Res Treat, 2024. \u003cstrong\u003e23\u003c/strong\u003e: p. 15330338231222331.\u003c/li\u003e\n\u003cli\u003eUddin, S., et al., \u003cem\u003eComparing different supervised machine learning algorithms for disease prediction.\u003c/em\u003e BMC Med Inform Decis Mak, 2019. \u003cstrong\u003e19\u003c/strong\u003e(1): p. 281.\u003c/li\u003e\n\u003cli\u003eSpicer, J. and A.N. Sanborn, \u003cem\u003eWhat does the mind learn? A comparison of human and machine learning representations.\u003c/em\u003e Curr Opin Neurobiol, 2019. \u003cstrong\u003e55\u003c/strong\u003e: p. 97-102.\u003c/li\u003e\n\u003cli\u003ePonce-Bobadilla, A.V., et al., \u003cem\u003ePractical guide to SHAP analysis: Explaining supervised machine learning model predictions in drug development.\u003c/em\u003e Clin Transl Sci, 2024. \u003cstrong\u003e17\u003c/strong\u003e(11): p. e70056.\u003c/li\u003e\n\u003cli\u003eBalachandran, V.P., et al., \u003cem\u003eNomograms in oncology: more than meets the eye.\u003c/em\u003e Lancet Oncol, 2015. \u003cstrong\u003e16\u003c/strong\u003e(4): p. e173-80.\u003c/li\u003e\n\u003cli\u003eHuang, J., et al., \u003cem\u003eUnveiling the ageing-related genes in diagnosing osteoarthritis with metabolic syndrome by integrated bioinformatics analysis and machine learning.\u003c/em\u003e Artif Cells Nanomed Biotechnol, 2025. \u003cstrong\u003e53\u003c/strong\u003e(1): p. 57-68.\u003c/li\u003e\n\u003cli\u003eMusumeci, G., M.A. Szychlinska, and A. Mobasheri, \u003cem\u003eAge-related degeneration of articular cartilage in the pathogenesis of osteoarthritis: molecular markers of senescent chondrocytes.\u003c/em\u003e Histol Histopathol, 2015. \u003cstrong\u003e30\u003c/strong\u003e(1): p. 1-12.\u003c/li\u003e\n\u003cli\u003eShane Anderson, A. and R.F. Loeser, \u003cem\u003eWhy is osteoarthritis an age-related disease?\u003c/em\u003e Best Pract Res Clin Rheumatol, 2010. \u003cstrong\u003e24\u003c/strong\u003e(1): p. 15-26.\u003c/li\u003e\n\u003cli\u003eZhang, C., et al., \u003cem\u003eAdipokine Signaling Pathways in Osteoarthritis.\u003c/em\u003e Front Bioeng Biotechnol, 2022. \u003cstrong\u003e10\u003c/strong\u003e: p. 865370.\u003c/li\u003e\n\u003cli\u003eNi, J., et al., \u003cem\u003eEvidence for causal effects of sleep disturbances on risk for osteoarthritis: a univariable and multivariable Mendelian randomization study.\u003c/em\u003e Osteoarthritis Cartilage, 2022. \u003cstrong\u003e30\u003c/strong\u003e(3): p. 443-450.\u003c/li\u003e\n\u003cli\u003eIrwin, M.R., R. Olmstead, and J.E. Carroll, \u003cem\u003eSleep Disturbance, Sleep Duration, and Inflammation: A Systematic Review and Meta-Analysis of Cohort Studies and Experimental Sleep Deprivation.\u003c/em\u003e Biol Psychiatry, 2016. \u003cstrong\u003e80\u003c/strong\u003e(1): p. 40-52.\u003c/li\u003e\n\u003cli\u003eNa, H.S., et al., \u003cem\u003eInterleukin-1-Interleukin-17 Signaling Axis Induces Cartilage Destruction and Promotes Experimental Osteoarthritis.\u003c/em\u003e Front Immunol, 2020. \u003cstrong\u003e11\u003c/strong\u003e: p. 730.\u003c/li\u003e\n\u003cli\u003eWang, S., et al., \u003cem\u003eIdentification of \u0026alpha;2-macroglobulin as a master inhibitor of cartilage-degrading factors that attenuates the progression of posttraumatic osteoarthritis.\u003c/em\u003e Arthritis Rheumatol, 2014. \u003cstrong\u003e66\u003c/strong\u003e(7): p. 1843-53.\u003c/li\u003e\n\u003cli\u003ePark, J.Y., et al., \u003cem\u003eEffects of hyaluronic acid and \u0026gamma;-globulin concentrations on the frictional response of human osteoarthritic articular cartilage.\u003c/em\u003e PLoS One, 2014. \u003cstrong\u003e9\u003c/strong\u003e(11): p. e112684.\u003c/li\u003e\n\u003cli\u003eHill, L.A., et al., \u003cem\u003eCorticosteroid-binding globulin is a biomarker of inflammation onset and severity in female rats.\u003c/em\u003e J Endocrinol, 2016. \u003cstrong\u003e230\u003c/strong\u003e(2): p. 215-25.\u003c/li\u003e\n\u003cli\u003eGill, D., et al., \u003cem\u003eCardiometabolic traits mediating the effect of education on osteoarthritis risk: a Mendelian randomization study.\u003c/em\u003e Osteoarthritis Cartilage, 2021. \u003cstrong\u003e29\u003c/strong\u003e(3): p. 365-371.\u003c/li\u003e\n\u003cli\u003eTseng, P.C., et al., \u003cem\u003eAdditive Interaction of Work-Related Stress and Sleep Duration on Arthritis Among Middle-Aged Civil Servants.\u003c/em\u003e Psychol Res Behav Manag, 2021. \u003cstrong\u003e14\u003c/strong\u003e: p. 2093-2101.\u003c/li\u003e\n\u003cli\u003eYan, Y., et al., \u003cem\u003eThe association between triglyceride glucose index and arthritis: a population-based study.\u003c/em\u003e Lipids Health Dis, 2023. \u003cstrong\u003e22\u003c/strong\u003e(1): p. 132.\u003c/li\u003e\n\u003cli\u003eHuang, H., et al., \u003cem\u003eOpsonization Inveigles Macrophages Engulfing Carrier-Free Bilirubin/JPH203 Nanoparticles to Suppress Inflammation for Osteoarthritis Therapy.\u003c/em\u003e Adv Sci (Weinh), 2024. \u003cstrong\u003e11\u003c/strong\u003e(22): p. e2400713.\u003c/li\u003e\n\u003cli\u003eMathieu, S., et al., \u003cem\u003eCardiovascular profile in osteoarthritis: a meta-analysis of cardiovascular events and risk factors.\u003c/em\u003e Joint Bone Spine, 2019. \u003cstrong\u003e86\u003c/strong\u003e(6): p. 679-684.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"549\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"3\" style=\"width: 83.7459%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTable 1. Diagnostic Criteria for Metabolic Syndrome (2009 Harmonized International Consensus)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd height=\"40\" style=\"width: 1.0276%;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 24.1476%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eComponent\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 48.4664%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDiagnostic Criteria\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.1319%;\"\u003e\n \u003cp\u003e\u003cstrong\u003eUnits\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd height=\"40\" style=\"width: 1.0276%;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"4\" style=\"width: 24.1476%;\"\u003e\n \u003cp\u003e1. Central Obesity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 48.4664%;\"\u003e\n \u003cp\u003eWaist circumference \u0026ge; specified thresholds by region/ethnicity:\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"4\" style=\"width: 11.1319%;\"\u003e\n \u003cp\u003ecm\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd height=\"40\" style=\"width: 1.0276%;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 48.4664%;\"\u003e\n \u003cp\u003e\u0026emsp;① Europe, South America, Middle East:\u0026nbsp;\u003cbr\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Men \u0026ge;94 cm;Women \u0026ge;80 cm\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd height=\"40\" style=\"width: 1.0276%;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 48.4664%;\"\u003e\n \u003cp\u003e\u0026emsp;②\u0026nbsp;South Asia, China, Japan:\u003cbr\u003e\u0026nbsp;Men \u0026ge;90 cm; Women \u0026ge;80 cm\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd height=\"40\" style=\"width: 1.0276%;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 48.4664%;\"\u003e\n \u003cp\u003e\u0026emsp;③\u0026nbsp;United States: Men \u0026ge;102 cm; Women \u0026ge;88 cm\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd height=\"40\" style=\"width: 1.0276%;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 24.1476%;\"\u003e\n \u003cp\u003e2. Elevated Triglycerides (TG)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 48.4664%;\"\u003e\n \u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026ge;150 mg/dL (1.7 mmol/L) or under specific\u0026nbsp;\u003cbr\u003e\u0026nbsp;treatment for hypertriglyceridemia\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.1319%;\"\u003e\n \u003cp\u003emg/dL (mmol/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd height=\"40\" style=\"width: 1.0276%;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" style=\"width: 24.1476%;\"\u003e\n \u003cp\u003e3. Reduced HDL Cholesterol\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" style=\"width: 48.4664%;\"\u003e\n \u003cp\u003eMen: \u0026lt;40 mg/dL (1.03 mmol/L); Women: \u0026lt;50 mg/dL (1.29 mmol/L) or under specific treatment for low HDL-C\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" style=\"width: 11.1319%;\"\u003e\n \u003cp\u003emg/dL (mmol/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd height=\"40\" style=\"width: 1.0276%;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd height=\"40\" style=\"width: 1.0276%;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 24.1476%;\"\u003e\n \u003cp\u003e4. Abnormal Blood Pressure\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 48.4664%;\"\u003e\n \u003cp\u003eSystolic \u0026ge;130 mmHg or diastolic \u0026ge;85 mmHg or diagnosed with hypertension and receiving treatment\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.1319%;\"\u003e\n \u003cp\u003emmHg\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd height=\"40\" style=\"width: 1.0276%;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 24.1476%;\"\u003e\n \u003cp\u003e5. Abnormal Fasting Glucose\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 48.4664%;\"\u003e\n \u003cp\u003e\u0026ge;100 mg/dL (5.6 mmol/L) or diagnosed with\u0026nbsp;\u003cbr\u003e\u0026nbsp;type 2 diabetes mellitus\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 11.1319%;\"\u003e\n \u003cp\u003emg/dL (mmol/L)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd height=\"40\" style=\"width: 1.0276%;\"\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"549\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"5\" style=\"width: 549px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eTable 2: Comparison of differences in research variables between all included patients and normal individuals (0= non-patients, 1=OA patients).\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eVariables\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTotal (n = 13250)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e0 (n = 10790)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e1 (n = 2460)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003eP\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eABSI, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e0.08 \u0026plusmn; 0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e0.08 \u0026plusmn; 0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e0.08 \u0026plusmn; 0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eAge, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e57.35 \u0026plusmn; 15.28\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e55.67 \u0026plusmn; 15.44\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e64.74 \u0026plusmn; 12.03\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eBMI, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e33.23 \u0026plusmn; 6.89\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e33.11 \u0026plusmn; 6.78\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e33.75 \u0026plusmn; 7.32\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eBFP, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e41.12 \u0026plusmn; 8.23\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e40.60 \u0026plusmn; 8.21\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e43.40 \u0026plusmn; 7.90\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eBRI, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e7.14 \u0026plusmn; 2.27\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e7.04 \u0026plusmn; 2.23\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e7.55 \u0026plusmn; 2.39\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003ePoverty, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e2.45 \u0026plusmn; 1.57\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e2.43 \u0026plusmn; 1.57\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e2.52 \u0026plusmn; 1.54\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e0.009\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eDII, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e1.75 \u0026plusmn; 1.77\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e1.74 \u0026plusmn; 1.76\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e1.79 \u0026plusmn; 1.80\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e0.181\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eHbA1c, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e6.38 \u0026plusmn; 1.47\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e6.39 \u0026plusmn; 1.52\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e6.33 \u0026plusmn; 1.24\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e0.053\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eBilirubin, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e10.32 \u0026plusmn; 4.94\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e10.45 \u0026plusmn; 4.98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e9.77 \u0026plusmn; 4.71\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eAlkaline, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e78.07 \u0026plusmn; 28.78\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e77.96 \u0026plusmn; 28.63\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e78.55 \u0026plusmn; 29.43\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e0.365\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eProtein, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e71.96 \u0026plusmn; 4.98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e72.23 \u0026plusmn; 4.96\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e70.75 \u0026plusmn; 4.89\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eAlbumin, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e41.22 \u0026plusmn; 3.35\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e41.30 \u0026plusmn; 3.34\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e40.83 \u0026plusmn; 3.36\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eGlobulin, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e30.75 \u0026plusmn; 4.84\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e30.93 \u0026plusmn; 4.83\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e29.92 \u0026plusmn; 4.79\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eCr, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e83.95 \u0026plusmn; 48.51\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e83.28 \u0026plusmn; 45.50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e86.90 \u0026plusmn; 59.90\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eUric acid, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e352.02 \u0026plusmn; 91.18\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e352.29 \u0026plusmn; 91.08\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e350.86 \u0026plusmn; 91.64\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e0.483\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eNa, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e139.44 \u0026plusmn; 2.64\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e139.41 \u0026plusmn; 2.58\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e139.56 \u0026plusmn; 2.87\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e0.019\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eP, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e1.17 \u0026plusmn; 0.18\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e1.17 \u0026plusmn; 0.18\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e1.19 \u0026plusmn; 0.19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eCa, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e2.34 \u0026plusmn; 0.10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e2.34 \u0026plusmn; 0.10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e2.35 \u0026plusmn; 0.10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e0.002\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eK, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e4.06 \u0026plusmn; 0.39\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e4.05 \u0026plusmn; 0.38\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e4.10 \u0026plusmn; 0.42\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eFe, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e14.52 \u0026plusmn; 5.87\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e14.56 \u0026plusmn; 5.98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e14.36 \u0026plusmn; 5.37\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e0.115\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eCl, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e102.57 \u0026plusmn; 3.38\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e102.63 \u0026plusmn; 3.31\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e102.29 \u0026plusmn; 3.63\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eOsmolality, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e280.63 \u0026plusmn; 5.85\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e280.47 \u0026plusmn; 5.74\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e281.37 \u0026plusmn; 6.26\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eH2CO3, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e24.87 \u0026plusmn; 2.50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e24.81 \u0026plusmn; 2.48\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e25.13 \u0026plusmn; 2.56\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eTC, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e5.01 \u0026plusmn; 1.19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e5.02 \u0026plusmn; 1.18\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e4.96 \u0026plusmn; 1.24\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e0.036\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eLldl, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e2.86 \u0026plusmn; 1.06\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e2.87 \u0026plusmn; 1.06\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e2.83 \u0026plusmn; 1.08\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e0.132\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eT fem, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e-0.57 \u0026plusmn; 1.38\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e-0.52 \u0026plusmn; 1.38\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e-0.79 \u0026plusmn; 1.37\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eT lum, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e-0.71 \u0026plusmn; 1.76\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e-0.74 \u0026plusmn; 1.75\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e-0.57 \u0026plusmn; 1.80\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eSleep hours, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e6.90 \u0026plusmn; 1.51\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e6.89 \u0026plusmn; 1.50\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e6.93 \u0026plusmn; 1.54\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e0.175\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eTyG, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e9.18 \u0026plusmn; 0.69\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e9.20 \u0026plusmn; 0.69\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e9.11 \u0026plusmn; 0.66\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eWHR, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e0.97 \u0026plusmn; 0.07\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e0.97 \u0026plusmn; 0.07\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e0.96 \u0026plusmn; 0.08\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eWHtR, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e0.67 \u0026plusmn; 0.08\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e0.66 \u0026plusmn; 0.08\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e0.68 \u0026plusmn; 0.09\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eCDAI, Mean \u0026plusmn; SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e0.18 \u0026plusmn; 3.73\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e0.18 \u0026plusmn; 3.76\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e0.17 \u0026plusmn; 3.57\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e0.921\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eCALLY, M (Q₁, Q₃)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e299.13 (135.79, 665.94)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e305.45 (136.67, 668.81)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e284.76 (129.74, 639.17)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e0.086\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eAlt, M (Q₁, Q₃)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e22.00 (16.00, 31.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e22.00 (17.00, 32.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e20.00 (15.00, 27.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eAst, M (Q₁, Q₃)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e22.00 (18.00, 28.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e22.00 (18.00, 28.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e22.00 (18.00, 27.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eGamma, M (Q₁, Q₃)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e24.00 (17.00, 38.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e25.00 (18.00, 38.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e23.00 (17.00, 36.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eCRP, M (Q₁, Q₃)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e0.35 (0.16, 0.73)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e0.35 (0.16, 0.72)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e0.38 (0.17, 0.77)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eHS CRP, M (Q₁, Q₃)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e3.67 (1.70, 7.13)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e3.66 (1.70, 7.09)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e3.73 (1.75, 7.34)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e0.194\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eSII, M (Q₁, Q₃)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e486.58 (348.22, 695.25)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e481.49 (347.22, 684.36)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e518.55 (352.75, 735.54)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eSex, n(%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eMale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e6027 (45.49)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e5207 (48.26)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e820 (33.33)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eFemale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e7223 (54.51)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e5583 (51.74)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e1640 (66.67)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eAlcohol, n(%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e0.196\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eNO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e2228 (16.82)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e1836 (17.02)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e392 (15.93)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eYES\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e11022 (83.18)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e8954 (82.98)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e2068 (84.07)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eCHD, n(%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eNO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e12138 (91.61)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e10023 (92.89)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e2115 (85.98)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eYES\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e1112 (8.39)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e767 (7.11)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e345 (14.02)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eCOPD, n(%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eNO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e12408 (93.65)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e10213 (94.65)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e2195 (89.23)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eYES\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e842 (6.35)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e577 (5.35)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e265 (10.77)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eCKD, n(%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eNO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e9054 (68.33)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e7533 (69.81)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e1521 (61.83)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eYES\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e4196 (31.67)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e3257 (30.19)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e939 (38.17)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eSleep Disorder, n(%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eNO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e7529 (56.82)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e6430 (59.59)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e1099 (44.67)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eYES\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e5721 (43.18)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e4360 (40.41)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e1361 (55.33)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eSmoke, n(%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eNO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e6782 (51.18)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e5641 (52.28)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e1141 (46.38)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eYES\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e6468 (48.82)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e5149 (47.72)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e1319 (53.62)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eEducation, n(%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003e\u0026lt;High\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e1764 (13.31)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e1533 (14.21)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e231 (9.39)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eHigh\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e5297 (39.98)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e4345 (40.27)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e952 (38.70)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003e\u0026gt;High\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e6189 (46.71)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e4912 (45.52)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e1277 (51.91)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eOP, n(%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026lt;.001\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eNO osteoporosis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e6890 (52.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e5677 (52.61)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e1213 (49.31)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eLow bone mass\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e3634 (27.43)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e2830 (26.23)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e804 (32.68)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 128px;\"\u003e\n \u003cp\u003eOsteoporosis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e2726 (20.57)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e2283 (21.16)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 126px;\"\u003e\n \u003cp\u003e443 (18.01)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 42px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003eTable 3: Results of multivariate logistic regression analysis excluding confounding.\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"387\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 142px;\"\u003e\n \u003cp\u003echaracteristics\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 57px;\"\u003e\n \u003cp\u003eOR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003eOR（95%CI）\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003eP\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 142px;\"\u003e\n \u003cp\u003eAge\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 57px;\"\u003e\n \u003cp\u003e1.051\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e1.047-1.055\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e＜\u003c/strong\u003e\u003cstrong\u003e0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 142px;\"\u003e\n \u003cp\u003eBFP\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 57px;\"\u003e\n \u003cp\u003e1.038\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e1.028-1.049\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e＜\u003c/strong\u003e\u003cstrong\u003e0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 142px;\"\u003e\n \u003cp\u003eBilirubin\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 57px;\"\u003e\n \u003cp\u003e0.979\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e0.965-0.992\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 142px;\"\u003e\n \u003cp\u003eCHD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 57px;\"\u003e\n \u003cp\u003e1.442\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e1.209-1.72\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e＜\u003c/strong\u003e\u003cstrong\u003e0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 142px;\"\u003e\n \u003cp\u003eCl\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 57px;\"\u003e\n \u003cp\u003e0.973\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e0.956-0.99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 142px;\"\u003e\n \u003cp\u003eCOPD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 57px;\"\u003e\n \u003cp\u003e1.435\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e1.18-1.746\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e＜\u003c/strong\u003e\u003cstrong\u003e0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 142px;\"\u003e\n \u003cp\u003eEducation\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 57px;\"\u003e\n \u003cp\u003e1.29\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e1.185-1.403\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e＜\u003c/strong\u003e\u003cstrong\u003e0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 142px;\"\u003e\n \u003cp\u003eGlobulin\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 57px;\"\u003e\n \u003cp\u003e0.955\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e0.936-0.974\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e＜\u003c/strong\u003e\u003cstrong\u003e0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 142px;\"\u003e\n \u003cp\u003eP\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 57px;\"\u003e\n \u003cp\u003e1.556\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e1.146-2.113\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.005\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 142px;\"\u003e\n \u003cp\u003eProtein\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 57px;\"\u003e\n \u003cp\u003e0.996\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e0.977-1.016\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003e0.683\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 142px;\"\u003e\n \u003cp\u003eSex\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 57px;\"\u003e\n \u003cp\u003e1.474\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e1.24-1.751\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e＜\u003c/strong\u003e\u003cstrong\u003e0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 142px;\"\u003e\n \u003cp\u003eSleep Disorder\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 57px;\"\u003e\n \u003cp\u003e1.821\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e1.625-2.04\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e＜\u003c/strong\u003e\u003cstrong\u003e0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 142px;\"\u003e\n \u003cp\u003eSmoke\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 57px;\"\u003e\n \u003cp\u003e1.227\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e1.091-1.38\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 142px;\"\u003e\n \u003cp\u003eT_lum\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 57px;\"\u003e\n \u003cp\u003e1.092\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e1.057-1.129\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e＜\u003c/strong\u003e\u003cstrong\u003e0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 142px;\"\u003e\n \u003cp\u003eTyG\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 57px;\"\u003e\n \u003cp\u003e0.859\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 113px;\"\u003e\n \u003cp\u003e0.786-0.938\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.001\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"623\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"8\" style=\"width: 623px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eTable 4: Comparison of Performance between machine Learning Models and Logistic Regression Models.\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 189px;\"\u003e\n \u003cp\u003eModel\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 84px;\"\u003e\n \u003cp\u003eAUC (Train)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 66px;\"\u003e\n \u003cp\u003eAUC (Test)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 55px;\"\u003e\n \u003cp\u003eAccuracy\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003eSensitivity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 61px;\"\u003e\n \u003cp\u003eSpecificity\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54px;\"\u003e\n \u003cp\u003ePrecision\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54px;\"\u003e\n \u003cp\u003eF1-Score\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 189px;\"\u003e\n \u003cp\u003eLogistic Multivariate regression, LM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 84px;\"\u003e\n \u003cp\u003e0.744\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 66px;\"\u003e\n \u003cp\u003e0.761\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 55px;\"\u003e\n \u003cp\u003e0.825\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e0.742\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 61px;\"\u003e\n \u003cp\u003e0.892\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54px;\"\u003e\n \u003cp\u003e0.538\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54px;\"\u003e\n \u003cp\u003e0.483\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 189px;\"\u003e\n \u003cp\u003eRandom Forest, RF\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 84px;\"\u003e\n \u003cp\u003e1.000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 66px;\"\u003e\n \u003cp\u003e0.786\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 55px;\"\u003e\n \u003cp\u003e0.814\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e0.871\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 61px;\"\u003e\n \u003cp\u003e0.908\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54px;\"\u003e\n \u003cp\u003e0.482\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54px;\"\u003e\n \u003cp\u003e0.431\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 189px;\"\u003e\n \u003cp\u003eXGBoost\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 84px;\"\u003e\n \u003cp\u003e0.791\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 66px;\"\u003e\n \u003cp\u003e0.761\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 55px;\"\u003e\n \u003cp\u003e0.802\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e0.787\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 61px;\"\u003e\n \u003cp\u003e0.806\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54px;\"\u003e\n \u003cp\u003e0.484\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54px;\"\u003e\n \u003cp\u003e0.599\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 189px;\"\u003e\n \u003cp\u003eK-Nearest Neighbor, KNN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 84px;\"\u003e\n \u003cp\u003e0.661\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 66px;\"\u003e\n \u003cp\u003e0.670\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 55px;\"\u003e\n \u003cp\u003e0.690\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e0.761\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 61px;\"\u003e\n \u003cp\u003e0.674\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54px;\"\u003e\n \u003cp\u003e0.351\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54px;\"\u003e\n \u003cp\u003e0.480\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 189px;\"\u003e\n \u003cp\u003eLightGBM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 84px;\"\u003e\n \u003cp\u003e0.820\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 66px;\"\u003e\n \u003cp\u003e0.760\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 55px;\"\u003e\n \u003cp\u003e0.722\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e0.773\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 61px;\"\u003e\n \u003cp\u003e0.710\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54px;\"\u003e\n \u003cp\u003e0.382\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54px;\"\u003e\n \u003cp\u003e0.511\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 189px;\"\u003e\n \u003cp\u003eCatBoost\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 84px;\"\u003e\n \u003cp\u003e0.774\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 66px;\"\u003e\n \u003cp\u003e0.764\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 55px;\"\u003e\n \u003cp\u003e0.656\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e0.775\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 61px;\"\u003e\n \u003cp\u003e0.628\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54px;\"\u003e\n \u003cp\u003e0.325\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54px;\"\u003e\n \u003cp\u003e0.458\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 189px;\"\u003e\n \u003cp\u003eNaive Bayes, NB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 84px;\"\u003e\n \u003cp\u003e0.751\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 66px;\"\u003e\n \u003cp\u003e0.747\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 55px;\"\u003e\n \u003cp\u003e0.664\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e0.709\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 61px;\"\u003e\n \u003cp\u003e0.654\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54px;\"\u003e\n \u003cp\u003e0.321\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54px;\"\u003e\n \u003cp\u003e0.442\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 189px;\"\u003e\n \u003cp\u003eNeural Network, NN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 84px;\"\u003e\n \u003cp\u003e0.750\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 66px;\"\u003e\n \u003cp\u003e0.759\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 55px;\"\u003e\n \u003cp\u003e0.662\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 60px;\"\u003e\n \u003cp\u003e0.722\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 61px;\"\u003e\n \u003cp\u003e0.648\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54px;\"\u003e\n \u003cp\u003e0.322\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 54px;\"\u003e\n \u003cp\u003e0.445\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"discover-medicine","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [Discover Medicine](https://link.springer.com/journal/44337)","snPcode":"44337","submissionUrl":"https://submission.springernature.com/new-submission/44337/3","title":"Discover Medicine","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Discover Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Metabolic syndrome, Osteoarthritis, NHANES, Predictive modeling, Machine learning","lastPublishedDoi":"10.21203/rs.3.rs-7261040/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7261040/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eObjective: The aim of this study was to develop a machine-learning-based predictive model for assessing osteoarthritis (OA) risk in patients with metabolic syndrome (MetS), to identify key predictors and develop a clinical risk assessment tool.\u003c/p\u003e\n\u003cp\u003eMethods: Data from the National Health and Nutrition Examination Survey (NHANES, 1999-2023) were utilized to screen the core predictors in combination with LASSO(Least Absolute Shrinkage and Selection Operator) regression, and predictive models were constructed by machine learning algorithms such as XGBoost. The SHAP framework was introduced to parse variable contributions, and a column-line diagram tool was developed to enable individualized risk assessment.\u003c/p\u003e\n\u003cp\u003eResults: The study included 13,250 patients with MetS and screened 14 core predictors including age, body fat percentage (BFP), and sleep disorders. The XGBoost model demonstrated the best predictive performance in the validation set (AUC=0.761), and the SHAP analysis showed that age (29.6% contribution) and BFP (14.5%) were the strongest risk drivers. Column line plots categorized risk into low, moderate, and high tertiles to guide targeted interventions.\u003c/p\u003e\n\u003cp\u003eConclusion: This study is the first to construct a dynamic prediction model of OA risk in patients with MetS, which highlights established metabolic factors contributing to OA risk and provides an evidence-based tool for the “metabolic-joint co-management” strategy, with significant potential for clinical translation.\u003c/p\u003e","manuscriptTitle":"A Machine Learning Framework for Osteoarthritis Risk Prediction in Metabolic Syndrome: NHANES-Based Model Development and Clinical Tool Validation","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-08-07 07:07:26","doi":"10.21203/rs.3.rs-7261040/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-08-18T12:23:24+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-08-18T03:21:33+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-08-16T02:56:30+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"304868378187871957296588128012124027528","date":"2025-08-13T09:01:34+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"197140012527821783711958308286812857332","date":"2025-08-12T15:47:08+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"235275844061574343142068178064482381933","date":"2025-08-07T14:41:38+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"121598854868928566613043393518451337151","date":"2025-08-05T01:20:37+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-08-04T19:23:29+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-08-01T15:01:25+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-08-01T09:06:09+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-08-01T09:05:32+00:00","index":"","fulltext":""},{"type":"submitted","content":"Discover Medicine","date":"2025-07-31T10:08:00+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"discover-medicine","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [Discover Medicine](https://link.springer.com/journal/44337)","snPcode":"44337","submissionUrl":"https://submission.springernature.com/new-submission/44337/3","title":"Discover Medicine","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Discover Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"6296c4db-08f6-49f3-abfb-bc0baccb7aec","owner":[],"postedDate":"August 7th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2025-10-31T07:38:23+00:00","versionOfRecord":[],"versionCreatedAt":"2025-08-07 07:07:26","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7261040","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7261040","identity":"rs-7261040","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}