Predicting Carotid Atherosclerosis in Latent Autoimmune Diabetes in Adult Patients Using Machine Learning Models: A Retrospective Study

Predicting Carotid Atherosclerosis in Latent Autoimmune Diabetes in Adult Patients Using Machine Learning Models: A Retrospective Study | 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 Predicting Carotid Atherosclerosis in Latent Autoimmune Diabetes in Adult Patients Using Machine Learning Models: A Retrospective Study Xiaoqin Chen, Zhitong Li, Xiaoying Fan, Yuanyuan Yan, Shiwei Liu This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6131962/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 03 Jul, 2025 Read the published version in BMC Cardiovascular Disorders → Version 1 posted 10 You are reading this latest preprint version Abstract Background: Latent autoimmune diabetes in adults (LADA) is a slowly progressing form of diabetes with autoimmune origins. Patients with LADA are at an elevated risk of developing cardiovascular diseases, including carotid atherosclerosis. While machine learning models have been widely used in predicting cardiovascular risks in Type 1 and Type 2 diabetes, research on LADA remains limited. Early prediction of carotid atherosclerosis using machine learning models could help in timely intervention and improved patient outcomes for this specific population. Methods: We conducted a retrospective cross-sectional analysis involving 142 LADA patients diagnosed within the endocrinology department at Shanxi Bethune Hospital, China. Various clinical, demographic, and laboratory variables were analyzed using univariate and multivariate logistic regression, complemented by LASSO regression for feature selection. Additionally, eight machine learning algorithms—logistic regression (LR), decision tree (DT), random forests (RF), k-nearest neighbors (KNN), support vector machine (SVM), neural networks (NNET), eXtreme gradient boosting (XGBoost), and light gradient boosting machine (LightGBM)—were employed to predict carotid atherosclerosis. Results: Significant risk factors for carotid atherosclerosis were identified, including age, smoking history, BMI, ALB, HDL-C, and ALT. Among the various machine learning models evaluated, the LR model exhibited the highest performance, achieving an area under the curve (AUC) of 0.936, alongside an accuracy of 86%. NNET and SVM models also demonstrated robust predictive capacities with AUC values of 0.919 and 0.918, respectively. Conclusions: This study highlights the critical role of identifying risk factors for carotid atherosclerosis in LADA patients. Our use of ML models builds on the growing body of work in diabetes-related cardiovascular risk prediction, and it offers a novel approach by specifically targeting the LADA population. Incorporating ML models into clinical practice could improve risk stratification and patient management in LADA. Future research should validate these models across diverse populations and investigate the underlying mechanisms linking LADA to cardiovascular risk. Carotid Atherosclerosis Machine Learning Latent Autoimmune Diabetes in Adults Figures Figure 1 Figure 2 Figure 3 Figure 4 1 Background Latent autoimmune diabetes in adults (LADA), also called “type 1.5 diabetes,” is a condition marked by autoimmune beta-cell destruction and insulin resistance ( 1 ). Epidemiological studies show that LADA accounts for 2–12% of patients initially diagnosed with type 2 diabetes (T2DM), and a survey found it represented 65% of new type 1 diabetes (T1DM) cases in China ( 2 ). Unlike T1DM, which typically appears in childhood, LADA develops in adulthood with a gradual onset, initially resembling T2DM and not requiring insulin therapy. Over time, however, beta-cell decline leads to insulin dependence. The presence of autoantibodies such as insulin autoantibodies (IAA), islet cell autoantibodies (ICA), and glutamic acid decarboxylase antibodies (GADA) distinguishes LADA from both T1DM and T2DM ( 3 , 4 ). LADA patients also have a higher risk of cardiovascular complications. Studies from Germany and Austria found significantly higher rates of hypertension (77.7%) and dyslipidemia (90.6%) in LADA patients compared to T2DM and T1DM ( 5 ). A Swedish study showed a 67% increased risk of cardiovascular disease in high autoimmune LADA patients ( 6 ). These findings highlight the need to focus on cardiovascular risk, particularly atherosclerosis, in LADA patients, an area often overlooked in favor of microvascular complications. Atherosclerosis is a major contributor to cardiovascular disease, with carotid atherosclerosis serving as a key marker for systemic vascular disease. Thickening and plaque formation in the carotid artery are strong predictors of stroke and coronary artery disease ( 7 ). Traditional risk factors such as hypertension, hyperlipidemia, and diabetes contribute to its development ( 8 ), while recent studies highlight the role of chronic inflammation, endothelial dysfunction, and genetic polymorphisms in accelerating the process ( 9 ). Inflammation markers like interleukin-6 are linked to the severity of carotid atherosclerosis, and in LADA patients, elevated biomarkers such as C-reactive protein and tumor necrosis factor-alpha correlate with increased cardiovascular risk ( 10 ). Insulin resistance-related markers like adiponectin and leptin are also abnormal in LADA, reinforcing the connection between metabolic dysfunction and atherosclerosis ( 11 ). Monitoring these biomarkers could help identify at-risk patients early and guide treatment. However, the silent progression of atherosclerosis limits the effectiveness of traditional prediction tools like the Framingham Risk Score in autoimmune diabetes populations ( 12 ). While carotid ultrasound is the gold standard for detecting plaques, its use as a universal screening tool is restricted by logistical and resource limitations, highlighting the need for alternative, precision-based prediction models. Machine learning models have transformed cardiovascular risk prediction by analyzing complex datasets and capturing nonlinear relationships often missed by traditional methods. The development of explainable frameworks like shapley additive explanations (SHAP) has enhanced clinical applicability by clarifying each feature’s contribution to predictions ( 13 ). While machine learning models have been successful in predicting macrovascular complications in T1DM and T2DM, their use in LADA-specific atherosclerosis remains unexplored, despite the unique pathophysiology of this group ( 14 , 15 ). This retrospective study aims to develop and validate machine learning models for predicting carotid atherosclerosis in LADA patients using demographic, clinical, and laboratory data. We compare eight algorithms—logistic regression (LR), decision tree (DT), random forests (RF), k-nearest neighbors (KNN), support vector machine (SVM), neural networks (NNET), eXtreme gradient boosting (XGBoost), and light gradient boosting machine (LightGBM)—to 1) identify novel predictors beyond traditional risk factors, 2) create an interpretable prediction framework for clinical use, and 3) offer pathophysiological insights through feature importance analysis. Our findings may redefine cardiovascular risk assessment in LADA and provide a personalized monitoring strategy for this high-risk, under-recognized group. 2 Methods Study population and design This retrospective cross-sectional study was conducted at Shanxi Bethune Hospital, Taiyuan, China, from April 2023 to December 2024, with approval from the hospital’s Ethics Committee (Approval No. YXLL-2025-026). Informed consent was waived as the study used preexisting data without intervention, in compliance with the Declaration of Helsinki. A total of 142 patients diagnosed with LADA were included from the endocrinology department, with diagnosis based on: ( 1 ) age ≥ 18 at diabetes onset, ( 2 ) presence of pancreatic islet autoantibodies or autoimmune T cells, and ( 3 ) no insulin dependence for at least six months. Participants were classified into two groups based on carotid ultrasound: those with and without carotid plaques. Exclusion criteria included: ( 1 ) severe cardiovascular disease, ( 2 ) other autoimmune diseases, ( 3 ) severe liver or kidney dysfunction, ( 4 ) active infections or malignancies, and ( 5 ) incomplete medical records. Data collection Clinical and demographic information were retrospectively obtained from electronic medical records, which included the participants’ age, gender, diabetes duration, smoking and drinking status, insulin usage, body mass index (BMI), as well as their systolic blood pressure (SBP) and diastolic blood pressure (DBP). Laboratory data encompassed measures such as fasting plasma glucose (FPG), hemoglobin A1c (HbA1c), fasting C-peptide (FCP), fasting insulin (FI), and lipid profiles (total cholesterol (TC), triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C)). Additionally, autoimmune markers (IAA, ICA, GADA), liver function tests (alanine aminotransferase (ALT), aspartate aminotransferase (AST)), renal parameters (blood urea nitrogen (BUN), serum creatinine (Scr), uric acid (UA)), albumin levels (ALB), white blood cell count (WBC), thyroid function (thyroid-stimulating hormone (TSH), free triiodothyronine (FT3), free thyroxine (FT4)), 25-hydroxyvitamin D concentrations (25(OH)D), triglyceride-glucose index (TyG), and atherosclerotic index of plasma (AIP) were also recorded. Carotid ultrasonography was conducted to assess the presence of atherosclerotic plaques in the carotid arteries. All patient data were anonymized to ensure confidentiality.The TyG index was calculated using the formula ( 16 ): \(\:TyG=In\left(\frac{TG(mg/dL)\times\:FPG(mg/dL)}{2}\right)\) The AIP index was calculated using the formula ( 17 ): \(\:AIP=log10\left(\frac{TG(mmol/L)}{HDL-C(mmol/L)}\right)\) Predictive variables The predictive variables were firstly selected based on univariate and multivariate logistic regression analyses. In the univariate analysis, associations between individual clinical variables and the presence of carotid plaques were evaluated. Significant predictors (P < 0.05) identified from the univariate analysis were then included in a multivariate logistic regression model to assess independent risk factors for carotid atherosclerosis. To further refine the selection of relevant variables and prevent overfitting, least absolute shrinkage and selection operator (LASSO) regression was performed to identify the most critical variables associated with carotid plaques. Specifically, LASSO regularization was performed on the dataset to select the most significant features from the initial set of 33 variables. The LASSO algorithm underwent 10-fold cross-validation to ensure the reliability and stability of the selected features, optimizing model performance while avoiding overfitting. LASSO is a regularization technique that penalizes the inclusion of irrelevant or redundant features by applying an L1 penalty. This process encourages sparsity in the model, driving some coefficients to zero and effectively excluding less important variables. By utilizing LASSO, we were able to retain only the most significant features associated with carotid atherosclerosis in patients with LADA, thereby reducing model complexity. This not only improved the interpretability of the model but also mitigated the risk of overfitting, leading to a more reliable and accurate predictive model. Finally, collinearity among the selected predictors was checked by calculating the variance inflation factor (VIF) to ensure the independence of the variables included in the model. Machine learning models Eight machine learning models were used to predict the presence of carotid atherosclerosis in LADA patients: LR, DT, RF, KNN, SVM, NNET, XGBoost, and LightGBM. To avoid data leakage, the training and testing datasets were independently normalized before modeling. The training set, which comprised 70% of the dataset, was used for model development, while the remaining 30% was reserved as a holdout test set for final model evaluation and adjustments. Hyperparameters for each model were optimized using a combination of grid search and cross-validation to achieve the best performance and ensure robust parameter tuning and model stability. The test set played a critical role in adjusting model parameters and estimating the model’s generalizability, ensuring that the models did not overfit and could perform well on unseen data. Model evaluation The performance of the machine learning models was comprehensively evaluated using metrics such as accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and F1 score, ensuring robustness and clinical utility. A 5-fold cross-validation was employed, with the area under the receiver operating characteristic (ROC) curve (AUC) serving as the primary performance evaluation metric, to identify the best estimator ( 18 ). Calibration was assessed with the Brier score, comparing predicted probabilities to actual outcomes, and calibration plots were generated for visual inspection. Decision curve analysis (DCA) was performed to evaluate the clinical utility by assessing the net benefit at different predicted risk thresholds, accounting for both true and false positives and the potential harm of unnecessary interventions. Machine interpretation The LR model can use a nomogram to visually represent the relationship between predictive variables and outcomes, helping clinicians estimate the likelihood of carotid atherosclerosis. The DT model can employ a decision tree structure to show feature relationships and predictions. For more complex models like RF, KNN, SVM, NNET, XGBoost, and LightGBM, the SHAP method interprets model predictions by attributing feature contributions based on Shapley values from game theory. SHAP provides a clear, quantitative measure of feature importance and how each predictor influences the outcome, offering valuable, interpretable insights for clinicians. Figure 1 illustrates the workflow of this study. Statistical analysis Statistical analysis was performed using SPSS (version 26.0) and R (version 4.4.2). Continuous variables were expressed as mean ± standard deviation or median (interquartile range), and categorical variables as frequencies and percentages. Group comparisons were made with the independent t-test or Mann-Whitney U test for continuous variables, and the chi-square test for categorical variables. Missing data (< 0.15%) were imputed using the Random Forest method. Outliers, identified by the Tukey method, were excluded to minimize bias. A p-value < 0.05 was considered statistically significant. 3 Results Data processing results We analyzed the clinical and biochemical characteristics of 142 LADA patients, splitting them into a training set (n = 100) and a testing set (n = 42) to develop and validate machine learning models for predicting carotid atherosclerosis. No significant differences were found between the two sets for most clinical variables (age, gender, diabetes duration, smoking and drinking status, insulin usage, and carotid atherosclerosis status) or biochemical markers, including BMI, blood pressure, HbA1c, FPG, lipid profiles, autoimmune markers, liver and kidney function, thyroid hormones, and vitamin D (P > 0.05 for all). These results demonstrate a well-balanced dataset for developing predictive models for carotid atherosclerosis in LADA patients (see Table S2). Feature selection We employed a multi-step feature selection process for predicting carotid atherosclerosis in LADA patients, combining regression analyses and LASSO ( 19 – 21 ). Univariate logistic regression first identified variables linked to carotid plaques (P < 0.05), and multivariate analysis confirmed age and smoking history as independent risk factors (Table S1 ). To prevent overfitting, LASSO regression was applied with normalized data, using carotid atherosclerosis as the dependent variable. LASSO utilized compressive coefficients and 10-fold cross-validation to establish the optimal penalty parameter, λ, with the value minimizing binomial deviance selected for a balanced model (Figures S1 A and S1B). From 33 variables, LASSO identified age, smoking history, BMI, ALB, HDL-C, and ALT as key predictors. The VIF for these variables were all below 2, indicating no significant multicollinearity. Model evaluation and comparison In this study, we evaluated eight machine learning models for predicting carotid atherosclerosis in LADA patients. The models tested include LR, DT, RF, KNN, SVM, NNET, XGBoost, and LightGBM. Table 1 presents the AUC, accuracy, sensitivity, specificity, PPV, NPV, F1 score, and Brier score for each model. Figure 2 A illustrates the AUC values for the training set, while Fig. 2 B shows the AUC values for the testing set. Table 1 Evaluation of the performance of the eight algorithms. Data set Algorithm AUC (95%CI) Accuracy Sensitivity Specificity PPV NPV F1 score Brier Score Train LR 0.956 (0.922–0.989) 0.85 0.81 0.88 0.83 0.86 0.82 Test LR 0.936 (0.866-1.000) 0.86 0.85 0.86 0.85 0.86 0.85 0.101 Train DT 0.843 (0.770–0.915) 0.84 0.86 0.82 0.79 0.89 0.82 Test DT 0.811 (0.691–0.932) 0.81 0.85 0.77 0.77 0.85 0.81 0.153 Train RF 0.999 (0.997-1) 0.97 0.95 0.98 0.98 0.97 0.96 Test RF 0.857 (0.746–0.967) 0.74 0.70 0.77 0.74 0.74 0.72 0.155 Train KNN 0.969 (0.943–0.995) 0.88 0.86 0.89 0.86 0.89 0.86 Test KNN 0.898 (0.805–0.990) 0.86 0.85 0.86 0.85 0.86 0.85 0.144 Train SVM 0.950 (0.913–0.986) 0.87 0.81 0.91 0.88 0.86 0.84 Test SVM 0.918 (0.833-1.000) 0.86 0.80 0.91 0.89 0.83 0.84 0.109 Train NNET 0.981 (0.961-1.000) 0.93 0.98 0.89 0.88 0.98 0.93 Test NNET 0.919 (0.834-1.000) 0.86 0.95 0.77 0.79 0.94 0.86 0.109 Train XGBoost 0.991 (0.979-1.000) 0.94 0.88 0.98 0.97 0.92 0.92 Test XGBoost 0.898 (0.800-0.995) 0.83 0.85 0.82 0.81 0.86 0.83 0.125 Train LightGBM 0.965 (0.934–0.995) 0.89 0.86 0.91 0.88 0.90 0.87 Test LightGBM 0.880 (0.773–0.986) 0.81 0.80 0.82 0.80 0.82 0.80 0.136 AUC: area under the curve; PPV: positive predictive value; NPV: negative predictive value; F1 score = (2 * Precision * Recall) / ( Precision + Recall). Among the evaluated models, LR, NNET, and SVM showed superior performance across multiple metrics. LR achieved an AUC of 0.936 (0.866 to 1), with accuracy, sensitivity, and specificity all at 86%, and an F1 score of 0.85, indicating good precision and recall. NNET had an AUC of 0.919 (0.834 to 1), with 86% accuracy, 95% sensitivity, and 77% specificity, along with a strong F1 score of 0.86. SVM performed similarly well with an AUC of 0.918 (0.833 to 1), 86% accuracy, 80% sensitivity, and 91% specificity, and an F1 score of 0.84. All three models showed low Brier scores, indicating good calibration. Other models like KNN, XGBoost, RF, and LightGBM had competitive performance but lower metrics. Random Forest showed the highest training AUC (0.999) but dropped in performance on the test set with an AUC of 0.857 and accuracy of 74%. XGBoost and LightGBM had moderate AUCs of 0.898 and 0.880, respectively, but their accuracy and sensitivity were not as high as LR, NNET, and SVM. The DT model, although simpler, provided insightful results in predicting carotid atherosclerosis in LADA patients. The final decision tree had a clear, binary classification rule based solely on a single feature: age, with a split criterion at 44.5 years of age. This model, due to its simplicity, had a lower AUC (0.811) compared to the other more complex models, but it still provided a valuable baseline for understanding the role of a single variable in predicting outcomes. Despite its lower overall performance, the DT model was useful for visualizing the relationship between age and the risk of carotid atherosclerosis, as demonstrated in the supplementary Figure S3. The simplicity of the DT also offers an advantage in interpretability, which may be beneficial in clinical practice, where transparency in decision-making is crucial. To compare the practical utility of the models, we performed calibration curve analysis to evaluate their performance in terms of calibration (see Fig. 2 C). The calibration curves for LR, NNET, and SVM showed excellent alignment with the ideal diagonal, indicating that these models were well-calibrated. The Brier scores for these models were among the lowest, further confirming their good calibration and ability to provide reliable probability estimates. This reinforces the conclusion that LR, NNET, and SVM are not only accurate in prediction but also offer well-calibrated results, which are crucial for making reliable clinical decisions. Additionally, we used DCA to assess the net benefit of the models across various threshold ranges (see Fig. 2 D). The DCA results revealed that LR, NNET, and SVM had the highest net benefit across most of the threshold ranges. This further supports the clinical value of these models, as they offer a higher overall benefit to patients, especially when the threshold for classification is optimized to balance between false positives and false negatives. Model interpretation For the LR model, we utilized nomograms to visually interpret the relationship between the model’s input features and its output. The nomogram (Fig. 3 ) illustrates how each of the six features—age, smoking history, BMI, ALB, HDL-C, and ALT—affects the likelihood of predicting carotid atherosclerosis. Each feature was assigned a score based on its corresponding coefficient in the model, and the total score was used to estimate the probability of atherosclerosis. This visualization allows clinicians to easily understand the relative importance of each variable, with higher scores corresponding to a greater likelihood of the disease. Notably, the nomogram showed that age was a particularly influential factor, with older age strongly associated with an increased risk of developing atherosclerosis. To address the challenge of explaining complex prediction models, we applied SHAP to quantify the contribution of each feature to the predictions. SHAP provides two types of explanations: global (overall model behavior) and local (individual prediction-level insights). For the NNET model, the SHAP summary bar plot (Fig. 4 A) displayed the mean SHAP values for each feature, ranked by their importance in the prediction. Age, ALB, and smoking history emerged as the most influential factors. The SHAP summary dot plot (Fig. 4 B) further illustrates the direction and strength of these influences, revealing that features such as advanced age, low ALB levels, and smoking significantly increased the risk of carotid atherosclerosis in LADA patients. The SHAP dependence plot (Fig. 4 C) provided more granular insights into how individual features directly impacted the model’s prediction. To facilitate understanding, we also presented a typical prediction example (Fig. 4 D), demonstrating the interpretability of the NNET model in predicting carotid atherosclerosis. For the SVM model, SHAP analysis identified age, smoking history, and ALB as the most critical features (see Figure S2). 4 Discussion This study aimed to develop and validate machine learning models for predicting carotid atherosclerosis in LADA patients, focusing on identifying novel predictors beyond traditional cardiovascular risk factors. Eight machine learning algorithms were tested, including LR, DT, RF, KNN, SVM, NNET, XGBoost, and LightGBM, to compare their performance for risk stratification. The LR, NNET, and SVM models demonstrated the best results, with LR achieving the highest AUC of 0.936 and excellent accuracy (86%), sensitivity (85%), and specificity (86%). NNET excelled in sensitivity (95%), while SVM showed strong specificity (91%). These models were well-calibrated, ensuring reliable predictions for clinical use. Key predictors identified for carotid atherosclerosis included age, smoking history, BMI, ALB, HDL-C, and ALT. Notably, ALB was identified as a new predictor, suggesting that low ALB levels may signal early vascular damage. Additionally, HDL-C and ALT further underscored the importance of lipid metabolism and liver function in the development of carotid atherosclerosis. These findings highlight the value of incorporating these markers into routine clinical assessments for better risk stratification and early intervention. Recent studies have increasingly applied machine learning models to predict cardiovascular risks and atherosclerosis in patients with T1DM and T2DM, leveraging diverse clinical and demographic data. In T2DM, machine learning algorithms, including support vector machines, random forests, and deep learning models, have demonstrated significant potential in predicting cardiovascular outcomes by analyzing risk factors such as blood glucose levels, lipid profiles, blood pressure, and patient history ( 22 , 23 ). Similarly, in T1DM, where patients are at a higher risk for atherosclerosis due to chronic hyperglycemia and autoimmune factors, machine learning approaches have been used to identify early signs of vascular damage, often incorporating biomarkers like C-reactive protein and advanced glycation end products ( 24 , 25 ). However, despite the robust application of machine learning in these populations, there is a paucity of research specifically targeting LADA, a subtype of diabetes with characteristics of both T1DM and T2DM. Our study aims to fill this gap by applying ML techniques to predict carotid atherosclerosis specifically in LADA patients, thus providing a unique contribution to the field and a comparison point to the existing literature on T1DM and T2DM. A key finding of our study is the significant role of age as the most prominent predictor of carotid atherosclerosis in LADA patients. Age is a well-established cardiovascular risk factor, and our results align with this, as aging increases the impact of chronic hyperglycemia, dyslipidemia, and hypertension on vascular damage ( 26 , 27 ). The autoimmune component and insulin resistance in LADA likely accelerate atherosclerosis, making age particularly relevant in this cohort. Our findings support previous studies linking age with cardiovascular disease progression, with age serving as a proxy for factors like diabetes duration and comorbidities such as hypertension. Older individuals often experience more severe metabolic syndrome and autoimmune dysfunction, reinforcing age’s clinical significance as a predictor in our models. The relationship between age and other clinical factors, such as insulin use, is complex. LADA’s gradual progression means that older patients often require insulin therapy, reflecting disease duration ( 28 , 29 ). Insulin use, which correlates with deteriorating glycemic control, further underscores age as a marker of disease progression and cardiovascular risk. This interplay between age, insulin use, and metabolic dysfunction contributes to the accelerated development of atherosclerosis in LADA patients ( 30 ). Our study identified several factors beyond age as significant predictors of carotid atherosclerosis in LADA, emphasizing the multifactorial nature of cardiovascular risk ( 31 ). Smoking, a known risk factor, accelerates endothelial dysfunction, oxidative stress, and inflammation, all contributing to plaque formation ( 32 ). Our study affirms smoking as a potent risk factor in LADA, highlighting the importance of smoking cessation in cardiovascular risk management. Similarly, obesity and high BMI are linked to insulin resistance, dyslipidemia, and chronic inflammation, which drive atherosclerosis ( 33 ). We confirm that a higher BMI predicts carotid atherosclerosis, reinforcing the need for weight management. Additionally, ALB, a marker of nutrition and inflammation, was inversely associated with atherosclerosis, suggesting that inflammation and poor nutritional status contribute to vascular damage in LADA patients ( 34 ). Low HDL-C levels, indicating poor metabolic control and lipid dysregulation, were also found to increase the risk of atherosclerosis ( 35 ). Finally, elevated ALT, a marker of hepatic metabolic dysfunction and NAFLD, was identified as a predictor of carotid atherosclerosis, linking liver health with cardiovascular risk in LADA patients ( 36 ). The predictive value of age, smoking, BMI, and other clinical variables in cardiovascular disease is well-documented in diabetes populations. However, this study is the first to apply machine learning models to LADA patients, a group underrepresented in predictive modeling research. Most studies have focused on T1DM or T2DM, leaving a gap in understanding the unique cardiovascular risks of LADA. Carotid atherosclerosis mechanisms differ between T1DM and T2DM: T1DM, typically diagnosed earlier, leads to early atherosclerosis due to prolonged hyperglycemia ( 37 ), while T2DM, affecting older adults with obesity, hypertension, and dyslipidemia, shows faster disease progression ( 38 ). T1DM patients exhibit higher carotid intima-media thickness and more plaque, but T2DM patients have more complex risk factors, resulting in higher cardiovascular event rates ( 39 ). Traditional carotid atherosclerosis prediction models rely on clinical factors such as age, gender, smoking, blood pressure, lipid levels, and diabetes presence, often using statistical methods like logistic regression or Cox regression. Tools like the Framingham and ASCVD Risk Scores integrate multiple clinical parameters to predict cardiovascular event risk ( 40 ). This study applies machine learning to identify cardiovascular risks in LADA patients, offering a novel approach to risk stratification. Carotid atherosclerosis, a key early indicator of cardiovascular disease, can be detected before symptoms appear. LADA patients, who face combined autoimmune and metabolic risks, benefit from early identification to prevent severe outcomes like stroke and heart attacks. Traditional risk assessments often rely on T2DM guidelines, which may overlook unique LADA factors ( 2 ). Our study shows that machine learning models can provide more accurate, individualized predictions, enabling earlier interventions such as lifestyle changes and pharmacotherapy. Additionally, liver function markers like ALT could be used in routine screening, complementing traditional risk factors. Machine learning can improve personalized medicine by incorporating diverse data to create comprehensive risk profiles, which can be integrated into electronic health records for real-time assessments. This could enhance preventive care and optimize patient outcomes, particularly in settings with limited access to advanced imaging techniques. It is also important to acknowledge the limitations of this study. First, while the sample size of 142 patients is adequate for exploratory analysis, it may limit the generalizability of the results. A larger, more diverse cohort would provide better validation of the predictive model and enhance its applicability across different ethnic groups. Additionally, the retrospective nature of the study may introduce biases in data collection, particularly concerning the completeness of clinical records or the underreporting of certain risk factors. Prospective studies with standardized data collection methods are necessary to further validate these findings. Furthermore, while this study incorporated a range of clinical and biochemical factors, the inclusion of additional variables, such as genetic markers, advanced imaging techniques, and more detailed lifestyle factors (e.g., dietary habits and physical activity levels), could enhance the accuracy of the models. Incorporating these factors would provide a more comprehensive approach to predicting atherosclerosis risk in LADA patients. Lastly, although the models showed strong performance in terms of AUC and calibration, they would require external validation in different clinical settings and populations to confirm their clinical applicability. Despite some limitations, this study demonstrates significant clinical potential for machine learning-based predictive models in identifying high-risk LADA patients. By detecting those at risk for carotid atherosclerosis, early interventions such as lipid-lowering therapies or immune-modulating treatments can be implemented to reduce cardiovascular events. These models can also help optimize healthcare resources by prioritizing high-risk patients for intensive screenings and treatments, while lower-risk patients require less intensive monitoring. Additionally, while traditional statistical methods provide valuable insights into the relationship between features and the outcome, machine learning models offer greater flexibility and the ability to capture more complex, non-linear interactions in the progress of feature selection. By combining both approaches, future studies can provide a more comprehensive understanding of cardiovascular risk and improve predictive accuracy, ultimately leading to more personalized and effective interventions. While this study emphasizes the technical aspects of machine learning, it is essential to also consider the ethical implications of applying these models in clinical practice. Ensuring that these models do not introduce bias, especially in diverse populations, is a critical concern. Bias can arise if training data lacks diversity, leading to inaccurate predictions for underrepresented groups. Additionally, clinicians may face situations where a model’s prediction contradicts their clinical judgment, and in such cases, the model should serve as a supplementary tool, rather than replacing expert decision-making. Ethical considerations, such as ensuring transparency, obtaining informed consent, and minimizing biases in predictions, must be addressed to foster trust and efficacy in these technologies. Future research should focus on the ethical, legal, and social aspects of implementing machine learning models in practice, as well as validating findings in larger, more diverse populations. Furthermore, including additional biomarkers could enhance model accuracy and help personalize cardiovascular risk assessments for LADA patients, thereby improving both the predictive power and fairness of the model in diverse clinical settings. . 5 Conclusions In conclusion, our study has successfully identified key demographic and clinical risk factors associated with carotid atherosclerosis in LADA patients while demonstrating the potential of machine learning models in enhancing predictive accuracy. The findings highlight the superior performance of models such as LR, NNET, and SVM, offering valuable tools for clinical decision-making. The integration of these predictive methodologies into routine practice could facilitate early intervention and improved management strategies for LADA patients at risk of cardiovascular complications. Future research should aim to validate these models in larger cohorts and explore their implementation in clinical settings to optimize patient outcomes. Abbreviations LADA Latent autoimmune diabetes in adults T2DM type 2 diabetes T1DM type 1 diabetes IAA insulin autoantibodies ICA islet cell autoantibodies GADA glutamic acid decarboxylase antibodies SHAP shapley additive explanations LR logistic regression DT decision tree RF random forests KNN k-nearest neighbors, SVM:support vector machine NNET neural networks XGBoost eXtreme gradient boosting LightGBM light gradient boosting machine BMI body mass index SBP systolic blood pressure DBP diastolic blood pressure FPG fasting plasma glucose HbA1c hemoglobin A1c FCP fasting C-peptide FI fasting insulin TC total cholesterol TG triglycerides HDL-C high-density lipoprotein cholesterol LDL-C low-density lipoprotein cholesterol ALT alanine aminotransferase AST aspartate aminotransferase BUN blood urea nitrogen Scr serum creatinine UA uric acid ALB albumin levels WBC white blood cell TSH thyroid-stimulating hormone FT3 free triiodothyronine FT4 free thyroxine 25(OH)D 25-hydroxyvitamin D TyG triglyceride-glucose index AIP atherosclerotic index of plasma LASSO least absolute shrinkage and selection operator VIF variance inflation factor PPV positive predictive value NPV negative predictive value ROC receiver operating characteristic AUC the area under the ROC curve DCA decision curve analysis Declarations Ethics approval and consent to participate Approval of the research protocol: The protocol for this research project has been approved by a suitably constituted Ethics Committee of the institution and it conforms to the provisions of the Declaration of Helsinki. Ethics Committee of Shanxi Bethune Hospital, Taiyuan, China, Approval No. YXLL-2025-026. Informed Consent: N/A Consent for publication Informed consent was waived for this study by the ethics committee/institutional review board, as it is a retrospective analysis of preexisting anonymized clinical data without direct patient intervention or use of identifying images. Funding This work was supported by Collaborative Traditional Chinese and Western Medicine for Chronic Disease Management Research Project (CXZH2024085); Shanxi Province Metabolic Disease (Type 1 Diabetes) Clinical Medical Research Center (20240410501001); Shanxi Province Science and Technology Achievements Transformation Guidance Special Fund (202304021301066); Shanxi Province Science and Technology Innovation Talent Team Special Plan (202204051002029); Shanxi Province Research Funding for Returned Overseas Scholars (2024 − 143); Shanxi Province Basic Research Program (202303021212330); Shanxi Province Key Laboratory of Endocrine and Metabolic Diseases (202404010920011). Author Contribution XC, XF, and YY contributed to data collection. XC and ZL contributed to data analysis. ZL and XC contributed to data interpretation. ZL, XC and SL collectively interpreted the results. ZL wrote the manuscript. XC and SL substantively revised it. All authors had access to the data, and reviewed and approved the final manuscript before submission. Acknowledgements Not applicable Clinical trial number Not applicable Data Availability The datasets analysed during the current study are available from the corresponding author on reasonable request. References Buzzetti R, Tuomi T, Mauricio D, Pietropaolo M, Zhou Z, Pozzilli P, et al. Management of Latent Autoimmune Diabetes in Adults: A Consensus Statement From an International Expert Panel. Diabetes. 2020;69(10):2037–47. Shuoming L, Zhiguang Z. Interpretation of Consensus of Chinese experts on the diagnosis and treatment of latent autoimmune diabetes in adults (2021 Edition). In: (Chinese), editor. Chinese Journal of Diabetes Mellitus2022. pp. 17–20. Liu B, Xiang Y, Liu Z, Zhou Z. Past, present and future of latent autoimmune diabetes in adults. Diabetes Metab Res Rev. 2020;36(1):e3205. Mishra R, Hodge KM, Cousminer DL, Leslie RD, Grant SFA. A Global Perspective of Latent Autoimmune Diabetes in Adults. Trends Endocrinol Metab. 2018;29(9):638–50. Golomb RC, Tittel SR, Welters A, Karges W, Meyhöfer S, Hummel M, et al. Increased cardiovascular risk in people with LADA in comparison to type 1 diabetes and type 2 diabetes: Findings from the DPV registry in Germany and Austria. Diabetes Obes Metab. 2025;27(2):563–73. Wei Y, Herzog K, Ahlqvist E, Andersson T, Nyström T, Zhan Y, et al. All-Cause Mortality and Cardiovascular and Microvascular Diseases in Latent Autoimmune Diabetes in Adults. Diabetes Care. 2023;46(10):1857–65. Sirimarco G, Amarenco P, Labreuche J, Touboul PJ, Alberts M, Goto S, et al. Carotid atherosclerosis and risk of subsequent coronary event in outpatients with atherothrombosis. Stroke. 2013;44(2):373–9. Lu SX, Wu TW, Chou CL, Cheng CF, Wang LY. Combined effects of hypertension, hyperlipidemia, and diabetes mellitus on the presence and severity of carotid atherosclerosis in community-dwelling elders: A community-based study. J Chin Med Assoc. 2023;86(2):220–6. Xie Y, Yu M, Qing T, Luo H, Shao M, Wei W, et al. Variants in genes related to inflammation and endothelial function can increase the risk for carotid atherosclerosis in southwestern China. Front Neurol. 2023;14:1174425. Almeida-Santiago C, Quevedo-Abeledo JC, Hernández-Hernández V, de Vera-González A, González-Delgado A, González-Gay M, et al. Circulating interleukin-6 and cardiovascular disease risk in patients with rheumatoid arthritis with low disease activity due to active therapy. Clin Exp Rheumatol. 2023;41(7):1537–43. Castelblanco E, Hernández M, Castelblanco A, Gratacòs M, Esquerda A, Molló À, et al. Low-grade Inflammatory Marker Profile May Help to Differentiate Patients With LADA, Classic Adult-Onset Type 1 Diabetes, and Type 2 Diabetes. Diabetes Care. 2018;41(4):862–8. Damen JA, Pajouheshnia R, Heus P, Moons KGM, Reitsma JB, Scholten R, et al. Performance of the Framingham risk models and pooled cohort equations for predicting 10-year risk of cardiovascular disease: a systematic review and meta-analysis. BMC Med. 2019;17(1):109. Ponce-Bobadilla AV, Schmitt V, Maier CS, Mensing S, Stodtmann S. Practical guide to SHAP analysis: Explaining supervised machine learning model predictions in drug development. Clin Transl Sci. 2024;17(11):e70056. Zhou Y, Liu C, Zhang Z, Chen J, Zhao D, Li L, et al. Identification and validation of diagnostic biomarkers of coronary artery disease progression in type 1 diabetes via integrated computational and bioinformatics strategies. Comput Biol Med. 2023;159:106940. Abegaz TM, Baljoon A, Kilanko O, Sherbeny F, Ali AA. Machine learning algorithms to predict major adverse cardiovascular events in patients with diabetes. Comput Biol Med. 2023;164:107289. Alizargar J, Bai CH, Hsieh NC, Wu SV. Use of the triglyceride-glucose index (TyG) in cardiovascular disease patients. Cardiovasc Diabetol. 2020;19(1):8. Yin B, Wu Z, Xia Y, Xiao S, Chen L, Li Y. Non-linear association of atherogenic index of plasma with insulin resistance and type 2 diabetes: a cross-sectional study. Cardiovasc Diabetol. 2023;22(1):157. Wang Z, Gu Y, Huang L, Liu S, Chen Q, Yang Y, et al. Construction of machine learning diagnostic models for cardiovascular pan-disease based on blood routine and biochemical detection data. Cardiovasc Diabetol. 2024;23(1):351. Ainiwaer A, Hou WQ, Kadier K, Rehemuding R, Liu PF, Maimaiti H, et al. A Machine Learning Framework for Diagnosing and Predicting the Severity of Coronary Artery Disease. Rev Cardiovasc Med. 2023;24(6):168. Barbieri MC, Grisci BI, Dorn M. Analysis and comparison of feature selection methods towards performance and stability. Expert Syst Appl. 2024;249:123667. Cai J, Luo J, Wang S, Yang S. Feature selection in machine learning: A new perspective. Neurocomputing. 2018;300:70–9. Dong W, Wan EYF, Fong DYT, Tan KC, Tsui WW, Hui EM, et al. Development and validation of 10-year risk prediction models of cardiovascular disease in Chinese type 2 diabetes mellitus patients in primary care using interpretable machine learning-based methods. Diabetes Obes Metab. 2024;26(9):3969–87. Huang Q, Zou X, Lian Z, Zhou X, Han X, Luo Y, et al. Predicting cardiovascular outcomes in Chinese patients with type 2 diabetes by combining risk factor trajectories and machine learning algorithm: a cohort study. Cardiovasc Diabetol. 2025;24(1):61. Zucchini S, Fabi M, Maltoni G, Zioutas M, Trevisani V, Di Natale V, et al. Adolescents with severe obesity show a higher cardiovascular (CV) risk than those with type 1 diabetes: a study with skin advanced glycation end products and intima media thickness evaluation. Acta Diabetol. 2020;57(11):1297–305. Słomiński B, Jankowiak M, Maciejewska A, Studziński M, Mączyńska A, Skrzypkowska M, et al. Genetic variation in C-reactive protein (CRP) gene is associated with retinopathy and hypertension in adolescents with type 1 diabetes. Cytokine. 2022;160:156025. López-Melgar B, Fernández-Friera L, Oliva B, García-Ruiz JM, Sánchez-Cabo F, Bueno H, et al. Short-Term Progression of Multiterritorial Subclinical Atherosclerosis. J Am Coll Cardiol. 2020;75(14):1617–27. Sarkar S, Prasanna VS, Das P, Suzuki H, Fujihara K, Kodama S, et al. The onset and the development of cardiometabolic aging: an insight into the underlying mechanisms. Front Pharmacol. 2024;15:1447890. Yohena S, Penas-Steinhardt A, Muller C, Faccinetti NI, Cerrone GE, Lovecchio S, et al. Immunological and clinical characteristics of latent autoimmune diabetes in the elderly. Diabetes Metab Res Rev. 2019;35(5):e3137. Jones AG, McDonald TJ, Shields BM, Hagopian W, Hattersley AT. Latent Autoimmune Diabetes of Adults (LADA) Is Likely to Represent a Mixed Population of Autoimmune (Type 1) and Nonautoimmune (Type 2) Diabetes. Diabetes Care. 2021;44(6):1243–51. Chaillous L, Bouhanick B, Kerlan V, Mathieu E, Lecomte P, Ducluzeau PH, et al. Clinical and metabolic characteristics of patients with latent autoimmune diabetes in adults (LADA): absence of rapid beta-cell loss in patients with tight metabolic control. Diabetes Metab. 2010;36(1):64–70. Wang Y, Li L, Li Y, Liu M, Gan G, Zhou Y, et al. The Impact of Dietary Diversity, Lifestyle, and Blood Lipids on Carotid Atherosclerosis: A Cross-Sectional Study. Nutrients. 2022;14(4):815. Higashi Y. Smoking cessation and vascular endothelial function. Hypertens Res. 2023;46(12):2670–8. Ferreira J, Cunha P, Carneiro A, Vila I, Cunha C, Silva C, et al. Is Obesity a Risk Factor for Carotid Atherosclerotic Disease?-Opportunistic Review. J Cardiovasc Dev Dis. 2022;9(5):162. Ishizaka N, Ishizaka Y, Nagai R, Toda E, Hashimoto H, Yamakado M. Association between serum albumin, carotid atherosclerosis, and metabolic syndrome in Japanese individuals. Atherosclerosis. 2007;193(2):373–9. Feig JE, Hewing B, Smith JD, Hazen SL, Fisher EA. High-density lipoprotein and atherosclerosis regression: evidence from preclinical and clinical studies. Circ Res. 2014;114(1):205–13. Zanetti D, Gustafsson S, Assimes TL, Ingelsson E. Comprehensive Investigation of Circulating Biomarkers and Their Causal Role in Atherosclerosis-Related Risk Factors and Clinical Events. Circ Genom Precis Med. 2020;13(6):e002996. Du M, Li S, Jiang J, Ma X, Liu L, Wang T, et al. Advances in the Pathogenesis and Treatment Strategies for Type 1 Diabetes Mellitus. Int Immunopharmacol. 2025;148:114185. Jun JE, Kang H, Hwang YC, Ahn KJ, Chung HY, Jeong IK. The association between lipoprotein (a) and carotid atherosclerosis in patients with type 2 diabetes without pre-existing cardiovascular disease: A cross-sectional study. Diabetes Res Clin Pract. 2021;171:108622. Vouillarmet J, Marsot C, Maucort-Boulch D, Riche B, Helfre M, Grange C. Vascular Events and Carotid Atherosclerosis: A 5-Year Prospective Cohort Study in Patients with Type 2 Diabetes and a Contemporary Cardiovascular Prevention Treatment. J Diabetes Res. 2019;2019:9059761. Xing L, Li R, Zhang S, Li D, Dong B, Zhou H, et al. High Burden of Carotid Atherosclerosis in Rural Northeast China: A Population-Based Study. Front Neurol. 2021;12:597992. Additional Declarations No competing interests reported. Supplementary Files SupplementaryMaterial.docx TableS1.docx TableS2.xlsx FigureS1.tif FigureS2.tif FigureS3.tif Cite Share Download PDF Status: Published Journal Publication published 03 Jul, 2025 Read the published version in BMC Cardiovascular Disorders → Version 1 posted Editorial decision: Revision requested 15 Apr, 2025 Editor assigned by journal 15 Apr, 2025 Reviewers agreed at journal 13 Apr, 2025 Reviews received at journal 08 Apr, 2025 Reviewers agreed at journal 08 Apr, 2025 Reviews received at journal 08 Apr, 2025 Reviewers agreed at journal 08 Apr, 2025 Reviewers invited by journal 07 Apr, 2025 Submission checks completed at journal 07 Apr, 2025 First submitted to journal 05 Apr, 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-6131962","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":440053386,"identity":"67d5efc2-576f-411d-b5a4-16396f058cbc","order_by":0,"name":"Xiaoqin Chen","email":"","orcid":"","institution":"Department of Endocrinology, Third Hospital of Shanxi Medical University, Shanxi Bethune Hospital, Shanxi Academy of Medical Sciences, Tongji Shanxi Hospital","correspondingAuthor":false,"prefix":"","firstName":"Xiaoqin","middleName":"","lastName":"Chen","suffix":""},{"id":440053387,"identity":"2f87d7a7-6e8b-45d9-985c-df13d2f5e336","order_by":1,"name":"Zhitong Li","email":"","orcid":"","institution":"Third Hospital of Shanxi Medical University, Shanxi Bethune Hospital, Shanxi Academy of Medical Sciences, Tongji Shanxi Hospital","correspondingAuthor":false,"prefix":"","firstName":"Zhitong","middleName":"","lastName":"Li","suffix":""},{"id":440053388,"identity":"ed5974f4-ad1d-439c-a091-3a6ec36506cb","order_by":2,"name":"Xiaoying Fan","email":"","orcid":"","institution":"Third Hospital of Shanxi Medical University, Shanxi Bethune Hospital, Shanxi Academy of Medical Sciences, Tongji Shanxi Hospital","correspondingAuthor":false,"prefix":"","firstName":"Xiaoying","middleName":"","lastName":"Fan","suffix":""},{"id":440053389,"identity":"8f5e14c7-bb27-4b88-a3ce-3bb54a66f253","order_by":3,"name":"Yuanyuan Yan","email":"","orcid":"","institution":"Third Hospital of Shanxi Medical University, Shanxi Bethune Hospital, Shanxi Academy of Medical Sciences, Tongji Shanxi Hospital","correspondingAuthor":false,"prefix":"","firstName":"Yuanyuan","middleName":"","lastName":"Yan","suffix":""},{"id":440053391,"identity":"23bf7322-3143-4c18-96bf-7518ec52ef02","order_by":4,"name":"Shiwei Liu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA4klEQVRIiWNgGAWjYBACPgglwcDA3nyA4YEBEVrY4Fp4jiUwJJCgBaTLx4AhgRiHsfGfMf5cuMciTz6C5+OHhILDidsZmB8+uoFPi0SOmfSMZxLFhrd7N0skGBxO3NnAZmycg1cL7zZmngMSiRvnnN0A1rLhAA+bNF4t/Gc3fwZrmZHz+AdxWhhyN0iDtMyXyGEj0haJ/G/SM4BaNvAcM7NIMEg33nCYgF/4+Y8lfy44UJc4v7358Y0Pf6xlNxxvfvgYnxYQYAYRBgfA7GYolxgt8g1gdh0R6kfBKBgFo2CkAQDpoE7awwJfjwAAAABJRU5ErkJggg==","orcid":"","institution":"Department of Endocrinology, Third Hospital of Shanxi Medical University, Shanxi Bethune Hospital, Shanxi Academy of Medical Sciences, Tongji Shanxi Hospital","correspondingAuthor":true,"prefix":"","firstName":"Shiwei","middleName":"","lastName":"Liu","suffix":""}],"badges":[],"createdAt":"2025-03-01 01:53:19","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6131962/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6131962/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1186/s12872-025-04786-6","type":"published","date":"2025-07-03T15:58:46+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":80580044,"identity":"c5d9045f-1b5f-4e5b-929f-55d40550b5d0","added_by":"auto","created_at":"2025-04-14 23:12:26","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":209517,"visible":true,"origin":"","legend":"\u003cp\u003eThe complete workflow of our study.\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-6131962/v1/0fb8349a205ad50735b4d965.png"},{"id":80580045,"identity":"bcad6f77-ad78-4f85-a0aa-90d738d55e72","added_by":"auto","created_at":"2025-04-14 23:12:27","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":1007246,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eComprehensive evaluation of machine learning models.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(A) ROC and AUC of the training set; (B) ROC and AUC of the testing set; (C) Calibration curve analysis of the testing set; (D) Decision curve analysis for net benefit across threshold ranges.\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-6131962/v1/dc342512d8c7477bbeea91f6.png"},{"id":80580091,"identity":"3315759a-b3f8-4181-89aa-6401cbca114f","added_by":"auto","created_at":"2025-04-14 23:12:30","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":49751,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eNomogram used for predicting carotid atherosclerosis in LADA patients.\u003c/strong\u003e This nomogram integrates clinical predictors (Age, Smoking status, BMI, ALB, HDL, and ALT) to estimate individualized risk probabilities. Example of clinical application: For a hypothetical 30-year-old patient who don’t smoke (Smoke=0), with a BMI of 28, ALB of 1.6 g/dL, HDL of 35 mg/dL, and ALT of 80 U/L: Age: 30 years → 15 points; Smoking: No → 0 points; BMI: 28 → 7 points; ALB: 30 g/L → 28 points; HDL: 1 mmol/L → 15 points; ALT: 80 U/L → 18 points. Total points: 15 + 0 + 7 + 28 + 15 + 18 = 83 points. Corresponding to a ~75% risk probability of carotid atherosclerosis. Clinicians can use this tool to visualize risk stratification and guide personalized interventions.\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-6131962/v1/d1a02b506b506f9440f142fc.png"},{"id":80580090,"identity":"7aa647a0-7177-4fb0-86de-48cfef0649d5","added_by":"auto","created_at":"2025-04-14 23:12:30","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":704348,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSHAP interprets the NNET model.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e(A) Ranking of variable importance based on the average SHAP value. (B) All samples and features are illustrated, with each row representing a feature and x-axis representing the SHAP value. The yellow dots represent higher feature values, while the purple dots represent lower feature values. (C) The SHAP dependence plot of the NNET model. (D) The SHAP waterfall plot of the NNET model.\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-6131962/v1/7420bb70bd911aabc2da4729.png"},{"id":86179616,"identity":"f4da07f2-9a55-46ae-a302-0bafe88e6f1d","added_by":"auto","created_at":"2025-07-07 16:17:52","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1557750,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6131962/v1/2933802e-3b5f-4238-8f88-8a09d522415f.pdf"},{"id":80582348,"identity":"2df3d1f3-6d2e-41be-b192-b61b135eddfb","added_by":"auto","created_at":"2025-04-14 23:28:27","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":630795,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterial.docx","url":"https://assets-eu.researchsquare.com/files/rs-6131962/v1/71c6c08f20a1681321131e40.docx"},{"id":80581298,"identity":"eb047e31-7138-445a-8604-7fff739247f2","added_by":"auto","created_at":"2025-04-14 23:20:27","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":33759,"visible":true,"origin":"","legend":"","description":"","filename":"TableS1.docx","url":"https://assets-eu.researchsquare.com/files/rs-6131962/v1/a6641e527c0f567fb8d4cbbe.docx"},{"id":80580054,"identity":"73e92de2-2373-4635-916e-789c8539e671","added_by":"auto","created_at":"2025-04-14 23:12:27","extension":"xlsx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":13026,"visible":true,"origin":"","legend":"","description":"","filename":"TableS2.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6131962/v1/a095dd2d25150eaee1829c1d.xlsx"},{"id":80581300,"identity":"08f2696a-89da-4344-b0db-7752cf21bd7c","added_by":"auto","created_at":"2025-04-14 23:20:27","extension":"tif","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":699927,"visible":true,"origin":"","legend":"","description":"","filename":"FigureS1.tif","url":"https://assets-eu.researchsquare.com/files/rs-6131962/v1/7f942b0d5aadde9896bb8760.tif"},{"id":80580057,"identity":"c5962e34-6e21-4953-9306-52c17beda2da","added_by":"auto","created_at":"2025-04-14 23:12:27","extension":"tif","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":1013180,"visible":true,"origin":"","legend":"","description":"","filename":"FigureS2.tif","url":"https://assets-eu.researchsquare.com/files/rs-6131962/v1/d9e6e7be14414b5591632367.tif"},{"id":80580055,"identity":"7dc881cd-323f-4c62-8f6b-58724a335773","added_by":"auto","created_at":"2025-04-14 23:12:27","extension":"tif","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":57733,"visible":true,"origin":"","legend":"","description":"","filename":"FigureS3.tif","url":"https://assets-eu.researchsquare.com/files/rs-6131962/v1/1183d2096dea65042971f13f.tif"}],"financialInterests":"No competing interests reported.","formattedTitle":"Predicting Carotid Atherosclerosis in Latent Autoimmune Diabetes in Adult Patients Using Machine Learning Models: A Retrospective Study","fulltext":[{"header":"1 Background","content":"\u003cp\u003eLatent autoimmune diabetes in adults (LADA), also called \u0026ldquo;type 1.5 diabetes,\u0026rdquo; is a condition marked by autoimmune beta-cell destruction and insulin resistance (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e). Epidemiological studies show that LADA accounts for 2\u0026ndash;12% of patients initially diagnosed with type 2 diabetes (T2DM), and a survey found it represented 65% of new type 1 diabetes (T1DM) cases in China (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e). Unlike T1DM, which typically appears in childhood, LADA develops in adulthood with a gradual onset, initially resembling T2DM and not requiring insulin therapy. Over time, however, beta-cell decline leads to insulin dependence. The presence of autoantibodies such as insulin autoantibodies (IAA), islet cell autoantibodies (ICA), and glutamic acid decarboxylase antibodies (GADA) distinguishes LADA from both T1DM and T2DM (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e). LADA patients also have a higher risk of cardiovascular complications. Studies from Germany and Austria found significantly higher rates of hypertension (77.7%) and dyslipidemia (90.6%) in LADA patients compared to T2DM and T1DM (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e). A Swedish study showed a 67% increased risk of cardiovascular disease in high autoimmune LADA patients (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e). These findings highlight the need to focus on cardiovascular risk, particularly atherosclerosis, in LADA patients, an area often overlooked in favor of microvascular complications.\u003c/p\u003e \u003cp\u003eAtherosclerosis is a major contributor to cardiovascular disease, with carotid atherosclerosis serving as a key marker for systemic vascular disease. Thickening and plaque formation in the carotid artery are strong predictors of stroke and coronary artery disease (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e). Traditional risk factors such as hypertension, hyperlipidemia, and diabetes contribute to its development (\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e), while recent studies highlight the role of chronic inflammation, endothelial dysfunction, and genetic polymorphisms in accelerating the process (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e). Inflammation markers like interleukin-6 are linked to the severity of carotid atherosclerosis, and in LADA patients, elevated biomarkers such as C-reactive protein and tumor necrosis factor-alpha correlate with increased cardiovascular risk (\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e). Insulin resistance-related markers like adiponectin and leptin are also abnormal in LADA, reinforcing the connection between metabolic dysfunction and atherosclerosis (\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e). Monitoring these biomarkers could help identify at-risk patients early and guide treatment. However, the silent progression of atherosclerosis limits the effectiveness of traditional prediction tools like the Framingham Risk Score in autoimmune diabetes populations (\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e). While carotid ultrasound is the gold standard for detecting plaques, its use as a universal screening tool is restricted by logistical and resource limitations, highlighting the need for alternative, precision-based prediction models.\u003c/p\u003e \u003cp\u003eMachine learning models have transformed cardiovascular risk prediction by analyzing complex datasets and capturing nonlinear relationships often missed by traditional methods. The development of explainable frameworks like shapley additive explanations (SHAP) has enhanced clinical applicability by clarifying each feature\u0026rsquo;s contribution to predictions (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e). While machine learning models have been successful in predicting macrovascular complications in T1DM and T2DM, their use in LADA-specific atherosclerosis remains unexplored, despite the unique pathophysiology of this group (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e). This retrospective study aims to develop and validate machine learning models for predicting carotid atherosclerosis in LADA patients using demographic, clinical, and laboratory data. We compare eight algorithms\u0026mdash;logistic regression (LR), decision tree (DT), random forests (RF), k-nearest neighbors (KNN), support vector machine (SVM), neural networks (NNET), eXtreme gradient boosting (XGBoost), and light gradient boosting machine (LightGBM)\u0026mdash;to 1) identify novel predictors beyond traditional risk factors, 2) create an interpretable prediction framework for clinical use, and 3) offer pathophysiological insights through feature importance analysis. Our findings may redefine cardiovascular risk assessment in LADA and provide a personalized monitoring strategy for this high-risk, under-recognized group.\u003c/p\u003e"},{"header":"2 Methods","content":"\u003cp\u003eStudy population and design\u003c/p\u003e\n\u003cp\u003eThis retrospective cross-sectional study was conducted at Shanxi Bethune Hospital, Taiyuan, China, from April 2023 to December 2024, with approval from the hospital\u0026rsquo;s Ethics Committee (Approval No. YXLL-2025-026). Informed consent was waived as the study used preexisting data without intervention, in compliance with the Declaration of Helsinki. A total of 142 patients diagnosed with LADA were included from the endocrinology department, with diagnosis based on: (\u003cspan class=\"CitationRef\"\u003e1\u003c/span\u003e) age\u0026thinsp;\u0026ge;\u0026thinsp;18 at diabetes onset, (\u003cspan class=\"CitationRef\"\u003e2\u003c/span\u003e) presence of pancreatic islet autoantibodies or autoimmune T cells, and (\u003cspan class=\"CitationRef\"\u003e3\u003c/span\u003e) no insulin dependence for at least six months. Participants were classified into two groups based on carotid ultrasound: those with and without carotid plaques. Exclusion criteria included: (\u003cspan class=\"CitationRef\"\u003e1\u003c/span\u003e) severe cardiovascular disease, (\u003cspan class=\"CitationRef\"\u003e2\u003c/span\u003e) other autoimmune diseases, (\u003cspan class=\"CitationRef\"\u003e3\u003c/span\u003e) severe liver or kidney dysfunction, (\u003cspan class=\"CitationRef\"\u003e4\u003c/span\u003e) active infections or malignancies, and (\u003cspan class=\"CitationRef\"\u003e5\u003c/span\u003e) incomplete medical records.\u003c/p\u003e\n\u003cp\u003eData collection\u003c/p\u003e\n\u003cp\u003eClinical and demographic information were retrospectively obtained from electronic medical records, which included the participants\u0026rsquo; age, gender, diabetes duration, smoking and drinking status, insulin usage, body mass index (BMI), as well as their systolic blood pressure (SBP) and diastolic blood pressure (DBP). Laboratory data encompassed measures such as fasting plasma glucose (FPG), hemoglobin A1c (HbA1c), fasting C-peptide (FCP), fasting insulin (FI), and lipid profiles (total cholesterol (TC), triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C)). Additionally, autoimmune markers (IAA, ICA, GADA), liver function tests (alanine aminotransferase (ALT), aspartate aminotransferase (AST)), renal parameters (blood urea nitrogen (BUN), serum creatinine (Scr), uric acid (UA)), albumin levels (ALB), white blood cell count (WBC), thyroid function (thyroid-stimulating hormone (TSH), free triiodothyronine (FT3), free thyroxine (FT4)), 25-hydroxyvitamin D concentrations (25(OH)D), triglyceride-glucose index (TyG), and atherosclerotic index of plasma (AIP) were also recorded. Carotid ultrasonography was conducted to assess the presence of atherosclerotic plaques in the carotid arteries. All patient data were anonymized to ensure confidentiality.The TyG index was calculated using the formula (\u003cspan class=\"CitationRef\"\u003e16\u003c/span\u003e):\u003c/p\u003e\n\u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:TyG=In\\left(\\frac{TG(mg/dL)\\times\\:FPG(mg/dL)}{2}\\right)\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n\u003cp\u003eThe AIP index was calculated using the formula (\u003cspan class=\"CitationRef\"\u003e17\u003c/span\u003e):\u003c/p\u003e\n\u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:AIP=log10\\left(\\frac{TG(mmol/L)}{HDL-C(mmol/L)}\\right)\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n\u003cp\u003ePredictive variables\u003c/p\u003e\n\u003cp\u003eThe predictive variables were firstly selected based on univariate and multivariate logistic regression analyses. In the univariate analysis, associations between individual clinical variables and the presence of carotid plaques were evaluated. Significant predictors (P \u0026lt; 0.05) identified from the univariate analysis were then included in a multivariate logistic regression model to assess independent risk factors for carotid atherosclerosis. To further refine the selection of relevant variables and prevent overfitting, least absolute shrinkage and selection operator (LASSO) regression was performed to identify the most critical variables associated with carotid plaques. Specifically, LASSO regularization was performed on the dataset to select the most significant features from the initial set of 33 variables. The LASSO algorithm underwent 10-fold cross-validation to ensure the reliability and stability of the selected features, optimizing model performance while avoiding overfitting. LASSO is a regularization technique that penalizes the inclusion of irrelevant or redundant features by applying an L1 penalty. This process encourages sparsity in the model, driving some coefficients to zero and effectively excluding less important variables. By utilizing LASSO, we were able to retain only the most significant features associated with carotid atherosclerosis in patients with LADA, thereby reducing model complexity. This not only improved the interpretability of the model but also mitigated the risk of overfitting, leading to a more reliable and accurate predictive model. Finally, collinearity among the selected predictors was checked by calculating the variance inflation factor (VIF) to ensure the independence of the variables included in the model.\u003c/p\u003e\n\u003cp\u003eMachine learning models\u003c/p\u003e\n\u003cp\u003eEight machine learning models were used to predict the presence of carotid atherosclerosis in LADA patients: LR, DT, RF, KNN, SVM, NNET, XGBoost, and LightGBM. To avoid data leakage, the training and testing datasets were independently normalized before modeling. The training set, which comprised 70% of the dataset, was used for model development, while the remaining 30% was reserved as a holdout test set for final model evaluation and adjustments. Hyperparameters for each model were optimized using a combination of grid search and cross-validation to achieve the best performance and ensure robust parameter tuning and model stability. The test set played a critical role in adjusting model parameters and estimating the model\u0026rsquo;s generalizability, ensuring that the models did not overfit and could perform well on unseen data.\u003c/p\u003e\n\u003cp\u003eModel evaluation\u003c/p\u003e\n\u003cp\u003eThe performance of the machine learning models was comprehensively evaluated using metrics such as accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and F1 score, ensuring robustness and clinical utility. A 5-fold cross-validation was employed, with the area under the receiver operating characteristic (ROC) curve (AUC) serving as the primary performance evaluation metric, to identify the best estimator (\u003cspan class=\"CitationRef\"\u003e18\u003c/span\u003e). Calibration was assessed with the Brier score, comparing predicted probabilities to actual outcomes, and calibration plots were generated for visual inspection. Decision curve analysis (DCA) was performed to evaluate the clinical utility by assessing the net benefit at different predicted risk thresholds, accounting for both true and false positives and the potential harm of unnecessary interventions.\u003c/p\u003e\n\u003cp\u003eMachine interpretation\u003c/p\u003e\n\u003cp\u003eThe LR model can use a nomogram to visually represent the relationship between predictive variables and outcomes, helping clinicians estimate the likelihood of carotid atherosclerosis. The DT model can employ a decision tree structure to show feature relationships and predictions. For more complex models like RF, KNN, SVM, NNET, XGBoost, and LightGBM, the SHAP method interprets model predictions by attributing feature contributions based on Shapley values from game theory. SHAP provides a clear, quantitative measure of feature importance and how each predictor influences the outcome, offering valuable, interpretable insights for clinicians. Figure\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e illustrates the workflow of this study.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eStatistical analysis\u003c/p\u003e\n\u003cp\u003eStatistical analysis was performed using SPSS (version 26.0) and R (version 4.4.2). Continuous variables were expressed as mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation or median (interquartile range), and categorical variables as frequencies and percentages. Group comparisons were made with the independent t-test or Mann-Whitney U test for continuous variables, and the chi-square test for categorical variables. Missing data (\u0026lt;\u0026thinsp;0.15%) were imputed using the Random Forest method. Outliers, identified by the Tukey method, were excluded to minimize bias. A p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered statistically significant.\u003c/p\u003e"},{"header":"3 Results","content":"\u003cp\u003eData processing results\u003c/p\u003e \u003cp\u003eWe analyzed the clinical and biochemical characteristics of 142 LADA patients, splitting them into a training set (n\u0026thinsp;=\u0026thinsp;100) and a testing set (n\u0026thinsp;=\u0026thinsp;42) to develop and validate machine learning models for predicting carotid atherosclerosis. No significant differences were found between the two sets for most clinical variables (age, gender, diabetes duration, smoking and drinking status, insulin usage, and carotid atherosclerosis status) or biochemical markers, including BMI, blood pressure, HbA1c, FPG, lipid profiles, autoimmune markers, liver and kidney function, thyroid hormones, and vitamin D (P\u0026thinsp;\u0026gt;\u0026thinsp;0.05 for all). These results demonstrate a well-balanced dataset for developing predictive models for carotid atherosclerosis in LADA patients (see Table S2).\u003c/p\u003e \u003cp\u003eFeature selection\u003c/p\u003e \u003cp\u003eWe employed a multi-step feature selection process for predicting carotid atherosclerosis in LADA patients, combining regression analyses and LASSO (\u003cspan additionalcitationids=\"CR20\" citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e). Univariate logistic regression first identified variables linked to carotid plaques (P\u0026thinsp;\u0026lt;\u0026thinsp;0.05), and multivariate analysis confirmed age and smoking history as independent risk factors (Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e). To prevent overfitting, LASSO regression was applied with normalized data, using carotid atherosclerosis as the dependent variable. LASSO utilized compressive coefficients and 10-fold cross-validation to establish the optimal penalty parameter, λ, with the value minimizing binomial deviance selected for a balanced model (Figures \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003eA and S1B). From 33 variables, LASSO identified age, smoking history, BMI, ALB, HDL-C, and ALT as key predictors. The VIF for these variables were all below 2, indicating no significant multicollinearity.\u003c/p\u003e \u003cp\u003eModel evaluation and comparison\u003c/p\u003e \u003cp\u003eIn this study, we evaluated eight machine learning models for predicting carotid atherosclerosis in LADA patients. The models tested include LR, DT, RF, KNN, SVM, NNET, XGBoost, and LightGBM. Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e presents the AUC, accuracy, sensitivity, specificity, PPV, NPV, F1 score, and Brier score for each model. Figure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA illustrates the AUC values for the training set, while Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB shows the AUC values for the testing set.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eEvaluation of the performance of the eight algorithms.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"12\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c10\" colnum=\"10\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c11\" colnum=\"11\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c12\" colnum=\"12\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eData set\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003eAlgorithm\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eAUC\u003c/p\u003e \u003cp\u003e(95%CI)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eAccuracy\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eSensitivity\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eSpecificity\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003ePPV\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c9\"\u003e \u003cp\u003eNPV\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c11\" namest=\"c10\"\u003e \u003cp\u003eF1 score\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c12\"\u003e \u003cp\u003eBrier Score\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eTrain\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eLR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.956 (0.922\u0026ndash;0.989)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.85\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.81\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.88\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.83\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.86\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.82\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c12\" namest=\"c11\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eTest\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eLR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.936 (0.866-1.000)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.86\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.85\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.86\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.85\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.86\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.85\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c12\" namest=\"c11\"\u003e \u003cp\u003e0.101\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eTrain\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDT\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.843 (0.770\u0026ndash;0.915)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.84\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.86\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.82\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.79\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.82\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c12\" namest=\"c11\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eTest\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDT\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.811 (0.691\u0026ndash;0.932)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.81\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.85\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.77\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.77\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.85\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.81\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c12\" namest=\"c11\"\u003e \u003cp\u003e0.153\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eTrain\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.999 (0.997-1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.97\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.95\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.98\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.98\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.97\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.96\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c12\" namest=\"c11\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eTest\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.857 (0.746\u0026ndash;0.967)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.74\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.70\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.77\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.74\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.74\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.72\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c12\" namest=\"c11\"\u003e \u003cp\u003e0.155\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eTrain\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eKNN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.969 (0.943\u0026ndash;0.995)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.88\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.86\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.86\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.86\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c12\" namest=\"c11\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eTest\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eKNN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.898 (0.805\u0026ndash;0.990)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.86\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.85\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.86\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.85\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.86\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.85\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c12\" namest=\"c11\"\u003e \u003cp\u003e0.144\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eTrain\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSVM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.950 (0.913\u0026ndash;0.986)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.87\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.81\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.91\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.88\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.86\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.84\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c12\" namest=\"c11\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eTest\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSVM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.918 (0.833-1.000)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.86\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.80\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.91\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.83\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.84\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c12\" namest=\"c11\"\u003e \u003cp\u003e0.109\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eTrain\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNNET\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.981 (0.961-1.000)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.93\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.98\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.88\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.98\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.93\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c12\" namest=\"c11\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eTest\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNNET\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.919 (0.834-1.000)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.86\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.95\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.77\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.79\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.94\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.86\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c12\" namest=\"c11\"\u003e \u003cp\u003e0.109\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eTrain\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eXGBoost\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.991 (0.979-1.000)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.94\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.88\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.98\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.97\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c12\" namest=\"c11\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eTest\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eXGBoost\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.898 (0.800-0.995)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.83\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.85\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.82\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.81\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.86\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.83\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c12\" namest=\"c11\"\u003e \u003cp\u003e0.125\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eTrain\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eLightGBM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.965 (0.934\u0026ndash;0.995)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.86\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.91\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.88\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.90\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.87\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c12\" namest=\"c11\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eTest\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eLightGBM\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.880 (0.773\u0026ndash;0.986)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.81\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.80\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.82\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.80\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e0.82\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c10\"\u003e \u003cp\u003e0.80\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c12\" namest=\"c11\"\u003e \u003cp\u003e0.136\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"12\"\u003eAUC: area under the curve; PPV: positive predictive value; NPV: negative predictive value; F1 score = (2 * Precision * Recall) / ( Precision\u0026thinsp;+\u0026thinsp;Recall).\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eAmong the evaluated models, LR, NNET, and SVM showed superior performance across multiple metrics. LR achieved an AUC of 0.936 (0.866 to 1), with accuracy, sensitivity, and specificity all at 86%, and an F1 score of 0.85, indicating good precision and recall. NNET had an AUC of 0.919 (0.834 to 1), with 86% accuracy, 95% sensitivity, and 77% specificity, along with a strong F1 score of 0.86. SVM performed similarly well with an AUC of 0.918 (0.833 to 1), 86% accuracy, 80% sensitivity, and 91% specificity, and an F1 score of 0.84. All three models showed low Brier scores, indicating good calibration. Other models like KNN, XGBoost, RF, and LightGBM had competitive performance but lower metrics. Random Forest showed the highest training AUC (0.999) but dropped in performance on the test set with an AUC of 0.857 and accuracy of 74%. XGBoost and LightGBM had moderate AUCs of 0.898 and 0.880, respectively, but their accuracy and sensitivity were not as high as LR, NNET, and SVM.\u003c/p\u003e \u003cp\u003eThe DT model, although simpler, provided insightful results in predicting carotid atherosclerosis in LADA patients. The final decision tree had a clear, binary classification rule based solely on a single feature: age, with a split criterion at 44.5 years of age. This model, due to its simplicity, had a lower AUC (0.811) compared to the other more complex models, but it still provided a valuable baseline for understanding the role of a single variable in predicting outcomes. Despite its lower overall performance, the DT model was useful for visualizing the relationship between age and the risk of carotid atherosclerosis, as demonstrated in the supplementary Figure S3. The simplicity of the DT also offers an advantage in interpretability, which may be beneficial in clinical practice, where transparency in decision-making is crucial.\u003c/p\u003e \u003cp\u003eTo compare the practical utility of the models, we performed calibration curve analysis to evaluate their performance in terms of calibration (see Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eC). The calibration curves for LR, NNET, and SVM showed excellent alignment with the ideal diagonal, indicating that these models were well-calibrated. The Brier scores for these models were among the lowest, further confirming their good calibration and ability to provide reliable probability estimates. This reinforces the conclusion that LR, NNET, and SVM are not only accurate in prediction but also offer well-calibrated results, which are crucial for making reliable clinical decisions. Additionally, we used DCA to assess the net benefit of the models across various threshold ranges (see Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eD). The DCA results revealed that LR, NNET, and SVM had the highest net benefit across most of the threshold ranges. This further supports the clinical value of these models, as they offer a higher overall benefit to patients, especially when the threshold for classification is optimized to balance between false positives and false negatives.\u003c/p\u003e \u003cp\u003eModel interpretation\u003c/p\u003e \u003cp\u003eFor the LR model, we utilized nomograms to visually interpret the relationship between the model\u0026rsquo;s input features and its output. The nomogram (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e) illustrates how each of the six features\u0026mdash;age, smoking history, BMI, ALB, HDL-C, and ALT\u0026mdash;affects the likelihood of predicting carotid atherosclerosis. Each feature was assigned a score based on its corresponding coefficient in the model, and the total score was used to estimate the probability of atherosclerosis. This visualization allows clinicians to easily understand the relative importance of each variable, with higher scores corresponding to a greater likelihood of the disease. Notably, the nomogram showed that age was a particularly influential factor, with older age strongly associated with an increased risk of developing atherosclerosis.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTo address the challenge of explaining complex prediction models, we applied SHAP to quantify the contribution of each feature to the predictions. SHAP provides two types of explanations: global (overall model behavior) and local (individual prediction-level insights). For the NNET model, the SHAP summary bar plot (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eA) displayed the mean SHAP values for each feature, ranked by their importance in the prediction. Age, ALB, and smoking history emerged as the most influential factors. The SHAP summary dot plot (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eB) further illustrates the direction and strength of these influences, revealing that features such as advanced age, low ALB levels, and smoking significantly increased the risk of carotid atherosclerosis in LADA patients. The SHAP dependence plot (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eC) provided more granular insights into how individual features directly impacted the model\u0026rsquo;s prediction. To facilitate understanding, we also presented a typical prediction example (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eD), demonstrating the interpretability of the NNET model in predicting carotid atherosclerosis. For the SVM model, SHAP analysis identified age, smoking history, and ALB as the most critical features (see Figure S2).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e"},{"header":"4 Discussion","content":"\u003cp\u003eThis study aimed to develop and validate machine learning models for predicting carotid atherosclerosis in LADA patients, focusing on identifying novel predictors beyond traditional cardiovascular risk factors. Eight machine learning algorithms were tested, including LR, DT, RF, KNN, SVM, NNET, XGBoost, and LightGBM, to compare their performance for risk stratification. The LR, NNET, and SVM models demonstrated the best results, with LR achieving the highest AUC of 0.936 and excellent accuracy (86%), sensitivity (85%), and specificity (86%). NNET excelled in sensitivity (95%), while SVM showed strong specificity (91%). These models were well-calibrated, ensuring reliable predictions for clinical use. Key predictors identified for carotid atherosclerosis included age, smoking history, BMI, ALB, HDL-C, and ALT. Notably, ALB was identified as a new predictor, suggesting that low ALB levels may signal early vascular damage. Additionally, HDL-C and ALT further underscored the importance of lipid metabolism and liver function in the development of carotid atherosclerosis. These findings highlight the value of incorporating these markers into routine clinical assessments for better risk stratification and early intervention.\u003c/p\u003e \u003cp\u003eRecent studies have increasingly applied machine learning models to predict cardiovascular risks and atherosclerosis in patients with T1DM and T2DM, leveraging diverse clinical and demographic data. In T2DM, machine learning algorithms, including support vector machines, random forests, and deep learning models, have demonstrated significant potential in predicting cardiovascular outcomes by analyzing risk factors such as blood glucose levels, lipid profiles, blood pressure, and patient history (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e). Similarly, in T1DM, where patients are at a higher risk for atherosclerosis due to chronic hyperglycemia and autoimmune factors, machine learning approaches have been used to identify early signs of vascular damage, often incorporating biomarkers like C-reactive protein and advanced glycation end products (\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e). However, despite the robust application of machine learning in these populations, there is a paucity of research specifically targeting LADA, a subtype of diabetes with characteristics of both T1DM and T2DM. Our study aims to fill this gap by applying ML techniques to predict carotid atherosclerosis specifically in LADA patients, thus providing a unique contribution to the field and a comparison point to the existing literature on T1DM and T2DM.\u003c/p\u003e \u003cp\u003eA key finding of our study is the significant role of age as the most prominent predictor of carotid atherosclerosis in LADA patients. Age is a well-established cardiovascular risk factor, and our results align with this, as aging increases the impact of chronic hyperglycemia, dyslipidemia, and hypertension on vascular damage (\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e). The autoimmune component and insulin resistance in LADA likely accelerate atherosclerosis, making age particularly relevant in this cohort. Our findings support previous studies linking age with cardiovascular disease progression, with age serving as a proxy for factors like diabetes duration and comorbidities such as hypertension. Older individuals often experience more severe metabolic syndrome and autoimmune dysfunction, reinforcing age\u0026rsquo;s clinical significance as a predictor in our models. The relationship between age and other clinical factors, such as insulin use, is complex. LADA\u0026rsquo;s gradual progression means that older patients often require insulin therapy, reflecting disease duration (\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e). Insulin use, which correlates with deteriorating glycemic control, further underscores age as a marker of disease progression and cardiovascular risk. This interplay between age, insulin use, and metabolic dysfunction contributes to the accelerated development of atherosclerosis in LADA patients (\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eOur study identified several factors beyond age as significant predictors of carotid atherosclerosis in LADA, emphasizing the multifactorial nature of cardiovascular risk (\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e). Smoking, a known risk factor, accelerates endothelial dysfunction, oxidative stress, and inflammation, all contributing to plaque formation (\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e). Our study affirms smoking as a potent risk factor in LADA, highlighting the importance of smoking cessation in cardiovascular risk management. Similarly, obesity and high BMI are linked to insulin resistance, dyslipidemia, and chronic inflammation, which drive atherosclerosis (\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e). We confirm that a higher BMI predicts carotid atherosclerosis, reinforcing the need for weight management. Additionally, ALB, a marker of nutrition and inflammation, was inversely associated with atherosclerosis, suggesting that inflammation and poor nutritional status contribute to vascular damage in LADA patients (\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e). Low HDL-C levels, indicating poor metabolic control and lipid dysregulation, were also found to increase the risk of atherosclerosis (\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e). Finally, elevated ALT, a marker of hepatic metabolic dysfunction and NAFLD, was identified as a predictor of carotid atherosclerosis, linking liver health with cardiovascular risk in LADA patients (\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe predictive value of age, smoking, BMI, and other clinical variables in cardiovascular disease is well-documented in diabetes populations. However, this study is the first to apply machine learning models to LADA patients, a group underrepresented in predictive modeling research. Most studies have focused on T1DM or T2DM, leaving a gap in understanding the unique cardiovascular risks of LADA. Carotid atherosclerosis mechanisms differ between T1DM and T2DM: T1DM, typically diagnosed earlier, leads to early atherosclerosis due to prolonged hyperglycemia (\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e), while T2DM, affecting older adults with obesity, hypertension, and dyslipidemia, shows faster disease progression (\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e). T1DM patients exhibit higher carotid intima-media thickness and more plaque, but T2DM patients have more complex risk factors, resulting in higher cardiovascular event rates (\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e). Traditional carotid atherosclerosis prediction models rely on clinical factors such as age, gender, smoking, blood pressure, lipid levels, and diabetes presence, often using statistical methods like logistic regression or Cox regression. Tools like the Framingham and ASCVD Risk Scores integrate multiple clinical parameters to predict cardiovascular event risk (\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThis study applies machine learning to identify cardiovascular risks in LADA patients, offering a novel approach to risk stratification. Carotid atherosclerosis, a key early indicator of cardiovascular disease, can be detected before symptoms appear. LADA patients, who face combined autoimmune and metabolic risks, benefit from early identification to prevent severe outcomes like stroke and heart attacks. Traditional risk assessments often rely on T2DM guidelines, which may overlook unique LADA factors (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e). Our study shows that machine learning models can provide more accurate, individualized predictions, enabling earlier interventions such as lifestyle changes and pharmacotherapy. Additionally, liver function markers like ALT could be used in routine screening, complementing traditional risk factors. Machine learning can improve personalized medicine by incorporating diverse data to create comprehensive risk profiles, which can be integrated into electronic health records for real-time assessments. This could enhance preventive care and optimize patient outcomes, particularly in settings with limited access to advanced imaging techniques.\u003c/p\u003e \u003cp\u003eIt is also important to acknowledge the limitations of this study. First, while the sample size of 142 patients is adequate for exploratory analysis, it may limit the generalizability of the results. A larger, more diverse cohort would provide better validation of the predictive model and enhance its applicability across different ethnic groups. Additionally, the retrospective nature of the study may introduce biases in data collection, particularly concerning the completeness of clinical records or the underreporting of certain risk factors. Prospective studies with standardized data collection methods are necessary to further validate these findings. Furthermore, while this study incorporated a range of clinical and biochemical factors, the inclusion of additional variables, such as genetic markers, advanced imaging techniques, and more detailed lifestyle factors (e.g., dietary habits and physical activity levels), could enhance the accuracy of the models. Incorporating these factors would provide a more comprehensive approach to predicting atherosclerosis risk in LADA patients. Lastly, although the models showed strong performance in terms of AUC and calibration, they would require external validation in different clinical settings and populations to confirm their clinical applicability.\u003c/p\u003e \u003cp\u003eDespite some limitations, this study demonstrates significant clinical potential for machine learning-based predictive models in identifying high-risk LADA patients. By detecting those at risk for carotid atherosclerosis, early interventions such as lipid-lowering therapies or immune-modulating treatments can be implemented to reduce cardiovascular events. These models can also help optimize healthcare resources by prioritizing high-risk patients for intensive screenings and treatments, while lower-risk patients require less intensive monitoring. Additionally, while traditional statistical methods provide valuable insights into the relationship between features and the outcome, machine learning models offer greater flexibility and the ability to capture more complex, non-linear interactions in the progress of feature selection. By combining both approaches, future studies can provide a more comprehensive understanding of cardiovascular risk and improve predictive accuracy, ultimately leading to more personalized and effective interventions.\u003c/p\u003e \u003cp\u003eWhile this study emphasizes the technical aspects of machine learning, it is essential to also consider the ethical implications of applying these models in clinical practice. Ensuring that these models do not introduce bias, especially in diverse populations, is a critical concern. Bias can arise if training data lacks diversity, leading to inaccurate predictions for underrepresented groups. Additionally, clinicians may face situations where a model\u0026rsquo;s prediction contradicts their clinical judgment, and in such cases, the model should serve as a supplementary tool, rather than replacing expert decision-making. Ethical considerations, such as ensuring transparency, obtaining informed consent, and minimizing biases in predictions, must be addressed to foster trust and efficacy in these technologies. Future research should focus on the ethical, legal, and social aspects of implementing machine learning models in practice, as well as validating findings in larger, more diverse populations. Furthermore, including additional biomarkers could enhance model accuracy and help personalize cardiovascular risk assessments for LADA patients, thereby improving both the predictive power and fairness of the model in diverse clinical settings. .\u003c/p\u003e"},{"header":"5 Conclusions","content":"\u003cp\u003eIn conclusion, our study has successfully identified key demographic and clinical risk factors associated with carotid atherosclerosis in LADA patients while demonstrating the potential of machine learning models in enhancing predictive accuracy. The findings highlight the superior performance of models such as LR, NNET, and SVM, offering valuable tools for clinical decision-making. The integration of these predictive methodologies into routine practice could facilitate early intervention and improved management strategies for LADA patients at risk of cardiovascular complications. Future research should aim to validate these models in larger cohorts and explore their implementation in clinical settings to optimize patient outcomes.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cdiv class=\"DefinitionList\"\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eLADA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eLatent autoimmune diabetes in adults\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eT2DM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003etype 2 diabetes\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eT1DM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003etype 1 diabetes\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eIAA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003einsulin autoantibodies\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eICA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eislet cell autoantibodies\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eGADA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eglutamic acid decarboxylase antibodies\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSHAP\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eshapley additive explanations\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eLR\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003elogistic regression\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eDT\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003edecision tree\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eRF\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003erandom forests\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eKNN\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ek-nearest neighbors, SVM:support vector machine\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eNNET\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eneural networks\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eXGBoost\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eeXtreme gradient boosting\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eLightGBM\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003elight gradient boosting machine\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eBMI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ebody mass index\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSBP\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003esystolic blood pressure\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eDBP\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ediastolic blood pressure\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eFPG\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003efasting plasma glucose\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eHbA1c\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ehemoglobin A1c\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eFCP\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003efasting C-peptide\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eFI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003efasting insulin\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eTC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003etotal cholesterol\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eTG\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003etriglycerides\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eHDL-C\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ehigh-density lipoprotein cholesterol\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eLDL-C\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003elow-density lipoprotein cholesterol\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eALT\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ealanine aminotransferase\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eAST\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003easpartate aminotransferase\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eBUN\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eblood urea nitrogen\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eScr\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eserum creatinine\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eUA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003euric acid\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eALB\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ealbumin levels\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eWBC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ewhite blood cell\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eTSH\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ethyroid-stimulating hormone\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eFT3\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003efree triiodothyronine\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eFT4\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003efree thyroxine\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003e25(OH)D\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003e25-hydroxyvitamin D\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eTyG\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003etriglyceride-glucose index\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eAIP\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eatherosclerotic index of plasma\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eLASSO\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eleast absolute shrinkage and selection operator\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eVIF\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003evariance inflation factor\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003ePPV\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003epositive predictive value\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eNPV\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003enegative predictive value\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eROC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ereceiver operating characteristic\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eAUC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ethe area under the ROC curve\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eDCA\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003edecision curve analysis\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"Declarations","content":" \u003cp\u003e \u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e \u003cp\u003eApproval of the research protocol: The protocol for this research project has been approved by a suitably constituted Ethics Committee of the institution and it conforms to the provisions of the Declaration of Helsinki. Ethics Committee of Shanxi Bethune Hospital, Taiyuan, China, Approval No. YXLL-2025-026.\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eInformed Consent:\u003c/strong\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/p\u003e\u003cp\u003e \u003ch2\u003eConsent for publication\u003c/h2\u003e \u003cp\u003e Informed consent was waived for this study by the ethics committee/institutional review board, as it is a retrospective analysis of preexisting anonymized clinical data without direct patient intervention or use of identifying images.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eFunding\u003c/h2\u003e \u003cp\u003eThis work was supported by Collaborative Traditional Chinese and Western Medicine for Chronic Disease Management Research Project (CXZH2024085); Shanxi Province Metabolic Disease (Type 1 Diabetes) Clinical Medical Research Center (20240410501001); Shanxi Province Science and Technology Achievements Transformation Guidance Special Fund (202304021301066); Shanxi Province Science and Technology Innovation Talent Team Special Plan (202204051002029); Shanxi Province Research Funding for Returned Overseas Scholars (2024\u0026thinsp;\u0026minus;\u0026thinsp;143); Shanxi Province Basic Research Program (202303021212330); Shanxi Province Key Laboratory of Endocrine and Metabolic Diseases (202404010920011).\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eXC, XF, and YY contributed to data collection. XC and ZL contributed to data analysis. ZL and XC contributed to data interpretation. ZL, XC and SL collectively interpreted the results. ZL wrote the manuscript. XC and SL substantively revised it. All authors had access to the data, and reviewed and approved the final manuscript before submission.\u003c/p\u003e\u003ch2\u003eAcknowledgements\u003c/h2\u003e \u003cp\u003eNot applicable\u003c/p\u003e \u003cp\u003eClinical trial number\u003c/p\u003e \u003cp\u003eNot applicable\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe datasets analysed during the current study are available from the corresponding author on reasonable request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eBuzzetti R, Tuomi T, Mauricio D, Pietropaolo M, Zhou Z, Pozzilli P, et al. Management of Latent Autoimmune Diabetes in Adults: A Consensus Statement From an International Expert Panel. Diabetes. 2020;69(10):2037\u0026ndash;47.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShuoming L, Zhiguang Z. Interpretation of Consensus of Chinese experts on the diagnosis and treatment of latent autoimmune diabetes in adults (2021 Edition). In: (Chinese), editor. Chinese Journal of Diabetes Mellitus2022. pp. 17\u0026ndash;20.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu B, Xiang Y, Liu Z, Zhou Z. Past, present and future of latent autoimmune diabetes in adults. Diabetes Metab Res Rev. 2020;36(1):e3205.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMishra R, Hodge KM, Cousminer DL, Leslie RD, Grant SFA. A Global Perspective of Latent Autoimmune Diabetes in Adults. Trends Endocrinol Metab. 2018;29(9):638\u0026ndash;50.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGolomb RC, Tittel SR, Welters A, Karges W, Meyh\u0026ouml;fer S, Hummel M, et al. Increased cardiovascular risk in people with LADA in comparison to type 1 diabetes and type 2 diabetes: Findings from the DPV registry in Germany and Austria. Diabetes Obes Metab. 2025;27(2):563\u0026ndash;73.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWei Y, Herzog K, Ahlqvist E, Andersson T, Nystr\u0026ouml;m T, Zhan Y, et al. All-Cause Mortality and Cardiovascular and Microvascular Diseases in Latent Autoimmune Diabetes in Adults. Diabetes Care. 2023;46(10):1857\u0026ndash;65.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSirimarco G, Amarenco P, Labreuche J, Touboul PJ, Alberts M, Goto S, et al. Carotid atherosclerosis and risk of subsequent coronary event in outpatients with atherothrombosis. Stroke. 2013;44(2):373\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLu SX, Wu TW, Chou CL, Cheng CF, Wang LY. Combined effects of hypertension, hyperlipidemia, and diabetes mellitus on the presence and severity of carotid atherosclerosis in community-dwelling elders: A community-based study. J Chin Med Assoc. 2023;86(2):220\u0026ndash;6.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXie Y, Yu M, Qing T, Luo H, Shao M, Wei W, et al. Variants in genes related to inflammation and endothelial function can increase the risk for carotid atherosclerosis in southwestern China. Front Neurol. 2023;14:1174425.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlmeida-Santiago C, Quevedo-Abeledo JC, Hern\u0026aacute;ndez-Hern\u0026aacute;ndez V, de Vera-Gonz\u0026aacute;lez A, Gonz\u0026aacute;lez-Delgado A, Gonz\u0026aacute;lez-Gay M, et al. Circulating interleukin-6 and cardiovascular disease risk in patients with rheumatoid arthritis with low disease activity due to active therapy. Clin Exp Rheumatol. 2023;41(7):1537\u0026ndash;43.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCastelblanco E, Hern\u0026aacute;ndez M, Castelblanco A, Gratac\u0026ograve;s M, Esquerda A, Moll\u0026oacute; \u0026Agrave;, et al. Low-grade Inflammatory Marker Profile May Help to Differentiate Patients With LADA, Classic Adult-Onset Type 1 Diabetes, and Type 2 Diabetes. Diabetes Care. 2018;41(4):862\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDamen JA, Pajouheshnia R, Heus P, Moons KGM, Reitsma JB, Scholten R, et al. Performance of the Framingham risk models and pooled cohort equations for predicting 10-year risk of cardiovascular disease: a systematic review and meta-analysis. BMC Med. 2019;17(1):109.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePonce-Bobadilla AV, Schmitt V, Maier CS, Mensing S, Stodtmann S. Practical guide to SHAP analysis: Explaining supervised machine learning model predictions in drug development. Clin Transl Sci. 2024;17(11):e70056.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhou Y, Liu C, Zhang Z, Chen J, Zhao D, Li L, et al. Identification and validation of diagnostic biomarkers of coronary artery disease progression in type 1 diabetes via integrated computational and bioinformatics strategies. Comput Biol Med. 2023;159:106940.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAbegaz TM, Baljoon A, Kilanko O, Sherbeny F, Ali AA. Machine learning algorithms to predict major adverse cardiovascular events in patients with diabetes. Comput Biol Med. 2023;164:107289.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlizargar J, Bai CH, Hsieh NC, Wu SV. Use of the triglyceride-glucose index (TyG) in cardiovascular disease patients. Cardiovasc Diabetol. 2020;19(1):8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYin B, Wu Z, Xia Y, Xiao S, Chen L, Li Y. Non-linear association of atherogenic index of plasma with insulin resistance and type 2 diabetes: a cross-sectional study. Cardiovasc Diabetol. 2023;22(1):157.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang Z, Gu Y, Huang L, Liu S, Chen Q, Yang Y, et al. Construction of machine learning diagnostic models for cardiovascular pan-disease based on blood routine and biochemical detection data. Cardiovasc Diabetol. 2024;23(1):351.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAiniwaer A, Hou WQ, Kadier K, Rehemuding R, Liu PF, Maimaiti H, et al. A Machine Learning Framework for Diagnosing and Predicting the Severity of Coronary Artery Disease. Rev Cardiovasc Med. 2023;24(6):168.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarbieri MC, Grisci BI, Dorn M. Analysis and comparison of feature selection methods towards performance and stability. Expert Syst Appl. 2024;249:123667.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCai J, Luo J, Wang S, Yang S. Feature selection in machine learning: A new perspective. Neurocomputing. 2018;300:70\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDong W, Wan EYF, Fong DYT, Tan KC, Tsui WW, Hui EM, et al. Development and validation of 10-year risk prediction models of cardiovascular disease in Chinese type 2 diabetes mellitus patients in primary care using interpretable machine learning-based methods. Diabetes Obes Metab. 2024;26(9):3969\u0026ndash;87.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuang Q, Zou X, Lian Z, Zhou X, Han X, Luo Y, et al. Predicting cardiovascular outcomes in Chinese patients with type 2 diabetes by combining risk factor trajectories and machine learning algorithm: a cohort study. Cardiovasc Diabetol. 2025;24(1):61.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZucchini S, Fabi M, Maltoni G, Zioutas M, Trevisani V, Di Natale V, et al. Adolescents with severe obesity show a higher cardiovascular (CV) risk than those with type 1 diabetes: a study with skin advanced glycation end products and intima media thickness evaluation. Acta Diabetol. 2020;57(11):1297\u0026ndash;305.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSłomiński B, Jankowiak M, Maciejewska A, Studziński M, Mączyńska A, Skrzypkowska M, et al. Genetic variation in C-reactive protein (CRP) gene is associated with retinopathy and hypertension in adolescents with type 1 diabetes. Cytokine. 2022;160:156025.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eL\u0026oacute;pez-Melgar B, Fern\u0026aacute;ndez-Friera L, Oliva B, Garc\u0026iacute;a-Ruiz JM, S\u0026aacute;nchez-Cabo F, Bueno H, et al. Short-Term Progression of Multiterritorial Subclinical Atherosclerosis. J Am Coll Cardiol. 2020;75(14):1617\u0026ndash;27.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSarkar S, Prasanna VS, Das P, Suzuki H, Fujihara K, Kodama S, et al. The onset and the development of cardiometabolic aging: an insight into the underlying mechanisms. Front Pharmacol. 2024;15:1447890.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYohena S, Penas-Steinhardt A, Muller C, Faccinetti NI, Cerrone GE, Lovecchio S, et al. Immunological and clinical characteristics of latent autoimmune diabetes in the elderly. Diabetes Metab Res Rev. 2019;35(5):e3137.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJones AG, McDonald TJ, Shields BM, Hagopian W, Hattersley AT. Latent Autoimmune Diabetes of Adults (LADA) Is Likely to Represent a Mixed Population of Autoimmune (Type 1) and Nonautoimmune (Type 2) Diabetes. Diabetes Care. 2021;44(6):1243\u0026ndash;51.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChaillous L, Bouhanick B, Kerlan V, Mathieu E, Lecomte P, Ducluzeau PH, et al. Clinical and metabolic characteristics of patients with latent autoimmune diabetes in adults (LADA): absence of rapid beta-cell loss in patients with tight metabolic control. Diabetes Metab. 2010;36(1):64\u0026ndash;70.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang Y, Li L, Li Y, Liu M, Gan G, Zhou Y, et al. The Impact of Dietary Diversity, Lifestyle, and Blood Lipids on Carotid Atherosclerosis: A Cross-Sectional Study. Nutrients. 2022;14(4):815.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHigashi Y. Smoking cessation and vascular endothelial function. Hypertens Res. 2023;46(12):2670\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFerreira J, Cunha P, Carneiro A, Vila I, Cunha C, Silva C, et al. Is Obesity a Risk Factor for Carotid Atherosclerotic Disease?-Opportunistic Review. J Cardiovasc Dev Dis. 2022;9(5):162.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIshizaka N, Ishizaka Y, Nagai R, Toda E, Hashimoto H, Yamakado M. Association between serum albumin, carotid atherosclerosis, and metabolic syndrome in Japanese individuals. Atherosclerosis. 2007;193(2):373\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFeig JE, Hewing B, Smith JD, Hazen SL, Fisher EA. High-density lipoprotein and atherosclerosis regression: evidence from preclinical and clinical studies. Circ Res. 2014;114(1):205\u0026ndash;13.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZanetti D, Gustafsson S, Assimes TL, Ingelsson E. Comprehensive Investigation of Circulating Biomarkers and Their Causal Role in Atherosclerosis-Related Risk Factors and Clinical Events. Circ Genom Precis Med. 2020;13(6):e002996.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDu M, Li S, Jiang J, Ma X, Liu L, Wang T, et al. Advances in the Pathogenesis and Treatment Strategies for Type 1 Diabetes Mellitus. Int Immunopharmacol. 2025;148:114185.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJun JE, Kang H, Hwang YC, Ahn KJ, Chung HY, Jeong IK. The association between lipoprotein (a) and carotid atherosclerosis in patients with type 2 diabetes without pre-existing cardiovascular disease: A cross-sectional study. Diabetes Res Clin Pract. 2021;171:108622.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVouillarmet J, Marsot C, Maucort-Boulch D, Riche B, Helfre M, Grange C. Vascular Events and Carotid Atherosclerosis: A 5-Year Prospective Cohort Study in Patients with Type 2 Diabetes and a Contemporary Cardiovascular Prevention Treatment. J Diabetes Res. 2019;2019:9059761.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXing L, Li R, Zhang S, Li D, Dong B, Zhou H, et al. High Burden of Carotid Atherosclerosis in Rural Northeast China: A Population-Based Study. Front Neurol. 2021;12:597992.\u003c/span\u003e\u003c/li\u003e\u003c/ol\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":"bmc-cardiovascular-disorders","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bcar","sideBox":"Learn more about [BMC Cardiovascular Disorders](http://bmccardiovascdisord.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bcar/default.aspx","title":"BMC Cardiovascular Disorders","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Carotid Atherosclerosis, Machine Learning, Latent Autoimmune Diabetes in Adults","lastPublishedDoi":"10.21203/rs.3.rs-6131962/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6131962/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eBackground:\u003c/strong\u003e Latent autoimmune diabetes in adults (LADA) is a slowly progressing form of diabetes with autoimmune origins. Patients with LADA are at an elevated risk of developing cardiovascular diseases, including carotid atherosclerosis. While machine learning models have been widely used in predicting cardiovascular risks in Type 1 and Type 2 diabetes, research on LADA remains limited. Early prediction of carotid atherosclerosis using machine learning models could help in timely intervention and improved patient outcomes for this specific population.\u003cstrong\u003e\u003cbr\u003e\n Methods:\u003c/strong\u003e We conducted a retrospective cross-sectional analysis involving 142 LADA patients diagnosed within the endocrinology department at Shanxi Bethune Hospital, China. Various clinical, demographic, and laboratory variables were analyzed using univariate and multivariate logistic regression, complemented by LASSO regression for feature selection. Additionally, eight machine learning algorithms—logistic regression (LR), decision tree (DT), random forests (RF), k-nearest neighbors (KNN), support vector machine (SVM), neural networks (NNET), eXtreme gradient boosting (XGBoost), and light gradient boosting machine (LightGBM)—were employed to predict carotid atherosclerosis.\u003cstrong\u003e\u003cbr\u003e\n Results:\u003c/strong\u003e Significant risk factors for carotid atherosclerosis were identified, including age, smoking history, BMI, ALB, HDL-C, and ALT. Among the various machine learning models evaluated, the LR model exhibited the highest performance, achieving an area under the curve (AUC) of 0.936, alongside an accuracy of 86%. NNET and SVM models also demonstrated robust predictive capacities with AUC values of 0.919 and 0.918, respectively.\u003cstrong\u003e\u003cbr\u003e\n Conclusions:\u003c/strong\u003e This study highlights the critical role of identifying risk factors for carotid atherosclerosis in LADA patients. Our use of ML models builds on the growing body of work in diabetes-related cardiovascular risk prediction, and it offers a novel approach by specifically targeting the LADA population. Incorporating ML models into clinical practice could improve risk stratification and patient management in LADA. Future research should validate these models across diverse populations and investigate the underlying mechanisms linking LADA to cardiovascular risk.\u003c/p\u003e","manuscriptTitle":"Predicting Carotid Atherosclerosis in Latent Autoimmune Diabetes in Adult Patients Using Machine Learning Models: A Retrospective Study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-04-14 23:12:22","doi":"10.21203/rs.3.rs-6131962/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-04-15T05:46:46+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-04-15T05:26:06+00:00","index":"","fulltext":""},{"type":"reviewerAgreed","content":"250701690398138055927642473961971290395","date":"2025-04-13T12:31:29+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-04-08T10:09:33+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"11610062968092861428788842163932724076","date":"2025-04-08T10:00:52+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-04-08T07:33:05+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"59873530625122799165364249117798581777","date":"2025-04-08T05:04:22+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-04-08T00:59:32+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-04-07T11:46:30+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Cardiovascular Disorders","date":"2025-04-05T14:11:16+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-cardiovascular-disorders","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bcar","sideBox":"Learn more about [BMC Cardiovascular Disorders](http://bmccardiovascdisord.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bcar/default.aspx","title":"BMC Cardiovascular Disorders","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"c44c3a18-1fad-4db0-858f-71e6c7fc037c","owner":[],"postedDate":"April 14th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2025-07-07T16:09:00+00:00","versionOfRecord":{"articleIdentity":"rs-6131962","link":"https://doi.org/10.1186/s12872-025-04786-6","journal":{"identity":"bmc-cardiovascular-disorders","isVorOnly":false,"title":"BMC Cardiovascular Disorders"},"publishedOn":"2025-07-03 15:58:46","publishedOnDateReadable":"July 3rd, 2025"},"versionCreatedAt":"2025-04-14 23:12:22","video":"","vorDoi":"10.1186/s12872-025-04786-6","vorDoiUrl":"https://doi.org/10.1186/s12872-025-04786-6","workflowStages":[]},"version":"v1","identity":"rs-6131962","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6131962","identity":"rs-6131962","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}