Scholarlink

Background <p>Increased arterial stiffness is a high-risk factor for cardiovascular diseases, making its early identification crucial. Based on data from an apparently healthy community-based population, this study employed machine learning algorithms to screen for factors influencing early arterial stiffness and constructed a risk prediction model for its development.</p> Methods <p>In this cross-sectional study, 2948 community participants were enrolled between June and December 2024. Twenty-four blood-derived indices spanning metabolic, lipid, and inflammatory domains were calculated. Least absolute shrinkage and selection operator (LASSO) regression was used to select and adjust for significant confounders among baseline clinical variables. Univariate and multivariate logistic regression analyses were then performed to identify indicators independently associated with early increased arterial stiffness, defined as a brachial-ankle pulse wave velocity (baPWV) ≥ 1400&#xa0;cm/s. A random forest model, combined with SHapley Additive exPlanations (SHAP) analysis, was employed to evaluate and rank the predictive importance of all candidate variables. Key predictors identified through these analyses were integrated to build a multivariate logistic regression model, visualized as a nomogram. Model performance was evaluated by discrimination, calibration, and decision curve analysis (DCA).</p> Results <p>Early increased arterial stiffness was identified in 1636 participants (55.5%). Multivariate analysis identified nine independent risk factors, including the monocyte-to-high-density lipoprotein cholesterol ratio (MHR), Castelli’s Risk Index I and II (CRI- I, CRI-II), atherogenic index (AI), TyG-waist circumference (TyG-WC), TyG-waist-to-height ratio (TyG-WHtR), neutrophil-to-HDL ratio (NHR), platelet-to-HDL ratio (PHR), and uric acid-to-HDL ratio (UHR). The random forest model highlighted age, systolic and diastolic blood pressure, TyG-WHtR, heart rate, fasting blood glucose, and glycated hemoglobin as top predictors. SHAP analysis confirmed the substantial contribution of TyG-WHtR. The resulting nomogram demonstrated excellent discrimination in the test set (AUC = 0.877, 95% CI: 0.865–0.889) and good calibration. A web-based calculator was developed for individualized risk estimation.</p> Conclusions <p>This study developed and internally validated a prediction model incorporating seven routine clinical indicators for assessing arterial stiffness risk in a community population. TyG-WHtR emerged as a key independent predictor. The model and its visual tools offer a practical means for early community-based screening of arterial stiffness.</p> Trial registration <p>Clinical cohort registration number：ChiCTR2300068117</p> <p>Registration date: 2023-02-07</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A machine learning model based on routine blood-derived indices for early arterial stiffness prediction in the community

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月23日

查看全文
  • Liuting Zheng,
  • Xiao Feng,
  • Keyun Wang,
  • Yue Niu,
  • Sifan Yang,
  • Junqian Wang,
  • Huanrong Li,
  • Luying Qiao,
  • Wang Lu,
  • Shuang Li,
  • Huidi Xie,
  • Tian Li,
  • Yuzhe Zhang,
  • Yajia Sun,
  • Xiaoxu Zhang,
  • Ying Zheng,
  • Peng Wang,
  • Jing Huang,
  • Li Zhang,
  • Xuefeng Sun,
  • Weiguang Zhang,
  • Zhe Feng,
  • Xiangmei Chen

摘要

Background

Increased arterial stiffness is a high-risk factor for cardiovascular diseases, making its early identification crucial. Based on data from an apparently healthy community-based population, this study employed machine learning algorithms to screen for factors influencing early arterial stiffness and constructed a risk prediction model for its development.

Methods

In this cross-sectional study, 2948 community participants were enrolled between June and December 2024. Twenty-four blood-derived indices spanning metabolic, lipid, and inflammatory domains were calculated. Least absolute shrinkage and selection operator (LASSO) regression was used to select and adjust for significant confounders among baseline clinical variables. Univariate and multivariate logistic regression analyses were then performed to identify indicators independently associated with early increased arterial stiffness, defined as a brachial-ankle pulse wave velocity (baPWV) ≥ 1400 cm/s. A random forest model, combined with SHapley Additive exPlanations (SHAP) analysis, was employed to evaluate and rank the predictive importance of all candidate variables. Key predictors identified through these analyses were integrated to build a multivariate logistic regression model, visualized as a nomogram. Model performance was evaluated by discrimination, calibration, and decision curve analysis (DCA).

Results

Early increased arterial stiffness was identified in 1636 participants (55.5%). Multivariate analysis identified nine independent risk factors, including the monocyte-to-high-density lipoprotein cholesterol ratio (MHR), Castelli’s Risk Index I and II (CRI- I, CRI-II), atherogenic index (AI), TyG-waist circumference (TyG-WC), TyG-waist-to-height ratio (TyG-WHtR), neutrophil-to-HDL ratio (NHR), platelet-to-HDL ratio (PHR), and uric acid-to-HDL ratio (UHR). The random forest model highlighted age, systolic and diastolic blood pressure, TyG-WHtR, heart rate, fasting blood glucose, and glycated hemoglobin as top predictors. SHAP analysis confirmed the substantial contribution of TyG-WHtR. The resulting nomogram demonstrated excellent discrimination in the test set (AUC = 0.877, 95% CI: 0.865–0.889) and good calibration. A web-based calculator was developed for individualized risk estimation.

Conclusions

This study developed and internally validated a prediction model incorporating seven routine clinical indicators for assessing arterial stiffness risk in a community population. TyG-WHtR emerged as a key independent predictor. The model and its visual tools offer a practical means for early community-based screening of arterial stiffness.

Trial registration

Clinical cohort registration number：ChiCTR2300068117

Registration date: 2023-02-07