Scholarlink

Background <p>Diabetes is a common chronic disease, with heavy metals potentially contributing to its development. However, large-scale population-based data on the association between body heavy metals and diabetes is lacking.</p> Objective <p>The purpose of this study is to evaluate the relationship between heavy metals in the body and diabetes.</p> Methods <p>This study utilized individual data from the National Health and Nutrition Examination Survey (NHANES, 2005–2016), focusing on heavy metals in urine and blood, along with other covariates. Nine machine learning models were trained, with random forest (RF) selected as the optimal model based on F1 scores for interpretability analysis. Shapley additive explanations and partial dependence plots were used to examine the interactions between key metal variables and diabetes prediction. Logistic regression further verified the significance of metals in diabetes.</p> Results <p>In the RF model, the area under the receiver operating characteristic curve for the test set was 0.83, with an accuracy of 79.3% and an F1 score of 0.54. Urinary cadmium (Cd) showed a significant positive correlation with diabetes, with increased risk in the range of 0.025–0.501&#xa0;µg/L. Conversely, urine barium (Ba) was significantly negatively correlated, reducing diabetes risk in the range of 0.04–3.82&#xa0;µg/L. Urine thallium demonstrated a reduced probability of diabetes, though not significantly. A synergistic effect between urinary Cd and Ba was observed. Subgroup analysis showed no heterogeneity among subgroups.</p> Conclusion <p>The findings revealed a significant positive correlation between urinary Cd and diabetes, and a significant negative correlation between urine Ba and diabetes.</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

In vivo heavy metal and diabetes association: A cross-sectional interpretable machine learning analysis of NHANES

  • 期刊论文

  • 发表日期：

    2025年7月3日

查看全文
  • Jianan He,
  • Wenhao Zhou,
  • Huanting Zhang,
  • Jie Shen

摘要

Background

Diabetes is a common chronic disease, with heavy metals potentially contributing to its development. However, large-scale population-based data on the association between body heavy metals and diabetes is lacking.

Objective

The purpose of this study is to evaluate the relationship between heavy metals in the body and diabetes.

Methods

This study utilized individual data from the National Health and Nutrition Examination Survey (NHANES, 2005–2016), focusing on heavy metals in urine and blood, along with other covariates. Nine machine learning models were trained, with random forest (RF) selected as the optimal model based on F1 scores for interpretability analysis. Shapley additive explanations and partial dependence plots were used to examine the interactions between key metal variables and diabetes prediction. Logistic regression further verified the significance of metals in diabetes.

Results

In the RF model, the area under the receiver operating characteristic curve for the test set was 0.83, with an accuracy of 79.3% and an F1 score of 0.54. Urinary cadmium (Cd) showed a significant positive correlation with diabetes, with increased risk in the range of 0.025–0.501 µg/L. Conversely, urine barium (Ba) was significantly negatively correlated, reducing diabetes risk in the range of 0.04–3.82 µg/L. Urine thallium demonstrated a reduced probability of diabetes, though not significantly. A synergistic effect between urinary Cd and Ba was observed. Subgroup analysis showed no heterogeneity among subgroups.

Conclusion

The findings revealed a significant positive correlation between urinary Cd and diabetes, and a significant negative correlation between urine Ba and diabetes.