Scholarlink

Background <p>Insulin resistance indices, lipid metabolism indices and obesity-related indices are emerging predictors of dyslipidaemia, but the relationship with dyslipidaemia remains unclear. Therefore, this study aimed to explore the potential correlation between these metrics and dyslipidaemia, as well as to develop machine learning models to predict key metrics.</p> Methods <p>In this prospective cohort study, 3,144 baseline study participants were included to assess the association of insulin resistance, lipid metabolism and obesity indicators with dyslipidemia by logistic regression and restricted cubic spline (RCS). After excluding participants who were diagnosed with dyslipidemia at baseline, the final study population included in the longitudinal analysis was 1,033. To compare the predictive ability of novel insulin resistance, lipid metabolism, and obesity indices for the risk of developing dyslipidemia in adults and to confirm the predictive value of optimal markers using Cox proportional risk regression modeling, RCS, and through the construction of multiple machine learning models.</p> Results <p>At a median follow-up duration of 1,874 days, 468 participants were diagnosed with dyslipidaemia. In the cross-sectional analysis, ABSI, CI, TyG, TyG-BMI, TyG-WC, TyG-WHtR, METS-IR, and TG/HDL-C were all positively associated with the risk of dyslipidaemia (<i>P</i> for trend &lt; 0.05). In longitudinal analysis, TyG, TyG-BMI, TyG-WC, TyG-WHtR, METS-IR, and TG/HDL-C were positively associated with the risk of new-onset dyslipidaemia (<i>P</i> for trend &lt; 0.05). Further RCS revealed that TyG-WHtR and METS-IR exhibited stable positive linear associations in both analyses (<i>P</i>-<i>nonlinear</i> &gt; 0.05), whereas TyG and TG/HDL-C consistently demonstrated non-linear relationships (<i>P</i>-<i>nonlinear</i> &lt; 0.05). The association patterns of TyG-WC was non-linear in the baseline analysis but shifted to linear in the longitudinal analysis. The association patterns of ABSI, CI, and TyG-BMI were linear in the baseline analysis but shifted to non-linear in the longitudinal analysis. The support vector machine model demonstrated optimal performance in predicting dyslipidaemia, with area under the receiver operating characteristic curve (AUC) values of 0.870 (95% CI: 0.845–0.896, <i>P</i> &lt; 0.001) for the training set and 0.791 (95% CI: 0.740–0.842, <i>P</i> &lt; 0.001) for the testing set. In addition, feature importance assessment and Shapley additive interpretation (SHAP) analysis further confirmed METS-IR as the best predictor of dyslipidemia with an importance score of 0.105.</p> Conclusion <p>METS-IR was the best predictors of the risk of developing dyslipidaemia, respectively, but further confirmation of these correlations through external validation is needed.</p> Clinical trial number <p>Not applicable.</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

Comparison of the predictive power of novel insulin resistance, lipid metabolism and obesity indices for the risk of developing dyslipidaemia in adults

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月6日

查看全文
  • Yujian Liang,
  • Jiansheng Cai,
  • Xiaoting Mo,
  • Qiumei Liu,
  • Jiahui Rong,
  • Linhai Zhao,
  • Lei Luo,
  • Tiantian Zhang,
  • Songju Wu,
  • Wenjia Jin,
  • Qinyi Guan,
  • Kaisheng Teng,
  • Liuyong Zhong,
  • Xiaolin Li,
  • Pengfeng Chen,
  • Tian Qiu,
  • Shuzhen Liu,
  • Jian Qin,
  • Zhiyong Zhang

摘要

Background

Insulin resistance indices, lipid metabolism indices and obesity-related indices are emerging predictors of dyslipidaemia, but the relationship with dyslipidaemia remains unclear. Therefore, this study aimed to explore the potential correlation between these metrics and dyslipidaemia, as well as to develop machine learning models to predict key metrics.

Methods

In this prospective cohort study, 3,144 baseline study participants were included to assess the association of insulin resistance, lipid metabolism and obesity indicators with dyslipidemia by logistic regression and restricted cubic spline (RCS). After excluding participants who were diagnosed with dyslipidemia at baseline, the final study population included in the longitudinal analysis was 1,033. To compare the predictive ability of novel insulin resistance, lipid metabolism, and obesity indices for the risk of developing dyslipidemia in adults and to confirm the predictive value of optimal markers using Cox proportional risk regression modeling, RCS, and through the construction of multiple machine learning models.

Results

At a median follow-up duration of 1,874 days, 468 participants were diagnosed with dyslipidaemia. In the cross-sectional analysis, ABSI, CI, TyG, TyG-BMI, TyG-WC, TyG-WHtR, METS-IR, and TG/HDL-C were all positively associated with the risk of dyslipidaemia (P for trend < 0.05). In longitudinal analysis, TyG, TyG-BMI, TyG-WC, TyG-WHtR, METS-IR, and TG/HDL-C were positively associated with the risk of new-onset dyslipidaemia (P for trend < 0.05). Further RCS revealed that TyG-WHtR and METS-IR exhibited stable positive linear associations in both analyses (P-nonlinear > 0.05), whereas TyG and TG/HDL-C consistently demonstrated non-linear relationships (P-nonlinear < 0.05). The association patterns of TyG-WC was non-linear in the baseline analysis but shifted to linear in the longitudinal analysis. The association patterns of ABSI, CI, and TyG-BMI were linear in the baseline analysis but shifted to non-linear in the longitudinal analysis. The support vector machine model demonstrated optimal performance in predicting dyslipidaemia, with area under the receiver operating characteristic curve (AUC) values of 0.870 (95% CI: 0.845–0.896, P < 0.001) for the training set and 0.791 (95% CI: 0.740–0.842, P < 0.001) for the testing set. In addition, feature importance assessment and Shapley additive interpretation (SHAP) analysis further confirmed METS-IR as the best predictor of dyslipidemia with an importance score of 0.105.

Conclusion

METS-IR was the best predictors of the risk of developing dyslipidaemia, respectively, but further confirmation of these correlations through external validation is needed.

Clinical trial number

Not applicable.