Cardiovascular

Machine learning-based prediction model for predicting the impact of insulin resistance on the risk of ischemic cardiomyopathy.

TL;DR

The TyG-BMI and TC/HDL-C ratio independently predict ischemic cardiomyopathy risk, with the XGBoost model identified as the most effective for ICM risk prediction, indicating substantial clinical applicability.

Key Findings

TyG-BMI index was identified as an independent risk factor for ischemic cardiomyopathy (ICM).

  • TyG-BMI was selected through univariable logistic regression (P < 0.05) and confirmed via multivariable logistic regression analysis.
  • TyG-BMI is a composite measure combining triglyceride glucose index with body mass index, used as a reliable indicator of insulin resistance.
  • The TG/HDL-C ratio was also investigated as an insulin resistance indicator but TyG-BMI emerged as the key independent predictor.
  • The study included 1,603 total subjects.

TC/HDL-C ratio was identified as an independent predictor of ICM risk.

  • TC/HDL-C was confirmed as an independent risk factor through both univariable and multivariable logistic regression analyses (P < 0.05).
  • TC/HDL-C was included among the variables used to construct the eight machine learning models.
  • Total cholesterol (TC) alone was also identified as an independent risk factor for ICM.
  • HDL-C was independently associated with ICM risk as well.

Eleven independent risk factors for ICM were identified through logistic regression analyses.

  • The independent risk factors identified were: TyG-BMI, age, ejection fraction, TC/HDL-C, sex, HDL-C, TC, BMI, hemoglobin, diabetes, and hypertension.
  • All variables had P < 0.05 in both univariable and multivariable logistic regression analyses.
  • Variables with P < 0.05 in univariable analysis were selected for entry into multivariable logistic regression.
  • These 11 variables were subsequently used to construct eight machine learning models.

The XGBoost (XGB) model was identified as the optimal machine learning model for ICM risk prediction among eight models evaluated.

  • Eight machine learning models were developed and compared using the 11 identified independent risk factors.
  • The XGB model was selected as the best-performing model based on its performance metrics.
  • SHAP (SHapley Additive exPlanations) values were visualized using the XGB model to interpret feature contributions.
  • An online calculator was developed based on the XGB model for practical clinical use.

The XGB model demonstrated strong calibration and clinical utility as validated by calibration plot and decision curve analysis (DCA).

  • The calibration plot indicated 'strong alignment between the model's predicted and actual values.'
  • Decision curve analysis (DCA) demonstrated the model's clinical utility.
  • Model validation included both calibration assessment and DCA to confirm practical applicability.
  • The study involved 1,603 subjects for model development and validation.

An online calculator was developed based on the optimal XGB model to facilitate clinical application of ICM risk prediction.

  • The online calculator incorporates the 11 identified independent risk factors: TyG-BMI, age, ejection fraction, TC/HDL-C, sex, HDL-C, TC, BMI, hemoglobin, diabetes, and hypertension.
  • SHAP value visualization was used alongside the online calculator to enhance interpretability of the model.
  • The tool was designed to support clinical decision-making for ICM risk assessment.
  • The model was characterized as having 'substantial clinical applicability.'

What This Means

This research suggests that insulin resistance — measured using a calculated index called TyG-BMI (which combines triglyceride, glucose, and body mass index values) — is a meaningful predictor of ischemic cardiomyopathy (ICM), a form of heart disease caused by reduced blood flow to the heart muscle. Researchers studied 1,603 patients and used statistical methods to identify 11 independent risk factors for ICM, including TyG-BMI, age, heart pumping function (ejection fraction), cholesterol ratios, sex, individual cholesterol measures, body mass index, hemoglobin levels, diabetes, and hypertension. They then used these factors to build and compare eight different machine learning (AI-based) prediction models. The XGBoost model outperformed the other seven models and was selected as the best tool for predicting ICM risk. The researchers validated this model using calibration plots (showing its predictions matched real-world outcomes) and decision curve analysis (showing it provides clinically meaningful guidance). To make the model accessible to healthcare providers, they also developed an online calculator and used a technique called SHAP values to visually explain which factors most strongly influenced each prediction. This research suggests that relatively simple blood tests and measurements — particularly those reflecting insulin resistance and cholesterol metabolism — could be combined into an accessible tool to help identify patients at elevated risk for ischemic cardiomyopathy. If validated in broader populations, such a tool could support earlier intervention and more targeted monitoring for at-risk individuals.

Have a question about this study?

Citation

Naman T, Cheng H, Yu X, Guo Z. (2026). Machine learning-based prediction model for predicting the impact of insulin resistance on the risk of ischemic cardiomyopathy.. BMC medical informatics and decision making. https://doi.org/10.1186/s12911-026-03693-w