Gao Ying, Zhang Meihui, Xu Chen, Wu Birong, Qiao Shan, Cai Yong, Hu Fan
Public Health Research Center, Tongren Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China.
Public Health Research Center, Tongren Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China; School of Public Health, Shanghai Jiao Tong University School of Medicine, Shanghai, China.
Diabetes Res Clin Pract. 2025 Sep;227:112414. doi: 10.1016/j.diabres.2025.112414. Epub 2025 Aug 7.
The triglyceride-glucose (TyG) index is a recognized surrogate marker for insulin resistance. This study explores the relationship between the baseline TyG index and the subsequent risk of stroke in middle-aged and older adults, while also examining the variable importance of various predictors potentially influencing stroke incidence.
We included 6863 participants from the China Health and Retirement Longitudinal Study (CHARLS) who had no history of stroke at the start of the study. To identify important predictors, we employed the Least Absolute Shrinkage and Selection Operator (Lasso) Cox regression model, followed by a multivariate Cox proportional hazards model to analyze the association between the TyG index and future stroke incidence. Subgroup analyses by age and gender were conducted. The significance of different predictors was assessed using explainable survival machine learning models that accounted for temporal changes.
Over a 9-year follow-up, 787 participants (11.5 %) experienced a first stroke. The baseline TyG index had an inverted U-shaped relationship with stroke risk. After adjustment for confounders, participants in the second, third, and highest quartiles of the baseline TyG index showed a higher stroke risk compared to those in the lowest quartile (P < 0.01), with adjusted hazard ratios (HR) [95 % confidence intervals (CI)] of 1.45 (1.16-1.82), 1.64 (1.31-2.04), and 1.36 (1.08-1.72), respectively. These associations were consistent across all subgroups except for individuals younger than 60 years. Notably, age emerged as the most significant predictor of stroke risk in the explainable machine learning analysis, with the TyG index also identified as a relatively important factor.
This research employs explainable machine learning to delineate factors that contribute to stroke risk, highlighting how the TyG index's impact on stroke risk varies by age and gender. As an established surrogate marker for insulin resistance, the TyG index monitoring may play a crucial role in stroke prevention and management strategies.
甘油三酯-葡萄糖（TyG）指数是公认的胰岛素抵抗替代指标。本研究探讨中年及老年人群基线TyG指数与随后发生中风风险之间的关系，同时考察各种潜在影响中风发病率的预测因素的可变重要性。
我们纳入了中国健康与养老追踪调查（CHARLS）中的6863名参与者，这些参与者在研究开始时无中风病史。为了确定重要的预测因素，我们采用了最小绝对收缩和选择算子（Lasso）Cox回归模型，随后使用多变量Cox比例风险模型分析TyG指数与未来中风发病率之间的关联。进行了按年龄和性别的亚组分析。使用考虑时间变化的可解释生存机器学习模型评估不同预测因素的显著性。
在9年的随访中，787名参与者（11.5%）发生了首次中风。基线TyG指数与中风风险呈倒U形关系。在调整混杂因素后，基线TyG指数处于第二、第三和最高四分位数的参与者与最低四分位数的参与者相比，中风风险更高（P<0.01），调整后的风险比（HR）[95%置信区间（CI）]分别为1.45（1.16-1.82）、1.64（1.31-2.04）和1.36（1.08-1.72）。除60岁以下个体外，所有亚组中的这些关联均一致。值得注意的是，在可解释的机器学习分析中，年龄是中风风险最显著的预测因素，TyG指数也被确定为一个相对重要的因素。
本研究采用可解释的机器学习来描绘导致中风风险的因素，突出了TyG指数对中风风险的影响如何因年龄和性别而异。作为胰岛素抵抗的既定替代指标，TyG指数监测可能在中风预防和管理策略中发挥关键作用。
