The value of triglyceride-glucose index-related indices in evaluating migraine: perspectives from multi-centre cross-sectional studies and machine learning models.

Background: This study employed representative data from the U.S. and China to delve into the correlation among migraine prevalence, the triglyceride‒glucose index, a marker of insulin resistance, and the composite indicator of obesity.

Methods: Cross-sectional data were acquired from the National Health and Nutrition Examination Survey conducted between 1999 and 2004, as well as from the China Longitudinal Study of Health and Retirement (CHARLS) performed from 2011 to 2012. Weighted logistic regression analysis, subgroup analysis, smooth curve fitting and threshold effect analysis were used to ascertain the intricate relationships among triglyceride glucose-body mass index (TyG-BMI), triglyceride glucose-waist circumference (TyG-WC), triglyceride glucose-waist height ratio (TyG-WHtR) and migraine. Boruta's algorithm and nine machine learning models were applied. SHapley Additive Explanations (SHAP) values were used to analyze leading models, highlighting influential features.

Results: The analysis included 6,204 U.S. participants and 9,401 Chinese participants. TyG-BMI as well as TyG-WHtR were shown to be strongly correlated with the incidence of migraine among U.S. adults (TyG-BMI: OR = 1.28, 95% CI 1.14-1.44, P < 0.001; TyG-WHtR: OR = 1.17, 95% CI 1.09-1.26, P < 0.001). However, this correlation was not detected in Chinese adults. TyG-BMI indicated a strong positive association beyond the threshold of 206, while TyG-WHtR demonstrated a significant positive link below the cutoff of 7.4. In addition, age was an important interaction factor between TyG-BMI and TyG-WHtR and migraine. The XGBoost model showed excellent performance, with higher AUC values for TyG-BMI than for TyG-WHtR (0.929/0.926).

Conclusions: The TyG-BMI, relative to the TyG-WHtR, may provide clinicians with information about patients' insulin sensitivity, thus helping to develop individualized treatment strategies. These findings contribute to population-level health interventions aimed at mitigating metabolic and neurological disease burdens, ensuring healthy lives and promoting well-being.