Association of the triglyceride-glucose series indices with hypertension and pre-hypertension in Anhui China

Insulin resistance (IR) and hypertension have a pivotal part in the pathophysiology of cardiovascular disease. The triglyceride–glucose (TyG) index and its derivatives—TyG–body mass index (BMI), TyG–waist circumference (WC), and TyG–WC-to-height ratio (WtHR)—have appeared as simple, cost-effective substitutes for evaluating IR. This research seeks to explore the associations between TyG combined with these obesity indicators and hypertension and pre-hypertension. Data of 9,132 participants in Bengbu (Anhui Province, China) 2019 was collected. The dose–response relationship between these parameters and hypertension or pre-hypertension were analyzed utilizing restricted cubic spline models with multiplex logistic regression (LR). The performance of these parameters in predicting hypertension or pre-hypertension was evaluated using LR and XGBoost model. LR analysis revealed that the TyG index and its derivatives were significantly associated with an increased risk of hypertension and pre-hypertension ( P < 0.05). Notably, the TyG-WtHR index demonstrated superior potential predictive capacity (AUC = 0.631, OR = 3.133, 95% CI = 2.688–3.652). Dose–response relationships were predominantly nonlinear ( P for nonlinear < 0.05), with some linear trends and sex-based differences. The XGBoost model outperformed the traditional LR model (AUC: LR—0.617, XGBoost—0.662), identifying the TyG–WC and TyG–WtHR indices as potential risk factors of hypertension. Exploratory interaction analysis shows that, high meat consumption was associated with higher hypertension likelihood in individuals with elevated TyG values. The TyG index and its derivatives, are contributive, sex-specific factors of hypertension in middle-aged and elder participants. Our study finds the potential of metabolic monitoring plus a personalized diet for intervening hypertension in at-risk individuals, the parameters can be integrated into machine learning models to augment facilitating early detection.