The joint association of the combined triglyceride-glucose index and atherogenic index of plasma with hypertension on stroke risk across different glycemic status: a prospective cohort study

Fig. 1. Flowchart of inclusion and exclusion criteria for this study

Fig. 2. Identification and Characterization of TyG-AIP Trajectory Clusters. This figure illustrates the identification and characterization of TyG-AIP trajectory clusters across three glycemia strata: total population (A–C), prediabetes (PDM, D-F), and normoglycemia (NDM, G-I). A, D, G The elbow method plots used to determine the optimal number of clusters (K) for K-means clustering within each stratum. B, E, H Scatter plots presenting the K-means clustering results, with TyG-AIP in 2012 plotted against 2015, and points colored by the assigned trajectory cluster. C, F, I Line plots depicting the mean TyG-AIP changes from 2012 to 2015 for each identified trajectory cluster within the respective strata. Three distinct trajectory patterns were identified across all strata: “Low level gradual decline,” “Medium level rapid ascent,” and “High level rapid decline.” The average silhouette coefficient for the overall clustering solution was 0.479. T2D, type 2 diabetes; PDM, prediabetes; NDM, newly diagnosed diabetes

Fig. 3. Results of TyG-AIP and Cumulative Incidence of Stroke Stratified by Sex, Age, Hypertension, Smoking, Alcohol Consumption, and BMI

Fig. 4. TyG, AIP and TyG-AIP as continuous and categorical variables with stroke risk in the total population, individuals with dysglycemia (PDM), and those with normoglycemia (NDM)

Fig. 5. The dose–response relationship between TyG, AIP, TyG-AIP and stroke risk in the total population, individuals with dysglycemia (PDM), and those with normoglycemia (NDM). Notably, the RCS curves for TyG–AIP, TyG, and AIP exhibit highly similar shapes. This reflects that the TyG–AIP index successfully integrates the core metabolic risk signal related to stroke shared by its components. The incremental value of this composite index lies primarily in its clinical convenience as a unified metric and its superior predictive performance in multivariable-adjusted models (see Table 2 and Supplementary tables)

Fig. 6. TyG-AIP index + hypertension and TyG-AIP index trajectory + hypertension with stroke risk in the total population, individuals with dysglycemia (PDM), and those with normoglycemia (NDM). The figure displays all groups relative to a common reference (Q1). For direct comparisons assessing the specific contributions of hypertension and elevated TyG-AIP within strata (e.g., Q4 vs. Q2; Q4 vs. Q3), please refer to Supplementary Tables 13, 14

Fig. 7. Superior predictive performance of a model combining TyG-AIP and hypertension for stroke risk. Receiver operating characteristic (ROC) curves compare three models: hypertension alone (blue), TyG-AIP index alone (green), and a combined model including both predictors (red). The area under the curve (AUC) for each model is shown in the legend. The combined model demonstrates the highest discriminatory ability, justifying the integration of both risk factors to improve prediction accuracy