Abstract
BACKGROUND: Stroke is a leading contributor to the high disease burden in low- and middle-income countries, making early prevention essential for risk reduction. The triglyceride-glucose (TyG) index and the estimated glucose disposal rate (eGDR) are surrogate markers of insulin resistance, and both have been associated with cardiometabolic risk. However, the predictive utility of their combined use for stroke risk across different glucose metabolism states remains unclear. METHODS: This prospective cohort study enrolled 5789 stroke-free participants aged ≥ 45 years, including 3490 with abnormal glucose regulation(AGR) and 2296 with normal glucose regulation(NGR). The TyG-eGDR index was dichotomised as TyG × eGDR. The TyG + eGDR combination was further partitioned into 12 strata defined by TyG tertiles crossed with eGDR quartiles. We evaluated the relationship between theseTyG, eGDR, TyG-eGDR and TyG + eGDR categories and incident stroke using cox proportional-hazards models, mediation analysis, restricted cubic splines, receiver-operating-characteristic curves, and subgroup analyses Query. RESULTS: Among 5789 participants followed for incident stroke, 460 new events were documented, corresponding to an overall incidence of 9.93 per 1000 person-years. We observed a negative association between lower eGDR index (OR = 0.70, 95% CI 0.62-0.79) and lower TyG-eGDR index (OR = 0.72, 95% CI 0.63-0.81) with stroke risk across all populations, while in the AGR subgroup, a higher TyG index (OR = 1.24, 95% CI 1.11-1.37) was associated with a higher risk of stroke. Compared to the reference group (low TyG and high eGDR), the combination of high TyG (≥ 8.90) and low eGDR (< 7.09) conferred the highest stroke hazard (HR = 3.75, 95% CI 2.29-6.15, p = 1.6 × 10⁻⁷), and this pattern was consistent in both AGR and NGR sub-cohorts. However, the TyG + eGDR combination exerted a stronger influence on stroke risk within the AGR group. RCS showed non-linear dose-response for TyG and TyG-eGDR with stroke (p < 0.05), but linear for eGDR. The TyG + eGDR composite model achieved superior discrimination (area under the ROC curve: 0.71). Mediation analysis indicated that eGDR accounted for 29.99% of the association between TyG and stroke risk, while TyG explained 5.81% of the association between eGDR and stroke risk; comparable mediation patterns were observed in the AGR population but not in the NGR population. CONCLUSIONS: Low eGDR and a low TyG-eGDR product independently predict a higher stroke risk, whereas elevated TyG predominantly increases stroke risk in individuals with AGR. Combining TyG and eGDR markedly improves stroke discrimination beyond either metric alone. Routine integration of these two inexpensive indices permits earlier, more precise risk stratification and can inform targeted interventions that substantially reduce stroke incidence.