Prognostic value of acute-to-chronic glycaemic ratio in EVT-treated large vessel occlusion stroke according to diabetes status: a machine learning-augmented analysis

The acute-to-chronic glycemic ratio (ACR) has emerged as a prognostic marker in acute ischemic stroke, but its predictive value in patients with large vessel occlusion (LVO) undergoing endovascular treatment (EVT) and the potential modification by diabetes remain unclear. We evaluated the association between ACR and clinical outcomes in EVT-treated LVO patients and explored the role of diabetes. We performed a post hoc analysis of 441 EVT-treated patients from the DETECT2-China registry. ACR was calculated as admission glucose divided by HbA1c-derived estimated average glucose and dichotomized by the median. Primary outcome was poor functional outcome at 90 days (mRS 3–6). Multivariable logistic regression, subgroup analyses, and ROC comparisons were conducted. SHapley Additive exPlanations (SHAP) were applied to explore non-linear dynamics and interactions. A total of 441 patients were included. Multivariable analysis showed that high ACR was independently associated with poor functional outcome at 90 days (adjusted odds ratio [aOR] 1.53, 95% CI 1.01–2.33, P = 0.047), early neurological deterioration (END) (aOR 1.99, 95% CI 1.14–3.50, P = 0.016), and 90-day mortality (aOR 1.47, 95% CI 1.01–2.12, P = 0.044). Subgroup analysis revealed a numerical trend toward a stronger association between high ACR and poor functional outcome in patients with diabetes (aOR 2.85, 95% CI 1.20–6.77, P = 0.017) than in those without diabetes (aOR 1.36, 95% CI 0.84–2.20, P = 0.209), although the interaction was not statistically significant ( P for interaction = 0.094). Adding ACR to the baseline clinical model significantly improved the area under the curve (AUC) for predicting poor outcome from 0.730 to 0.747 (DeLong P = 0.035), while adding absolute glucose yielded no significant improvement. Exploratory SHAP analysis confirmed a non-linear, sharp escalation in the risk contribution of ACR, and at high ACR levels, diabetic patients exhibited a significantly higher risk contribution compared to non-diabetic patients. Outperforming absolute glucose, elevated ACR independently predicts 90-day poor functional outcome, early neurological deterioration, and mortality. These exploratory subgroup observations suggest a potential differential sensitivity to relative stress hyperglycemia according to chronic glycemic background, warranting hypothesis-generating interpretation.