Abstract TP85: Clinical And Imaging Biomarkers Of Ageing And Stroke Severity, And Their Interaction With Clinical Outcomes

Outcomes following endovascular thrombectomy (EVT) for large vessel occlusion stroke remain variable. Clinical predictors include presenting NIHSS and age, surrogate metrics of stroke severity and pre-stroke brain health, respectively. Imaging biomarkers can also predict outcome, but prognostic models typically only include biomarkers of acute stroke severity. This study aimed at exploring how incorporating both imaging biomarkers of stroke severity and brain frailty, automatically derived from routine imaging, impacts outcome prediction following successful EVT. In this single-center retrospective study of 215 patients presenting with anterior circulation stroke, we evaluated prognostic significance of automated analysis outputs from e-Stroke. Acute ischemic volume (AIV) and e-ASPECTS were used to define stroke severity. Cerebral atrophy was used as a marker of brain frailty. Models incorporating imaging biomarkers, clinical features, procedural factors, and combinations of these were evaluated using multivariate regression models predicting mRS 0-2 at 30 and 90-days, NIHSS improvement, mortality and sICH. Incorporation of brain atrophy into models negated the effect of age on patient outcome. The interaction of stroke severity with age was similar to that of infarct volume and brain atrophy. Models incorporating age, presenting NIHSS, AIV, atrophy, reperfusion status and time to recanalization outperformed clinical models at predicting NIHSS improvement (ROC AUC 0.81 and 0.57, respectively) and 90-day mRS. Combinations of imaging and clinical features were superior for predicting any outcome compared with clinical or imaging features in isolation. Models incorporating automated image analysis of acute infarct volume and atrophy outperform models based solely on clinical characteristics for predicting short- and long-term outcome following EVT. Atrophy as a marker of brain frailty is associated with poorer long-term outcomes and negates the effect of age in multivariate models. AIV tempers the effect of intervention on early neurological improvement. Better understanding the relationship between stroke severity and baseline health status may improve patient selection strategies for reperfusion therapies.