A Weighted Cortical ASPECTS Compared Against Total ASPECTS in Outcome Prediction of Anterior Large Vessel Occlusion Thrombectomy
Pratit Patel1, Manisha Koneru2, Mary Penckofer2, Thanh Nguyen3, khalife Jane1, Renato Oliveira4, Mohamad Abdalkader3, Piers Klein5, Nicholas Vigilante6, kamen Scott7, Ajith Thomas1, James Siegler1
1Cooper University Hospital, 2Rowan School of Medicine, 3BMC, 4Cooper university Hospital, 5BU, 6Rowan School of medicine, 7Rowan School of medicineh
Objective:
Compare outcome prediction using Cortical Alberta Stroke Program Early Computed Tomography Score(ASPECTS) against total ASPECTS, in large vessel occlusion strokes undergoing Thrombectomy.
Background:
The 10-point ASPECTS accounts for subcortical structures that may be more likely to infarct and differentially associated with long-term outcomes when compared to injured cortical regions. More specific, cortically-weighted 7-point ASPECTS score could more accurately predict long-term functional outcomes following successful recanalization.
Design/Methods:
Registry data from 2 Comprehensive Stroke Centers(site 1:2016-2021; site 2:2019-2021) including patients with M1 or internal carotid artery(ICA) occlusions who underwent thrombectomy were consolidated for analysis. The primary outcome was a shift in 90-day modified Rankin(mRS), which was adjusted a priori for age, pre-stroke mRS, baseline National Institutes of Health Stroke Scale, M1 versus internal carotid artery (ICA) occlusion, and successful recanalization(TICI 2b/3). The training model of 10-point ASPECTS(tASPECTS) vs. 7-point cortical-weighted ASPECTS(cASPECTS) was generated in site 1 and validated in site 2, with models compared using receiver operator characteristics(ROC) and corrected Akaike’s Information Criterion(AICc).
Results:
Of 328 included patients (site 1/training cohort=181; site 2/validation=147), median age was 71y(IQR 61.25-82), of whom 119(36.28%) had ICA occlusion, with a median tASPECTS of 9(IQR 8-10) and median cASPECTS of 6(IQR 5-7). In derivation cohort for the cASPECTS, multi-class adjusted ROC curves were generated, with areas under the curve (AUC) for predicting mRS shift ranging from 0.8546-0.9110(AICc=690.078, r^2=0.5936). In validation cohort, the misclassification rate was 71.6%, however, when expanded for flexibility(+/-1 mRS unit), model classifies validation site data correctly 55.8% of time. In training cohort for the tASPECTS, the AUC for predicted mRS classes ranged from 0.8812-0.9250(AICc=693.646, r^2=0.6165), with no difference in model performance against cASPECTS (p=0.14). In validation cohort, the misclassification rate was 73.47%, and improved to 50.3% correct when expanded for flexibility(+/-1 mRS unit).
Conclusions:
The cASPECTS and tASPECTS were similarly predictive of 90-day mRS scores following successful recanalization.