Abstract WP48: Inter-rater Reliability for Radiological Studies Used to Select Patients for Endovascular Stroke Therapy
Introduction: The selection of acute ischemic stroke patients relies on the interpretation of radiological studies, commonly including noncontrast CT (ASPECTS score), CTA, and in some circumstances, CT perfusion (CTP) scans.
Hypothesis: The interrater reliability between readers of noncontrast head CT will be moderate and not as good as that between readers of CT perfusion scans.
Methods: Retrospective study of consecutive ischemic stroke patients who were considered for endovascular stroke therapy at an academic medical center from April 2010 - November 2014. All patients underwent CT, CTA, and CTP. Images were independently reviewed by four readers (A, B, C, D) in total (2 experienced neuroradiologists, one experienced neurointensivist, and one neuroradiology fellow). Each rated half of the cohort. All readers performed online training in ASPECTS. Moderate-to-good collateral circulation was defined as filling of 50% or more of the middle cerebral artery pial arterial territory on CTA, as in ESCAPE trial. An “acceptable core” on CTP was defined as in EXTEND-IA trial with core volume < 70 mL, mismatch ratio >1.2, and absolute mismatch volume >10 mL but was qualitatively estimated. Interrater agreement was assessed for total ASPECTS, dichotomized ASPECTS (≥6 vs <6), evaluation of collaterals, and evaluation of core. Agreement was assessed using kappa statistic, which was weighted for the total ASPECTS analysis.
Results: Our cohort included 119 patients with acute stroke and large vessel occlusion. Interrater agreement for total ASPECTS was very good between readers C and D (' 0.70) and good between A and B (' 0.55). For dichotomized ASPECTS, agreement was fair-to-good (' 0.30 for C/D; ' 0.41 for A/B). Agreement for grading collaterals on CTA was good to very good (' 0.64 for C /D; ' 0.54 for A/B). Agreement for rating core on CTP was very good (' 0.78 for C/D; ' 0.62 for A/D).
Conclusions: Interrater reliability is good to very good for total ASPECTS score, dichotomous grading of collaterals on CTA, and subjective estimation of core size on CTP. Agreement is only fair to good for dichotomized ASPECTS.
Author Disclosures: J.E. Fugate: None. W. Brinjikji: None. D.F. Kallmes: None. H. Cloft: None. A.A. Rabinstein: None.
- © 2016 by American Heart Association, Inc.