Simplified Edinburgh CT Criteria for Identification of Lobar Intracerebral Hemorrhage Associated With Cerebral Amyloid Angiopathy

Abstract

Background and Objectives:In patients with lobar intracerebral hemorrhage(ICH) etiological characterization represents a trade-off between feasibility, resource allocation, and diagnostic certainty. This study investigated the accuracy and clinical utility of the simplified Edinburgh CT criteria to identify underlying cerebral amyloid angiopathy(CAA).Methods:This external validation analyzed 210 consecutive patients with lobar ICH and available CT&MRI studies from a prospective single-center observational cohort study(2006-2015,UKER-ICH,NCT03183167). We investigated the simplified Edinburgh CT-based criteria’s inter-rater variability and diagnostic accuracy for identification of ICH associated with probable CAA according to MRI-based modified Boston criteria as a reference standard. We evaluated the simplified Edinburgh criteria’s utility by decision curve analysis, comparing the theoretical clinical net-benefit(weighted benefit–harm at varying threshold probabilities) of the high-risk category(finger-like-projections and subarachnoid hemorrhage) for ruling-in and the low-risk category(neither finger-like-projections nor subarachnoid hemorrhage) for ruling-out with the assumptions of no or all patients having CAA(default-strategies).Results:Of 210 patients, 70(33.3%) had high-risk, 67(31.9%) had medium-risk, and 73(34.8%) had low-risk for CAA associated ICH according to simplified Edinburgh CT criteria, showing moderate inter-rater variability. Discrimination was good(AUROC:0.74,95%CI 0.67–0.81) without evidence of poor calibration(Hosmer–Lemeshow,p=0.54) for validation of MRI-based diagnosis of probable CAA(n=94/210,44.8%). The rule-in criteria, i.e. high-risk, had 87.1%(79.3-92.3) specificity, and the rule-out criteria, i.e. low-risk, had 80.9%(71.1-88.0) sensitivity. Decision curve analysis suggested a theoretical clinical net-benefit for ruling-in but not for ruling-out probable CAA compared to default-strategies.Discussion:Applying the simplified Edinburgh CT criteria during diagnostic work-up seems clinically useful and may accurately identify CAA in patients with lobar ICH.Classification of Evidence:This study provides Class II evidence that in patients with lobar hemorrhages, simplified Edinburgh Criteria accurately identifies those at high risk of CAA.

CT Diagnosis of Cerebral Amyloid Angiopathy

“…5 Another limitation is the interrater reliability of finger-like projections detection, a concern directly addressed by the Edinburgh group through an online training tool. Sembill et al 6 reported only moderate interrater agreement for finger-like projections (Cohen kappa of 0.66 vs 0.80 for SAH and 0.81 for the Boston criteria diagnosis), similar to that in the original publication.…”

“…This possibility is supported by the observed association of SAH with cSS, 6 an imaging finding linked to severe leptomeningeal CAA and vascular degeneration. 7 Sembill et al 6 also noted association between cSS and finger-like projections, an ICH feature that is not understood at a mechanistic level but also potentially a reflection of bleeding from a leptomeningeal source.…”

“…Use of the criteria comes with limitations, however, notably that a substantial subset of patients with lobar ICH (;1/3 in the current study 6 ) will have only 1 of the 2 CT features and thus fit into neither category, and that both false-positives and falsenegatives can occur, with highest false-negative likelihood at smaller ICH volumes. Balancing the benefits of a true diagnosis vs the risks of a false one, modeled by decision curve analysis in the study by Sembill et al, 6 is likely to depend closely on the clinical particulars of a patient's situation. When MRI is available, it appears to remain the more accurate and generalizable diagnostic test for CAA (pending further headto-head radiologic-pathologic validation data) because of its ability to detect an individual's lifetime record of prior hemorrhagic lesions, its applicability to non-ICH as well as ICH clinical presentations, its ability to detect some secondary ICH causes such as cavernous malformations, and its interrater reliability.…”

Simplified Edinburgh and modified Boston criteria in relation to amyloid PET for lobar intracerebral hemorrhage

“… The simplified Edinburgh criteria are based on analysis of the non-contrast CT. ( Rodrigues et al, 2018 ) In the original study, ( Rodrigues et al, 2018 ) no patients had fingerlike projections in isolation, but given the association between CAA and fingerlike projections, fingerlike projections in isolation were (in line with previous studies) ( Schwarz et al, 2022 , Sembill et al, 2022 ) categorized as medium risk. …”

“…Neuroimaging with Computed Tomography (CT) or Magnetic Resonance Imaging (MRI) can reveal characteristics suggestive of CAA. The simplified Edinburgh criteria ( Table 1 ) ( Rodrigues et al, 2018 , Schwarz et al, 2022 , Sembill et al, 2022 ) and the modified Boston criteria ( Table 2 ) ( Smith and Greenberg, 2003 , Linn et al, 2010 , Greenberg and Charidimou, 2018 ) evaluate and categorize these features to make a clinical diagnosis of CAA when tissue analysis is lacking. Validation studies of these criteria against histopathological evidence revealed 88–100% specificity for ruling in “high risk CAA” and “probable CAA” ( Rodrigues et al, 2018 , Greenberg and Charidimou, 2018 ).…”