Potential and Failure of the ABCD Score in Stroke Risk Prediction After Transient Ischemic Attack

See related articles, pages 844–850 and 851–856. Patients with transient ischemic attacks (TIAs) are a heterogeneous group in terms of risk factors, symptomatology, underlying disorders, and prognosis.1 The importance of recognizing this common condition lies in the high risk of early stroke that TIAs carry (3.1% at 2 days, 5.2% at 7 days, 8.0% at 30 days, and 9.2% at 90 days).2,3 Recent evidence suggests that the early stroke risk after a TIA may be estimated from prediction scores based on clinical features, etiology, vascular imaging, or diffusion magnetic resonance imaging.1,4,5 The ABCD system (ABCD/ABCD: [Age { 60 years 1 point}, Blood pressure {systolic 140 mm Hg and/or diastolic 90 mm Hg 1}, Clinical features {unilateral weakness 2, speech disturbance without weakness 1, other symptoms 0}, Duration of symptoms { 10 minutes 0, 10 to 59 minutes 1, 60 minutes 2}, and Diabetes mellitus {yes 1}]) is a simple clinical tool that has been recently developed to predict individual risk of stroke and to triage TIA patients on first presentation to medical attention.6,7 Current international guidelines have rapidly adopted the former score in risk stratification of TIA patients, advocating immediate hospitalization and emergent diagnostic evaluation of TIA patients with an ABCD score of 3 or higher in the United States (American Heart Association recommendations)4 and of 4 or higher in the United Kingdom (National Institute of Clinical Excellence Guidelines).5 However, after excluding the original derivation studies of the ABCD system using data from the same 2 large, populationbased TIA registries in California and Oxfordshire,6,7 multiple external validation studies have yielded inconsistent results,8–21 ranging from insignificant to limited8,15,18 to satisfactory to excellent10,13,16,17 predictive ability. The extremely varied methodology in the different validation studies may contribute to these divergent findings, including a limited number of outcome events (as low as 2 in 1 study),15 the clinical specialty of the initial evaluation (neurologists10–13,15–17 vs emergency department physicians9,14,21), retrospective data collection by chart review,9,10,14,19 and lack of follow-up after the first 7 days.9,19 Notably, a recent systematic review and meta-analysis of 11 studies reporting the performance of the ABCD system in a total of 5938 TIA patients has reported a satisfactory predictive ability for both the ABCD (pooled estimate of the area under the curve [AUC] for prediction of 7-day risk 0.74; 95% CI, 0.68 to 0.81) and the ABCD (pooled estimate of the AUC for prediction of 7-day risk 0.77; 95% CI, 0.63 to 0.91) scores.22 In this issue of Stroke, Sheehan et al23 and Chandratheva et al24 report the results of 2 independent, relatively large, prospective, population-based studies investigating the diagnostic utility of ABCD for prediction of early stroke risk after TIA. More specifically, Sheehan and colleagues23 used data from the North Dublin TIA Study (including all TIA cases identified during a 3-year period in a prospective, population-based cohort of 294 529 inhabitants of North Dublin city) to externally validate the ABCD score and to evaluate whether carotid stenosis or atrial fibrillation might add to the prognostic information yielded by the ABCD score. Interestingly, they noted that the degree of carotid stenosis was linearly associated with increased stroke risk after TIA, whereas atrial fibrillation was not. In addition, they documented an agreeable predictive utility of the score in nonspecialist-suspected TIA patients (n 700; AUC for prediction of the 90-day risk 0.61; 95% CI, 0.52 to 0.71). In contrast, they showed that the predictive ability of the ABCD score was no better than chance in TIA cases confirmed by stroke specialists (n 443, AUC for prediction of the 90-day risk 0.55; 95% CI, 0.45 to 0.64), largely related to the 24.2% (8/33) of recurrences documented in patients with low ABCD scores. (0–3) These findings are at odds with the currently supported notion that TIA patients with low ABCD scores carry an insignificant risk of stroke and highlight the importance of emergent carotid evaluation in all TIA patients, independent of the presenting ABCD scores. The main strengths of this timely study are related to the prospective and population-based design, the large sample size, the “hot pursuit” strategy used for TIA identification, and the regular follow-up assessments. On the other hand, certain methodological shortcomings need to be acknowledged: (1) the lack of brain imaging data in a substantial portion of confirmed TIA patients (14.4%); and (2) the absence of a uniform predefined protocol regarding secondary prevention strategies, which were delivered according to the practice of the treating physician and patient preference. Therefore, potential variations in TIA clinical management may have accounted for the higher risk of stroke documented in the low-ABCD-score group who fared poorly. Third, previous investigators have indicated that the ABCD system appears to identify TIA patients with 50% carotid artery stenosis or atrial fibrillation who are at high risk of early stroke.25 However, in the present series, the potential interaction between ABCD score and the degree of carotid artery The opinions expressed in this editorial are not necessarily those of the editors or of the American Heart Association. From the Department of Neurology, Democritus University of Thrace, Alexandroupolis, Greece. Correspondence to Georgios Tsivgoulis, MD, FESO, Kapodistriou 3, Nea Xili, Alexandroupolis, Greece, 68100. E-mail [email protected] (Stroke. 2010;41:836-838.) © 2010 American Heart Association, Inc.