How useful is the ABCD2 TIA score in predicting and preventing stroke?

Abstract

Background and objectives: Independent validations of the ABCD2 score used to predict stroke development have reported conflicting results, and besides expert opinion as to proper diagnostic approach and best treatment differs widely. A model predictive power can be modified by the concomitant use of effective diagnostic and pharmacological treatments. We aimed to determine the predictive power of the ABCD2 score while simultaneously providing patients with current urgent recommended treatments and recording their early and long term health outcomes. Methods: Data were retrospectively collected from all the patients presenting with a TIA for a whole year and were followed for another whole year. Physicians completed data forms with the ABCD2 score when patients arrived at the emergency department (ED).We calculated sensitivity, specificity for predicting stroke at 7 and 30 days after visiting the ED using the high-risk cutpoint of an ABCD2 score ≥ 4.Univariate Cox proportional hazards regression modelling was performed for ABCD2 score to estimate the hazard ratios relative to the low-risk category and to assess the effect of the individual components of the ABCD2 score and other potential risk factors to predict stroke development. Results: We enrolled 172 patients (mean age 71 yr, 51 % women) with a new incident diagnosis of TIA. The mean (SD) ABCD2 score was 4.2 (1.4). There were 7 new TIA, 17 non fatal strokes and 3 fatal strokes. Intrahospital mortality was 1.7% and 8.7% during the 1 year follow up. An ABCD2 score of ≥ 4 had a sensitivity of 88% and 82 % for a stroke at 7 and 30 days respectively, a poor specificity of 30%. Negative predictive value at 7 days was 98%. ABCD2 score ≥ 4 had no significant predictive value for stroke within 7 days (hazard ratio [HR], 3.49; 95%CI, 0.42 to 27.93) and 30 days (HR, 1.97; 95%CI, 0.43 to 9.13) of the event. Only diabetes predicted an increased likelihood of stroke over the first week (HR, 5.47; 95%IC 1.43 to 20.95) and over the first month (HR, 3.60; 95%IC 1.08 to 12). Conclusions: An ABCD2 score of < 4 has a good negative predictive value (98%) for stroke development within the first 7 days. However, the low positive predictive value of the ABCD2 score fails to predict with a high level of confidence the future occurrence of a stroke. It was only being diabetic that was significantly related to the probability of stroke development. TIA probably justifies early accurate identification of the underlying TIA etiology for nearly all presentations. We recommend adding a systematic Brain CT, carotid ultrasound and ECG within 24 hours while concomitantly starting urgent treatment.