Introduction Stroke is a leading cause of death and acquired disability in every society in which it has been studied. Stroke and transient ischaemic attack (TIA) arise from identical aetiologies and a number of studies have demonstrated that TIAs carry a significant risk of stroke. Several independent predictors of stroke have been incorporated into models such as the ABCD² clinical prediction rule, which is used to predict risk of stroke following TIA. This systematic review assessed the predictive value of the ABCD² rule in relation to 7 and 90 day risk of stroke.

Methods A computerised systematic literature search was performed to retrieve articles that validated the ABCD² rule. The original derivation study was used as a predictive model and applied to all validation studies, with observed and predicted number of strokes at 7 and 90 days stratified by risk group (0-3 low, 4-5 moderate, 6-7 high). Results from the studies were pooled and risk ratios (RR) with 95% CI produced. Forest plots were used to graphically display the data. A RR score of 1 represents correct prediction by the ABCD² rule, <1 represents under-prediction and >1 over-prediction.

Results Nine validation studies (n=5626) predicted 7 day stroke risk. The ABCD² rule correctly predicted occurrence of stroke at 7 days across all three risk strata: low risk (n=1933) — RR 1.12, 95% CI (0.61 to 2.05); moderate risk (n=2640)—RR 1.11, 95% CI (0.74 to 1.68); high risk (n=1053)—RR 0.98, 95% CI (0.69 to 1.41). There were 318 strokes predicted and 288 strokes observed at 7 days across all three risk strata. Data on five studies (n=4897) were pooled to predict 90 day stroke risk. The ABCD² rule over-predicted the occurrence of stroke across all three risk strata — low risk (n=1660), RR 1.50, 95% CI (0.86 to 2.62); moderate risk (n=2214), RR 2.24, 95% CI (1.29 to 3.91); high risk (n=1033), RR 2.00, 95% CI (0.90 to 4.45). There were 268 strokes observed at 90 days in contrast to 404 predicted strokes. The chi-squared trend for analysis indicated that as the trichotomised ABCD² score increased, the rate of stroke increased (p<0.0001).

Conclusion The ABCD² score correctly predicts 7 day risk of stroke across all risk strata but over-predicts 90 day risk of stroke in all groups. The variation in the study setting and design needs to be considered in the interpretation of these findings. ABCD² is a useful CPR, particularly in relation to 7 day risk of stroke.