Twenty-one studies were included in the review. A case series from the authors' own institution was included. Twelve studies included patients with crescendo transient ischaemic attacks , and 16 included patients with stroke-in-evolution. The methodological quality was considered poor.
Crescendo transient ischaemic attacks (n=176 patients): The pooled perioperative stroke rate was 6.5% (95% CI 3.4 to 10.4); the pooled combined stroke/death rate was 9.0% (95% CI 4.3 to 15.1); and the pooled combined stroke/death/major cardiac event was 10.9% (95% CI 5.5 to 17.9).
Stroke-in-evolution (n=114 patients): The pooled perioperative stroke rate was 16.9% (95% CI 9.2 to 26.2); the pooled combined stroke/death rate was 20.0% (95% CI 12.5 to 28.6); and the pooled combined stroke/death/major cardiac event was 20.8% (95% CI 13.2 to 29.6).
All analyses for both groups of patients used a random-effects model, although statistically significant heterogeneity was not detected. This was due to the presence of selection bias in the majority of included studies.