Thirteen comparative cohort studies were included (3,556 participants, range 72 to 1,135). All were prospective and had well defined methodology, four studies did not clearly report whether participants were enrolled consecutively, 11 compared the groups at baseline for most potential confounders, nine adjusted statistically for baseline differences and seven described protocol violations. Only two of eleven studies that reported death and recovery rates at three months described losses to follow-up.
By three month follow-up, the over 80 years old group were significantly more likely to die (OR 2.77, 95% CI 2.25 to 3.40; 10 studies) and less likely to have functional recovery (OR 0.49, 95% CI 0.40 to 0.61; eight studies). There was no significant difference between the groups in the incidence of symptomatic intracranial haemorrhage (OR 1.31, 95% CI 0.93 to 1.84; 13 studies). There was no significant heterogeneity for any of these analyses (P>0.1) and sensitivity analyses did not substantially change the main findings. There was no indication of publication bias.
Pooling of the subgroup of studies that defined recovery as modified Rankin scale 0-2 resulted in significant heterogeneity (p=0.06) so were considered invalid.