Background Pre-treatment re-bleeding following aneurysmal subarachnoid hemorrhage (aSAH) affects up to 7.2% of patients even with ultra-early treatment within 24 hours. We retrospectively compared the utility of three published re-bleed prediction models and individual predictors between cases who re-bled matched to controls using size and parent vessel location from a cohort of patients treated in an ultra-early, ‘endovascular first’ manner.

Methods On retrospective analysis of our 9-year cohort of 707 patients suffering 710 episodes of aSAH, there were 53 episodes of pre-treatment re-bleeding (7.5%). Forty-seven cases who had a single culprit aneurysm were matched to 141 controls. Demographic, clinical and radiological data were extracted and predictive scores calculated. Univariate, multivariate, area under the receiver operator characteristic curve (AUROCC) and Kaplan–Meier (KM) survival curve analyses were performed.

Results The majority of patients (84%) were treated using endovascular techniques at a median 14.5 hours post-diagnosis. On AUROCC analysis the score of Liu et al. had minimal utility (C-statistic 0.553, 95% confidence interval (CI) 0.463 to 0.643) while the risk score of Oppong et al. (C-statistic 0.645 95% CI 0.558 to 0.732) and the ARISE-extended score of van Lieshout et al. (C-statistic 0.53 95% CI 0.562 to 0.744) had moderate utility. On multivariate modeling, the World Federation of Neurosurgical Societies (WFNS) grade was the most parsimonious predictor of re-bleeding (C-statistic 0.740, 95% CI 0.664 to 0.816).

Conclusions For aSAH patients treated in an ultra-early timeframe matched on size and parent vessel location, WFNS grade was superior to three published models for re-bleed prediction. Future re-bleed prediction models should incorporate the WFNS grade.