Introduction Pre-treatment re-bleeding is a devastating complication following aneurysmal subarachnoid haemorrhage (aSAH) with reported mortality up to 73% and a reported rate of poor neurological outcome in up to 84% of patients. Even with ultra-early treatment within 24 hours of ictus it affects up to 7.2% of patients. Three scoring systems have been described incorporating accepted clinical and radiological re-bleed predictors in various combinations. In this study we compared the predictive utility of each scoring system and individual established clinical and radiological predictors.

Methods Retrospective analysis of our single centre 9-year cohort of 642 consecutive aSAH patients treated within an ultra-early timeframe, 36 had pre-treatment re-bleeding (5.6%). Thirty suitable cases who had a single culprit saccular aneurysm and pre-treatment non-invasive cerebrovascular imaging were matched based on size and parent vessel location to 90 controls who did not re-bleed on a 1:3 basis. Demographic, clinical and radiological data were extracted and predictive scores calculated. Univariate, multivariate and area under the receiver operator characteristic curve (AUROCC) analyses were performed.

Results The majority of all patients (95%) were treated using endovascular techniques at a median time 15.25 hours post-ictus. Both of these metrics did not differ in a statistically significant manner between cases and control on univariate analysis. Median World Federation of Neurological Surgeons (WFNS) grade at presentation for cases was 4 compared to 2 for controls (p=0.021) whilst the median Fisher grade at presentation was 4 for cases and 3 for controls (p=0.001). At presentation 60% of cases had an ICH compared to 26% of controls (p=0.001). Symptomatic hydrocephalus requiring the placement of an external ventricular drain prior to aneurysm treatment was present for 60% of cases compared to 38% of controls (p=0.033). On AUROCC analysis the score of Liu et al. displayed minimal predictive utility (AUROCC 0.526, 95%CI 0.416 - 0.636) whilst the score of Oppong et al. (AUROCC 0.648 95% CI 0.541 - 0.754) and the ARISE-extended score (AUROCC 0.652 95% CI 0.529 - 0.775) displayed moderate utility. On multivariate modelling WFNS grade at presentation alone was the most parsimonious predictive model for re-bleed prediction (AUROCC 0.787, 95% CI 0.0.687 - 0.887) with superiority over all other potential factors in single, two or three factor combinations and each of the three described predictive scores.

Conclusion In aSAH patients treated within an ultra-early timeframe, WFNS grade at presentation displayed optimal utility for re-bleed prediction compared to three published scoring systems incorporating varying combinations of multiple re-bleed predictors. Future re-bleed prediction models should incorporate the WFNS. Furthermore, earlier aneurysm occlusion should not be withheld from high grade aSAH patients with favourable prognostic factors as these patients may stand to benefit most from re-bleed prevention.

Disclosures A. Dissanayake: None. E. Burrows: None. K. Ho: None. T. Phillips: None. S. Honeybul: None. G. Hankey: None.