Neurological event prediction for patients with symptomatic cerebral cavernous malformation: the BLED2 score.

Abstract

Retrospective patient cohort studies have identified risk factors associated with recurrent focal neurological events in patients with symptomatic cerebral cavernous malformations (CCMs). Using a prospectively maintained database of patients with CCMs, this study identified key risk factors for recurrent neurological events in patients with symptomatic CCM. A simple scoring system and risk stratification calculator was then created to predict future neurological events in patients with symptomatic CCMs. This was a dual-center, prospectively acquired, retrospectively analyzed cohort study. Adult patients who presented with symptomatic CCMs causing focal neurological deficits or seizures were uniformly treated and clinically followed from the time of diagnosis onward. Baseline variables included age, sex, history of intracerebral hemorrhage, lesion multiplicity, location, eloquence, size, number of past neurological events, and duration since last event. Stepwise multivariable Cox regression was used to derive independent predictors of recurrent neurological events, and predictive accuracy was assessed. A scoring system based on the relative magnitude of each risk factor was devised, and Kaplan-Meier curve analysis was used to compare event-free survival among patients with different score values. Subsequently, 1-, 2-, and 5-year neurological event rates were calculated for every score value on the basis of the final model. In total, 126 (47%) of 270 patients met the inclusion criteria. During the mean (interquartile range) follow-up of 54.4 (12-66) months, 55 patients (44%) experienced recurrent neurological events. Multivariable analysis yielded 4 risk factors: bleeding at presentation (HR 1.92, p = 0.048), large size ≥ 12 mm (HR 2.06, p = 0.016), eloquent location (HR 3.01, p = 0.013), and duration ≤ 1 year since last event (HR 9.28, p = 0.002). The model achieved an optimism-corrected c-statistic of 0.7209. All factors were assigned 1 point, except duration from last event which was assigned 2 points. The acronym BLED2 summarizes the scoring system. The 1-, 2-, and 5-year risks of a recurrent neurological event ranged from 0.6%, 1.2%, and 2.3%, respectively, for patients with a BLED2 score of 0, to 48%, 74%, and 93%, respectively, for patients with a BLED2 score of 5. The BLED2 risk score predicts prospective neurological events in symptomatic CCM patients.