Validation of the DIAS³ Score in Cerebral Venous Thrombosis Patients Treated by Decompressive Surgery

Introduction Cerebral venous thrombosis (CVT) is associated with a risk of late seizures, particularly in patients with severe disease requiring decompressive surgery. The DIAS3 score was developed to predict late seizures after CVT, but its performance in surgically treated patients is unknown. We aimed to validate the DIAS3 score in a high-risk cohort of CVT patients treated by decompressive surgery. Methods This was a substudy of the DECOMPRESS2 - a prospective, multicentre cohort, including patients with CVT treated with decompressive surgery. Patients with available data on the occurrence and timing of seizures following CVT were included. Late seizures were defined as those occurring more than 7 days after CVT diagnosis, up to 1 year follow-up. We calculated DIAS3 score for all included patients. Predictive performance was assessed only for the 1-year risk of late seizures. Discrimination was evaluated using receiver operating characteristic analysis, and calibration by comparing predicted and observed seizure risks. Diagnostic performance was assessed across multiple probability thresholds. Results Eighty-nine patients were included (median age 38 years; 69.7% female). Late seizures occurred in 22 patients (24.7%). The mean predicted 1-year seizure risk according to the DIAS3 score was 28.4%. The DIAS3 score showed poor discrimination (area under the curve 0.44, 95% confidence interval 0.30-0.59) and poor calibration. Across all tested thresholds, sensitivity and specificity showed an expected trade-off, but overall predictive accuracy was limited. Conclusion The DIAS3 score demonstrated poor discrimination and calibration in CVT patients treated by decompressive surgery. These findings suggest that the DIAS3 score should not be used to predict late seizures in this high-risk surgical subgroup.