Development and external validation of the LEAN score to predict late seizures after intracerebral haemorrhage

Introduction: Predicting the occurrence of late seizures after intracerebral haemorrhage may help in making clinical decisions about treatment. Currently, the CAVE score is the best performing risk score. We aimed to design a different, pragmatic risk prediction score and compared it to the CAVE score. Patients and methods: The South Limburg (Netherlands) intracerebral haemorrhage registry, consisting of patients with a primary intracerebral haemorrhage in 2004–2009, was used for the derivation cohort. We made a prediction model using Cox proportional hazard analyses; comparisons between models were made with the c-statistic. We validated our model externally in three independent cohorts. Results: Our derivation cohort consisted of 781 patients, of whom 78 (10%) developed late seizures. We found the following independent predictors for late seizures: any neurosurgical procedure, age < 65 years, lobar haemorrhage, and early seizures (occurring within the first week). These formed our new prediction score (LEAN score), which had an optimism-corrected c-statistic of 0.80 (95%-confidence interval 0.78–0.86). The LEAN score predicts late seizure risk as 0.7%, 1.6%, 8.8%, 22.0%, 29.8%, 43.5%, 100% for the increasing score groups respectively. External validation showed comparable optimism-corrected c-statistics for both the LEAN score and the CAVE score. Conclusion: The newly developed LEAN score consists of easily available clinical variables and performs equally to the CAVE score. Additionally, the high risk of late seizures in patients with the maximum LEAN score might make a diagnosis of epilepsy possible according to international guidelines despite these patients only had early seizures.