Hemispherectomy Outcome Prediction Scale: Development and validation of a seizure freedom prediction tool

Abstract

Objective: To develop and validate a model to predict seizure freedom in children undergoing cerebral hemispheric surgery for the treatment of drug-resistant epilepsy. Methods: We analyzed 1267 hemispheric surgeries performed in pediatric participants across 32 centers and 12 countries to identify predictors of seizure freedom at 3 months after surgery. A multivariate logistic regression model was developed based on 70% of the dataset (training set) and validated on 30% of the dataset (validation set). Missing data were handled using multiple imputation techniques. Results: Overall, 817 of 1237 (66%) hemispheric surgeries led to seizure freedom (median follow-up = 24 months), and 1050 of 1237 (85%) were seizure-free at 12 months after surgery. A simple regression model containing age at seizure onset, presence of generalized seizure semiology, presence of contralateral 18-fluoro-2-d eoxyglucose-positron emission tomography hypometabolism, etiologic substrate, and previous nonhemispheric resective surgery is predictive of seizure freedom (area under the curve = .72). A Hemispheric Surgery Outcome Prediction Scale (HOPS) score was devised that can be used to predict seizure freedom.

Development and validation of machine learning models for prediction of seizure outcome after pediatric epilepsy surgery

“…Our logistic regression model achieved similar performance. Among the features that we identified as important for predicting seizure freedom, age at seizure onset and MRI lesion were also identified by Weil et al 33 as significant predictors, although unlike their study we did not identify seizure semiology (generalized seizures) or contralateral PET hypometabolism (discordant PET findings) to be important. Differences in findings could be related to our inclusion of a wider spectrum of surgery types, rather than restricting the analysis to hemispherectomy, so that our findings are applicable to a broader pediatric epilepsy surgery population.…”

“…There is a paucity of studies utilizing ML algorithms to predict seizure outcome after pediatric epilepsy surgery. Weil et al 33 developed and validated a prediction tool to assess seizure freedom after hemispherectomy in children using a multivariate logistic regression model. Others have utilized nomograms and simple seizure‐freedom scores for predicting seizure freedom in mixed adult and pediatric populations 34,35 .…”

“…Weil et al 33 showed that their multivariate logistic regression model had an AUC of .72 for predicting seizure‐free outcome at 1 year following hemispherectomy. Our logistic regression model achieved similar performance.…”

Not surgical technique, but etiology, contralateral MRI, prior surgery, and side of surgery determine seizure outcome after pediatric hemispherotomy

“…range reported by a recently published large multicentric study 21,22 and three meta-analyses, 10,18,19 all including patients who underwent hemispheric procedures as early as the 1970s. This finding is particularly reassuring, considering the broader spectrum of etiologies and the increased complexity of surgical cases in recent years.…”

Surgery for pediatric drug resistant epilepsy: a narrative review of its history, surgical implications, and treatment strategies

“…Various techniques have been described since including the hemispherotomy technique described by Delalande, the endoscopically assisted hemispherectomy technique and endoscopic transventricular approach used for hypothalamic hamartoma being most recently described which aim to reduce the amount of brain tissue removed ( 84 , 85 ). A multicenter analysis described the evolution of the Hemispherectomy Outcome Prediction Scale evaluating outcomes of over 1,200 patients ( 86 ). With utilization of this score, dysplastic etiologies or patients with infantile spasms do worse after hemispherectomy and the use of preoperative MRI was a significant component in determining indication for surgery.…”