Abstracts

Rationale: The Hemispherectomy Outcome Prediction Scale (HOPS) was developed to aid both clinicians and patients in determining chance of success after hemispheric surgery for medically refractory epilepsy. Its multi-institutional, large cohort format stratified study patients well. Evidence suggests that methodologies utilized to create such predictive models including cross-validation as well as stratification utilizing the same data employed for model generation may be at risk of an undesirable modeling phenomenon known as overfitting. We describe our experience of hemispherectomy outcomes and investigate the HOPS protocol to see if retrospectively there would be any evidence of overfitting in predicting outcomes.

Methods: We performed a retrospective, observational chart review using patient data ranging from June 2014 to December 2022 at Nicklaus Children’s Hospital (NCH), Miami FL. A comprehensive search of NCH’s neurosurgical database was conducted in REDCap where qualifying patients were gathered. The percentage of patients in our cohort with complete seizure freedom at the time of last follow-up was tabulated for each HOPS score category and compared to our actual data.

Results: We found HOPS to stratify our limited post-operative patients adequately. However, the likelihood of complete seizure freedom among the patients predicted by HOPS to be at greatest chance of success was ~75%, about 20 points lower than the original HOPS cohort (Table 1). This reduction in absolute chance of success predicted by HOPS may represent some degree of overfitting. Our small cohort provides preliminary evidence suggesting the need for collection of data from a large cohort of patients entirely external to the original HOPS study for external validation.

Conclusions: The results of this study provide preliminary evidence that the HOPS criteria may not be universally generalizable and may overestimate the clinical effectiveness of this surgery. The current HOPS criteria may suffer from some degree of statistical overfitting which can only truly be revealed and overcome with larger external validation efforts in the future. We suggest that a more comprehensive approach to the external validation of HOPS would enhance its utility by delineating more precisely how generalizable its output may be and therefore how predictive it can be for diverse patients in the real world. Revision of the HOPS protocol to avoid overfitting would be critical to accurately determine surgical success and patient recovery after epileptic surgery.

Funding: N/A