Ineligible by 1,000 Cuts: Identifying Epilepsy Clinical Trial Ineligibility from Structured Electronic Health Records
Wesley Kerr1, Katherine McFarlane1, Dane Prince1, Vijayalakshmi Rajasekaran1, James Castellano1, Laura Kirkpatrick2
1University of Pittsburgh, 2UPMC Children's Hospital of Pittsburgh
Objective:
We evaluated how structured information in the electronic health records identified the most common co-occurring conditions and vital sign reasons for epilepsy clinical trial ineligibility.
Background:
Clinical trials are necessary to evaluate the efficacy and safety of novel epilepsy treatments. New treatments are necessary for the estimated 30% of patients with continued seizures. However, recruitment per site has lessened, requiring more trial sites and increasing costs.
Design/Methods:
Adults with epilepsy seen in an outpatient encounter by an epileptologist within a 1-year period from 2024 to 2025 were identified from the University of Pittsburgh Medical Center’s electronic health record (EHR) using any International Classification of Diseases, 10th Revision (ICD-10) code for epilepsy (G40.*) or two codes for convulsions (R56.*) separated by 30 days. Based on criteria for currently recruiting trials, excluded co-occurring conditions were identified using ICD-10 codes. Vital signs and demographic information were extracted from the EHR.
Results:
We identified 13,678 patients aged between 17.5 and 74.5 years without developmental and epileptic encephalopathy (DEE) codes. Epileptologists saw 26% (3,510) of patients. Of patients who saw epileptologists, eligibility was most limited by co-occurring mental health disorders (excluded 81%), metabolic illness (excluded 57%), and neoplasia (excluded 42%). Body mass index excluded 22% (18% above 35). Only 7% (230) of patients had no exclusions and 14% (497) had one exclusion.
Conclusions:
Only 7% of patients seen by epileptologists appeared eligible for clinical trials after evaluating vital signs and co-occurring conditions. Limiting recruitment to epileptologists markedly reduced the number and representativeness of potentially eligible patients. However, reliance on ICD-10 coding not only may have false positives, but also unnecessarily exclude patients with mild co-occurring disease. Integration and analysis of non-structured patient data (e.g., clinical notes) will be necessary for successful application of inclusion and exclusion criteria to achieve the ultimate goal of increased enrollment.
Disclaimer: Abstracts were not reviewed by Neurology® and do not reflect the views of Neurology® editors or staff.