Optimal choice of antiseizure medication: Agreement among experts and validation of a web‐based decision support application

Abstract

ObjectiveOptimal choice of antiseizure medication (ASM) depends on seizure type, syndrome, age, gender, comorbidities and co‐medications. There are no fixed rules on how to weigh these factors; choices are subjective and experience‐driven. We investigated agreement among experts in selecting ASM as monotherapy and used their prevailing choices to validate a web‐based decision‐support application.MethodsTwenty‐four international experts, blinded to the app, selected the optimal ASM for 25 individual patient‐cases covering a wide variation of seizure types and other factors influencing ASM selection. The app ranked ASMs in order of likely appropriateness for each case. In a second step, experts rated anonymously the choices of the app.ResultsOf the 25 patient‐cases (age 13‐74 years), 13 were female, 18 (72%) had comorbidities, six (24%) were on contraceptives, and 13 (52%) had other co‐medications. The median number of experts who selected the same ASM for a given case was 15 (62.5%) and interquartile range (IQR) 13‐18 (54%‐75%). Gwet's agreement coefficient among experts was 0.38 (95% confidence interval [CI] 0.32‐0.44), corresponding to a "fair" agreement. Agreement between the app and the prevailing expert choice for each case was 0.48 (95% CI 0.29‐0.67), corresponding to a “moderate” beyond chance agreement. The percent agreement between the highest ranked selections of the app and the expert selections was 73% (95% CI 64%‐82%). Ninety‐five percent of the experts considered that no incorrect or potentially harmful ASMs were ranked highest by the app, and most experts strongly agreed with the app's selections.SignificanceThis app, now validated by experts, provides an objective, reproducible method for selecting ASM that accounts for relevant clinical features. It is freely available at: https://epipick.org.

Potential for improved retention rate by personalized antiseizure medication selection: A register‐based analysis

“…Real‐world data from registers could be of great complementary use for clinicians, in addition to RCT data and expert‐based algorithms 3,9,10 . An exciting prospect is the potential of using register data for machine learning.…”

“…The challenges posed by the complexity of ASM selection are well recognized, and selection algorithms have been developed based on expert opinions 3,9,10 . Real‐world data on ASM effectiveness would be of complementary benefit, but collection of long‐term randomized evidence is difficult, especially for particular subsets of patients.…”

The EpiPick algorithm to select appropriate antiseizure medications in patients with epilepsy: Validation studies and updates

“…Agreement between the app and the majority decision of the experts was higher than the agreement among the experts. 3 Ninetyfive percent of the experts considered that no incorrect or potentially harmful ASMs were ranked the highest by the application, and most experts strongly agreed with the app's selections. 3 Finally, in a more recent study, we conducted an external validation of the app using an independent real-life retrospective data set to assess whether ASMs recommended by the algorithm were associated with better outcomes than ASMs considered less desirable by the algorithm.…”

A web‐based algorithm to rapidly classify seizures for the purpose of drug selection

“…These seizure types and combinations of seizure types were then used to develop a therapeutic decision support application. 7,8 Each expert participating in the Delphi process suggested a set of questions to identify these seizure types. Questions comprised seizure semiology, as well as results of neuroimaging investigations.…”

“…6 However, such detailed differentiation is not required for ASM selection and a much simpler categorization is sufficient for therapeutic decisions. 7,8 A correct diagnosis of the main seizure types, and distinguishing between epileptic seizures and other paroxysmal events, is essential to plan proper therapy. 9 For example, the patient who stares while unresponsive could have either an absence seizure or a focal impaired awareness seizure, and treatment is quite different for the two.…”