Antiseizure medication prescribing in people with Dravet syndrome: An analysis of real‐time administrative data

Abstract

Since 2018, three new antiseizure medications (ASMs) received FDA approval for Dravet syndrome (DS) in the U.S: cannabidiol, stiripentol, and fenfluramine. Yet, the uptake of these ASMs in routine clinical practice is unknown. We use new ICD‐10 codes for DS (implemented in 2020) to estimate ASM receipt in patients with DS. We analyzed the TriNetX Network, a real‐time electronic health record‐based dataset linked to prescription data encompassing all 50 states of the U.S. After identifying patients with health care encounters for DS in 2021 and 2022 (via ICD‐10 codes), we examined ASM prescribing in the year following a DS claim: 2022 and 2023, respectively. We retrieved 387 and 451 patients receiving claims for DS in 2021 and 2022, respectively. Clobazam, diazepam, valproate, midazolam, clonazepam, levetiracetam, and cannabidiol were the most common ASMs used (29%–44%). Stiripentol and fenfluramine prescribing was limited (7%–16%); these two ASMs, considered second‐line therapies in DS, were prescribed less often than ASMs considered third‐line or beyond. Cannabidiol, stiripentol, and fenfluramine prescribing rates remained nearly identical in the 2021 and 2022 cohorts. Our data suggests that stiripentol, fenfluramine, and, to an extent, cannabidiol may be underused in a large, diverse, primarily U.S.‐based population of patients with DS.

Plain Language Summary

In an analysis of routinely‐collected health care claims in the U.S., we found that the uptake of new antiseizure medications for Dravet Syndrome (i.e., stiripentol, fenfluramine, and cannabidiol) has been limited since 2022. Even though stiripentol and fenfluramine are considered second‐line treatments for Dravet syndrome, we found they were prescribed less frequently than medicines considered third‐line or beyond. These findings raise concern for underutilization of new antiseizure medications for Dravet syndrome in the United States.

1. INTRODUCTION

Dravet syndrome (DS) is a rare epileptic encephalopathy characterized by drug‐resistant epilepsy and neurodevelopmental problems. ¹ Fortunately, three antiseizure medications (ASMs) have received FDA approval, since 2018, for DS treatment in the US: pharmaceutical‐grade cannabidiol, ² , ³ stiripentol, ⁴ , ⁵ and fenfluramine. ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ All three treatments are well tolerated, with cannabidiol resulting in 40% reductions ² , ³ and fenfluramine and stiripentol resulting in 70% reductions in monthly seizure frequency. ⁴ , ⁵ , ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ In 2022, a panel of experts and caregivers selected fenfluramine and stiripentol as 2nd‐line antiseizure therapy for DS, and cannabidiol as 3rd‐line (Figure 1) via a Delphi consensus. ¹¹

FIGURE 1.

Treatment algorithm for maintenance therapies for the management of seizures in Dravet syndrome. Adapted from Wirrell et al. 11. KGD, ketogenic diet.

It is unknown how many patients with DS are potentially accessing fenfluramine, stiripentol, and cannabidiol, and there is a need for up‐to‐date estimates of ASM receipt in patients with DS. Because recruitment for DS research is challenging by its low prevalence (1.2–6.5 per 100 000 people ¹² ), the analysis of data produced from the delivery of routine health care (“real‐world data”) ¹³ may help optimize outcomes for people with DS. In October 2020, diagnostic codes for DS were added to the International Classification of Diseases‐10 (ICD‐10), ¹⁴ representing an opportunity for epidemiologic surveillance of DS treatments using administrative data. In this study, we used electronic health record (EHR) data from more than 110 million people, predominantly in the U.S., to describe rates of ASM receipt among people receiving claims for DS in 2021 and 2022.

2. METHODS

We used TriNetX Analytics, a large international database (90% based in the US) of electronic health records (EHR) linked to claims and pharmacy data. ¹⁵ This analysis was exempt from human subjects review by the Carilion Clinic Institutional Review Board. The authors confirm that they read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

A detailed outline of the TriNetX data is included in eTable 1 and eFigure—Data S1. We identified DS using ICD‐10 claims (G40.83), requiring at least 2 claims for DS at any time point in the database following the first claim. The first DS claim was defined as an “index event”, occurring between January 1, 2021 and December 31, 2021 (cohort #1) for inclusion. A parallel and non‐overlapping cohort who received the claims between January 1, 2022 and December 31, 2022 (cohort #2) was also generated to assess changes in ASM uptake over time. Because the ICD‐10 claim for DS was not approved until October 2020, a claim for DS does not necessarily reflect the timing of patients' first‐ever diagnosis of DS; thus, the ICD‐10 claims for DS were intended to capture both individuals receiving long‐term treatment for DS and those diagnosed after 2020.

TriNetX Analytics uses privacy‐preserving record linkage techniques (i.e., cryptographic tokens) to link individuals across different health care organizations and insurance plans. Subsequent deduplication steps are taken to minimize the risk of identifying repeated patients across the two cohorts. ¹⁵ Detailed data quality assessments conducted by TriNetX Analytics (including assessments overseen by multiple independent programmers using different programming languages) have been previously described. ¹⁶ , ¹⁷ We evaluated individuals for inpatient or office‐based prescriptions of ASMs (see eTable 1—Data S1 for a full list) during the year following the index claim (see eFigure—Data S1). We ranked the most common ASMs received. We included people of all ages, sexes and genders, and races and ethnicities available in the TriNetX network.

3. RESULTS

The sample consists of two parallel, non‐overlapping cohorts: individuals, spanning all 50 states of the U.S., receiving claims for a DS claim in 2021 (n = 387, median age at time of first DS claim = 12.5 years, 72.9% with ASM scripts after first claim) and 2022 (n = 451, median age at claim = 12.2 years, 70.7% with ASM scripts after first claim). Both cohorts were approximately 50% male and predominantly non‐Hispanic and White (Table 1). In cohort #1 (claim for DS in 2021), the most common ASMs prescribed in the following year (2022) were clobazam (prescribed to 44.2% patients in this cohort), diazepam (41.6%), valproate (33.9%), midazolam (33.6%), clonazepam (32.0%), and levetiracetam (29.2%). Other common medications included cannabidiol (28.9%), lorazepam (19.4%), topiramate (18.3%), and fenfluramine (16.3%). Prescriptions for stiripentol, zonisamide, perampanel, brivaracetam, rufinamide, lacosamide, and barbiturates such as phenobarbital were less common (each <10%). This data is summarized in Table 1. Overall, similar estimates were computed for ASM receipt in cohort #2 (claim for DS in 2022) compared to patients with DS in cohort #1. Notably, rates of cannabidiol (28.9% in cohort #1; 27.7% in cohort #2), fenfluramine (16.3% vs. 16.0%), and stiripentol (7.0% vs. 7.5%) prescribing were nearly identical in the two cohorts.

TABLE 1.

Real‐World Prescription Medication Profiles Among People with Dravet Syndrome (DS), receiving claims in 2021 (cohort #1) and 2022 (cohort #2).

	Cohort #1 Receiving first claim for DS in 2021		Cohort #2 Receiving first claim for DS in 2022
Demographics of people receiving claims of DS	N = 387	%	N = 451	%
Median age at time of first claim for DS, years	12.5 ± 9.75	Range (0, 61)	12.2 ± 9.25	Range (0, 57)
Gender
Female	187	48.3%	219	48.6%
Male	197	50.9%	228	50.6%
Ethnicity
Not Hispanic or Latino	291	75.2%	341	75.6%
Unknown Ethnicity	37	9.6%	42	9.3%
Hispanic or Latino	59	15.2%	68	15.1%
Race (not mutually exclusive with ethnicity)
White	255	65.9%	280	62.1%
Black or African American	35	9.0%	48	10.6%
Other Race	26	6.7%	39	8.6%
Asian	18	4.7%	19	4.2%
Most common antiseizure medications received after DS claim
Clobazam	171	44.2%	209	46.3%
Diazepam	161	41.6%	179	39.7%
Valproate	131	33.9%	145	32.2%
Midazolam	130	33.6%	150	33.3%
Clonazepam	124	32.0%	137	30.4%
Levetiracetam	113	29.2%	132	29.3%
Cannabidiol	112	28.9%	125	27.7%
Lorazepam	75	19.4%	76	16.9%
Topiramate	71	18.3%	78	17.3%
Fenfluramine	63	16.3%	72	16.0%
Stiripentol	27	7.0%	34	7.5%
Zonisamide	24	6.2%	24	5.3%
Perampanel	18	4.7%	15	3.3%
Brivaracetam	17	4.4%	21	4.7%
Phenobarbital	17	4.4%	32	7.1%
Rufinamide	14	3.6%	<10	<2.2%
Lacosamide	11	2.8%	16	3.5%

Note: Per TriNetX's compliance to HIPAA, cells with 10 individuals or fewer are rounded to “<10.”

4. DISCUSSION

We provide descriptive data on patterns of ASM prescribing in people with DS, predominantly in the U.S., over the years of 2022 and 2023. Clobazam, diazepam, valproate, midazolam, clonazepam, levetiracetam, and cannabidiol are among the most commonly prescribed ASMs for DS (29%–44%). The use of stiripentol and fenfluramine was limited (7%–16%) and stable over both years studied. Although stiripentol and fenfluramine are considered second‐line therapies for DS, ¹¹ they were used less frequently than several ASMs considered third‐line or beyond (e.g., topiramate, levetiracetam) despite expert consensus published in 2022. ¹¹ Additionally, the receipt rates for cannabidiol, stiripentol, and fenfluramine remained nearly identical in 2022 and 2023.

There are several limitations stemming specifically from TriNetX's lack of refinement capabilities. First, TriNetX's de‐identified data lacks the ability to assess regional variation in uptake of ASMs. Other administrative databases (i.e., Merative MarketScan, CMS Medicaid, U.S. Veterans Affairs data) would permit us to refine our analyses and analyze in‐depth geographical and regional predictors of ASM uptake; however, such databases are also far more limited in generalizability than TriNetX, as they require continuous insurance enrollment, do not provide real‐time estimates, and lack the vast payer and geographic catchment of the TriNetX data. Second, we are unable to distinguish between inpatient versus outpatient medications, rescue (i.e., midazolam) versus chronic ASMs, and whether the first claim for DS necessarily equates the first diagnosis. Third, while our data suggests that relatively few individuals with DS are receiving new FDA‐approved treatments, we do not know exactly where individuals are in their treatment trajectory (i.e., whether they are newly‐diagnosed versus considered treatment‐resistant). Fourth, although we required at least two claims during the recruitment timeframe (a common approach used to increase the specificity of epilepsy codes ¹⁸ ), we cannot rule out possible misdiagnosis of DS. Fifth, due to the nature of our study design, we are not capable of capturing those patients who were temporarily on cannabidiol, stiripentol, and/or fenfluramine before 2022. Indeed, patients may have been diagnosed with DS before the ICD‐10 code for DS was adopted; however, we posit that it is not likely that most patients with DS were receiving cannabidiol, stiripentol, and fenfluramine (outside of a research trial) prior to having the ICD‐10 code entered, as insurance companies in the U.S. are often reluctant to approve these ASMs without the DS diagnosis. ¹⁴ In fact, the development of the ICD code for DS was motivated by the need to help people with DS gain access to cannabidiol, stiripentol, and fenfluramine, when they would otherwise be unable to obtain insurance reimbursement for these ASMs without an ICD code for DS in the chart. Finally, the possibility of duplicate medical records cannot be completely ruled out, although this risk is mitigated by TriNetX's intensive procedures for deduplication ¹⁵ and subsequent quality control. ¹⁶ , ¹⁷

Overall, we believe estimates from administrative databases are coarse metrics that need to be validated with additional clinical studies. Because ICD‐10 codes for DS are relatively new, in‐depth validation studies are pending. A recent chart review of patients with genetic epilepsies, including DS, demonstrated consistency between those identified via Natural Language Processing, ICD‐10 codes, and manual chart review at the Cleveland Clinic, ¹⁹ but multi‐site estimations of sensitivity, specificity, and predictive values are still needed. Further, it is possible that people who have an assigned DS‐specific ICD‐10 code are being cared for by clinicians who are knowledgeable about DS; because we may be missing DS diagnoses, our results may overestimate the rate at which people with DS gain access to fenfluramine, stiripentol, and cannabidiol, as opposed to the entire DS population in the US, where individuals may be undiagnosed or without access to the standard of care. Despite these limitations, leveraging TriNetX allowed us to study a large cohort of patients with DS from diverse sites, thus reducing selection bias.

4.1. Conclusions and implications

Delays in the adoption of consensus recommendations are not unique to DS (i.e., stroke care ²⁰ ) and signify a time‐sensitive need for rigorous dissemination, implementation, and health outcomes research. Overall, we believe the reasons for the underuse of newer DS ASMs, especially fenfluramine and stiripentol, are multifactorial, spanning prescriber‐, patient‐, family‐, organizational‐, and policy‐level barriers. At a policy and health systems level, stiripentol was not US FDA‐approved and available for compassionate use until 2019, and its current availability often remains limited to specialty pharmacies that require protocols outside standard prescribing practices. Preliminary 2023 data from specialty pharmacies stocking stiripentol raised concern that the medication was often not co‐prescribed (as intended) with clobazam and was commonly discontinued due to adverse effects and lack of efficacy. ²¹ Importantly, FDA guidelines suggest titration of stiripentol to 50 mg/kg/day, which may contribute to an increased side effect burden if clobazam and valproate are not dose‐reduced. Against this backdrop, clinicians may not be familiar with the complex dosing, drug–drug interactions, and side effects of stiripentol to feel comfortable prescribing it. Additionally, these three new ASMs may require special handling from an administrative standpoint, requiring above‐average out‐of‐clinic effort. Further, there may be additional financial burdens for patients even if the medications are covered by insurance.

Such organizational and policy‐level challenges are likely compounded by a myriad of patient‐, family‐, and clinician‐level barriers to care. Prior studies have raised concern that adult neurologists, including epilepsy specialists, often feel unprepared to treat patients with childhood‐onset seizure disorders transitioning from pediatric to adult neurology care. ²² In the case of DS specifically, caregivers of people with DS may encounter communication and education barriers with clinicians, especially adult physicians who were described as “less collaborative with caregivers than their pediatric counterparts.” ²³ In a recent caregiver‐based survey study of patients transferring from pediatrics to adult DS care, only half of participants reported that adult neurology/epilepsy teams were knowledgeable about DS in adults. ²⁴ Concern was also raised that adult neurologists seldom interfaced with multidisciplinary teams or addressed important issues related to guardianship and self‐advocacy capabilities. ²⁴ Such barriers to DS care during the pediatric‐to‐adult transition are well‐documented, but less is known about pediatric barriers to DS treatment, an important knowledge gap that warrants attention. While there are recent promising initiatives to deliver educational materials to clinicians and ancillary staff, ²⁵ more rigorous dissemination and implementation scholarship is needed to optimize DS care. Irrespective of the cause(s), we believe that all three medications should be used as often as necessary given the high‐quality evidence supporting these drugs' high efficacy, tolerability, meaningful improvements in quality of life for patients and caregivers, and potentially reduction in sudden unexpected death in epilepsy risk. ⁴ , ¹⁰

AUTHOR CONTRIBUTIONS

Dr. Binx Yezhe Lin had full access to all of the data in the study; Drs. Binx Yezhe Lin and Fábio A. Nascimento take responsibility for the integrity of the data and the accuracy of the data analysis. Concept and design: Fábio A. Nascimento. Acquisition, analysis, or interpretation of data: Fábio A. Nascimento, Kevin Y Xu, Binx Yezhe Lin, M. Scott Perry. Drafting of the manuscript: Kevin Y Xu, Fábio A. Nascimento. Critical revision of the manuscript for important intellectual content: Fábio A. Nascimento, Kevin Y Xu, Binx Yezhe Lin, M. Scott Perry. Statistical analysis: Binx Yezhe Lin. Obtained funding: Kevin Y Xu. Administrative, technical, or material support: Kevin Y Xu, Fábio A. Nascimento. Supervision: Kevin Y Xu, Fábio A. Nascimento.

FUNDING INFORMATION

American Psychiatric Association (Dr. Xu, Dr. Lin). National Institutes of Health (NIH K12 DA041449, Dr. Xu).

CONFLICT OF INTEREST STATEMENT

None of the authors has any conflict of interest to disclose.

ETHICS STATEMENT

We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

Supporting information

Data S1.

Data S2.

Xu KY, Lin BY, Perry MS, Nascimento FA. Antiseizure medication prescribing in people with Dravet syndrome: An analysis of real‐time administrative data. Epilepsia Open. 2025;10:336–341. 10.1002/epi4.13105

DATA AVAILABILITY STATEMENT

TriNetX data is proprietary and can be accessed via request at www.trinetx.com.