PDF: SKYLINE
(2026)Objective
To evaluate whether adjunctive soticlestat reduces convulsive seizure frequency vs placebo in children and young adults (2-21 y) with Dravet syndrome and treatment-resistant convulsive seizures.
Study Summary
• Responder rate (≥50% reduction) 27.4% soticlestat vs 9.9% placebo (nominal p=0.008); Care GI-I and CGI-I improvements favored soticlestat (nominal p≤.004).
• Safety: any TEAE 80.8% vs 74.6%; serious TEAEs 9.6% vs 14.1%; discontinuations 15.1% vs 5.6%; 1 SUDEP on soticlestat.
Intervention
Soticlestat (cholesterol-24-hydroxylase inhibitor) up to 300 mg BID (weight-adjusted) vs matching placebo, added to standard of care.
Inclusion Criteria
Age 2-21 y; weight ≥10 kg; clinical diagnosis of Dravet syndrome adjudicated by TESC; treatment-resistant convulsive seizures on 0-4 stable concomitant ASMs; ≥12 convulsive seizures in the 12 wk before screening and ≥4/28 days during the 4-6 wk screening.
Study Design
Arms: Soticlestat (n=73) vs Placebo (n=71)
Patients per Arm: 73 vs 71 (N=144 randomized; modified intent-to-treat = all randomized who received ≥1 dose and had ≥1 day of seizure data)
Outcome
• Primary (maintenance): -23.29% vs -11.99%; difference -14.29% (95% CI -30.51 to 1.53), p=0.089.
• ≥50% responder rate: 27.4% vs 9.9% (nominal p=0.008); >75% responders 10 vs 1.
• SCN1A+ subgroup (n=58 soticlestat vs 56 placebo): -23.34% vs -12.70%, diff -17.38% (95% CI -34.29 to -1.21, nominal p=0.045); responders 27.6% vs 8.9% (OR 4.31, 95% CI 1.32-14.02, nominal p=0.011).
• Care GI-I, CGI-I, CGI-I Seizure Intensity/Duration all favored soticlestat (nominal p≤.004); CGI-I Non-Seizure Symptoms and QI-Disability not different.
• Safety: any TEAE 80.8% (59/73) vs 74.6% (53/71); treatment-related TEAEs 43.8% vs 26.8%; serious TEAEs 9.6% (7/73) vs 14.1% (10/71); discontinuations for TEAE 15.1% (11/73) vs 5.6% (4/71); most common drug-related events with soticlestat somnolence 12.3%, change in seizure presentation 9.6%, decreased appetite 6.8%, insomnia 5.5%; 1 SUDEP on soticlestat.
Bottom Line
In this global phase 3 trial, adjunctive soticlestat narrowly missed statistical significance for its primary endpoint of convulsive seizure frequency reduction in Dravet syndrome (placebo-adjusted -15.64%, p=0.061), though the responder rate (≥50% reduction) and multiple caregiver/clinician-reported outcomes numerically favored soticlestat with nominal significance. Safety was consistent with prior data. The totality of the data suggest at most a modest antiepileptic effect that did not clear the phase 3 primary bar.
Major Points
- Global, multicenter (53 sites, 17 countries), 1:1 randomized, double-blind, placebo-controlled phase 3 trial in children and young adults (2-21 y) with Dravet syndrome; 144 randomized (soticlestat 73, placebo 71); enrollment Dec 7, 2021 - Apr 11, 2024.
- Weight-adjusted soticlestat titrated over 4 weeks to a maximum of 300 mg BID (in participants ≥45 kg), followed by 12 weeks of maintenance (16-week total treatment).
- Primary endpoint (full treatment): median % change in convulsive seizure frequency/28 d was -22.16% soticlestat vs -8.64% placebo; placebo-adjusted -15.64% (95% CI -31.30 to 0.24), p=0.061 - narrowly missed.
- Primary endpoint (maintenance, EMA co-primary): -23.29% vs -11.99%; placebo-adjusted -14.29% (95% CI -30.51 to 1.53), p=0.089 - also not significant.
- Key secondary responder rate (≥50% reduction) 27.4% soticlestat vs 9.9% placebo (nominal p=0.008); >75% responders 10 vs 1.
- Care GI-I, CGI-I, and CGI-I Seizure Intensity and Duration all favored soticlestat (all nominal p≤.004); CGI-I Non-Seizure Symptoms and QI-Disability showed no meaningful difference.
- SCN1A+ exploratory subgroup (79% of participants; n=58 soticlestat vs 56 placebo): -23.34% vs -12.70%, diff -17.38% (95% CI -34.29 to -1.21, nominal p=0.045); responders 27.6% vs 8.9% (OR 4.31, 95% CI 1.32-14.02, nominal p=0.011).
- Post-hoc ASM-quartile analysis: benefit largest in participants with fewer prior ASMs (Q1: -34.9%; Q2: -30.1%; Q3: -22.5%) and adverse in the most refractory quartile (Q4 with ≥9 ASMs: +15.0% vs placebo).
- Any TEAE 80.8% (soticlestat) vs 74.6% (placebo); treatment-related TEAEs 43.8% vs 26.8%; most common drug-related events with soticlestat were somnolence (12.3%), change in seizure presentation (9.6%), decreased appetite (6.8%), and insomnia (5.5%).
- Serious TEAEs 9.6% (7/73) soticlestat vs 14.1% (10/71) placebo; TEAE-related discontinuations 15.1% (11/73) vs 5.6% (4/71); one SUDEP occurred on soticlestat, no deaths on placebo; no new safety signals.
Study Design
- Study Type
- Phase 3, global, multicenter, 1:1 randomized, double-blind, placebo-controlled, parallel-group, add-on trial
- Randomization
- Yes
- Blinding
- Double-blind (participants, caregivers, investigators, and sponsor personnel blinded; matching placebo administered orally or via enteral feeding tube)
- Sample Size
- 144
- Follow-up
- 16-week treatment (4-week titration + 12-week maintenance); 1-week taper + 2-week follow-up for those not entering open-label extension
- Centers
- 53
- Countries
- 17 countries (global; specific list not enumerated in main text)
Primary Outcome
Definition: Percentage change from baseline in convulsive seizure frequency per 28 days during the full 16-week treatment period (and, for EMA registration, during the 12-week maintenance period)
| Control | Intervention | HR/OR | P-value |
|---|---|---|---|
| -8.64% (n=71) full treatment; -11.99% maintenance | -22.16% (n=73) full treatment; -23.29% maintenance | - (-31.30 to 0.24) | 0.061 |
Limitations & Criticisms
- Primary endpoint not met - the trial narrowly missed statistical significance for both the full treatment period (p=0.061) and the maintenance period (p=0.089), so hierarchical testing means all downstream 'positive' key secondary results are nominal only.
- Placebo response was substantially higher than in the phase 2 ELEKTRA study (which showed a 50% placebo-adjusted reduction), likely diluting the observable treatment effect in phase 3.
- Population was more treatment-refractory than ELEKTRA (median 8 prior/current ASMs, range up to 27); post-hoc quartile analysis suggests the drug performs worst in patients on ≥9 ASMs (+15% vs placebo), raising questions about the target population.
- SCN1A genotype status was based on documented mutations - the ~20% classified as SCN1A- may include misdiagnosed patients and confounded the primary analysis; the SCN1A+ signal is exploratory and not adjusted for multiplicity.
- Sponsor (Takeda) designed, funded, and analyzed the trial, and 7 of 15 authors are Takeda employees/stockholders - risk of interpretation bias in framing a negative primary result as 'clinically meaningful.'
- Short 16-week treatment duration (4-week titration + 12-week maintenance) may not fully capture durability or long-term tolerability, particularly important for a chronic pediatric epilepsy.
- Post-hoc ASM-quartile analysis was not prespecified and is hypothesis-generating.
- Country/site-specific effects and cross-cultural differences in caregiver-reported outcomes (Care GI-I, CGI-I) across 17 countries were not detailed.
Citation
Epilepsia 2026;67(6):2796-2807. DOI: 10.1002/epi.70164