← Back

SKYWAY

A phase 3, randomized clinical trial of soticlestat as adjunctive therapy for Lennox–Gastaut syndrome

Year of Publication: 2026

Authors: Guerrini R, Marsh ED, Liao WP, ..., Murthy V

Journal: Epilepsia

Citation: Guerrini R, et al. A phase 3, randomized clinical trial of soticlestat as adjunctive therapy for Lennox–Gastaut syndrome. Epilepsia. 2026. doi:10.1002/epi.70216

Link: https://doi.org/10.1002/epi.70216

Clinical Question

Does soticlestat, a first-in-class cholesterol 24-hydroxylase (CH24H) inhibitor, reduce major motor drop seizure frequency as adjunctive therapy in patients with Lennox–Gastaut syndrome?

Bottom Line

Soticlestat did not demonstrate efficacy over placebo in reducing major motor drop seizure frequency in LGS; inhibition of CH24H should not be pursued as a pharmacotherapeutic strategy for this condition.

        Major Points
        The primary endpoint was not met: soticlestat produced no meaningful reduction in major motor drop (MMD) seizure frequency compared to placebo (−1.17% difference, p = .785 over full treatment; 2.43% difference, p = .778 over maintenance).
Hierarchical testing was stopped at the primary endpoint; all key secondary efficacy endpoints are therefore interpreted as nominal only.
Nominal signals were seen for CGI-I Disruptive Behaviors (OR 1.91, p = .032) and CGI-I Seizure Intensity and Duration (OR 1.67, p = .029), but these were not adjusted for multiplicity.
TEAEs were slightly more frequent with soticlestat (74.6%) than placebo (68.4%), mostly mild to moderate; somnolence was the most frequent soticlestat-related TEAE.
Serious treatment-related TEAEs occurred in 3 soticlestat-treated vs 1 placebo-treated participant.
95% of participants were on ≥2 background ASMs, reflecting the refractory nature of the LGS population studied.
The results suggest that CH24H inhibition is not a suitable pharmacotherapeutic target for LGS, despite earlier phase 2 signals.

      

Design

Study Type: Phase 3, global, multicenter, randomized, double-blind, placebo-controlled, parallel-group

Randomization: 1

Blinding: Double-blind (soticlestat and placebo tablets were indistinguishable in number and type)

Allocation: 1:1 randomization via interactive web response system; stratified by age group (≤6 vs >6 years) and country (China, Japan, rest of world)

Enrollment Period: November 2021 – January 2024

Follow-up Duration: 16-week treatment period (4-week titration + 12-week maintenance); optional open-label extension thereafter; 1-week taper + 2-week safety follow-up for discontinuers

Centers: 84

Countries: China, Japan, United States, Italy, Serbia, Hungary, Canada, France, and 10 additional countries (18 total)

Sample Size: 270

Analyzed: 270

Analysis: Modified intent-to-treat (mITT): all randomized participants who received ≥1 dose and were assessed for seizures ≥1 day during treatment. Rank ANCOVA for primary endpoint (Hodges–Lehmann estimator); Cochran–Mantel–Haenszel for responder rate; ordinal logistic regression for CGI-I outcomes; mixed model for repeated measures for QI-Disability.

Power Calculation: Sample size based on phase 2 study data and dose selection rationale (detailed in Methods S1)

Registration: NCT04938427; EudraCT 2021-002481-40

Inclusion Criteria

Age 2–55 years
Body weight ≥10 kg at screening
Documented clinical diagnosis of LGS adjudicated by the Epilepsy Study Consortium, supported by: onset of seizures between ages 1–8 years; presence of multiple seizure types; history of abnormal EEG; developmental delay or intellectual disability consistent with LGS
≥8 MMD seizures per month during the 3 months preceding screening
≥8 MMD seizures per 28 days during the prospective baseline period (4–6 weeks)

Exclusion Criteria

Current enrollment in a clinical trial of a different investigational drug
Participation in a clinical trial of a different study drug in the past 30 days before screening
Prior receipt of soticlestat
Admission to a medical facility and intubation for treatment of status epilepticus ≥2 times in the 3 months before screening
Use of strong CYP3A inducers (except ASMs such as carbamazepine, phenobarbital, phenytoin, and topical preparations)
Use of herbal preparations as ASMs

Arms

Field	Control	Soticlestat
N	136	134
Intervention	Matching placebo tablets (indistinguishable from soticlestat) twice daily added to stable background ASMs	Soticlestat (TAK-935) ≤300 mg twice daily, weight-adjusted (20-mg minitablets for <45 kg; 20-mg minitablets or 100-mg tablets titrated to maximum 300 mg for ≥45 kg), added to stable background ASMs
Duration	16 weeks (4-week titration + 12-week maintenance)	16 weeks (4-week titration + 12-week maintenance)

Outcomes

Outcome	Type	Control	Intervention	HR / OR / RR	P-value
Percentage change from baseline in major motor drop (MMD) seizure frequency per 28 days during the full 16-week treatment period (titration + maintenance); MMD seizure defined as a seizure resulting in or likely to result in a fall due to involvement of major body areas	Primary	Not reported as standalone arm value	Not reported as standalone arm value		Full treatment: p = .785; Maintenance: p = .778
	Secondary
	Secondary
	Secondary
	Secondary			1.91	nominal p = .032
	Secondary			1.67	nominal p = .029
	Secondary
	Secondary
	Secondary
	Secondary
	Safety
	Safety
	Safety
74.6%					Adverse
68.4%					Adverse
Somnolence					Adverse
3 participants					Adverse
1 participant					Adverse
Mostly mild or moderate					Adverse

Subgroup Analysis

Pharmacodynamic biomarker (plasma 24S-hydroxycholesterol) was measured at screening, Day 57, and Day 113 to estimate average 24HC reduction during a dosing interval per participant; subgroup results not reported in available source text.

Criticisms

The primary endpoint selection was based on a post-hoc analysis of the phase 2 ELEKTRA study (drop seizures leading to or likely to lead to a fall), introducing risk of bias in endpoint definition.
Phase 2 efficacy signals were modest and did not reach statistical significance in the LGS subgroup (placebo-adjusted median reduction 17.08%, p = .1160), raising questions about whether the phase 3 was adequately powered for a drug with likely small effect sizes.
The heterogeneous LGS population (highly varied etiology, age range 2–55 years, polytherapy burden) may have increased outcome variability and diluted any treatment signal.
High background polytherapy (95% on ≥2 ASMs) limits the ability to detect incremental benefit from adjunctive soticlestat.
Nominal secondary signals (disruptive behaviors, seizure intensity) could reflect chance findings and should not be interpreted as evidence of efficacy given primary endpoint failure and lack of multiplicity adjustment.

Funding

Takeda Development Center Americas, Inc. (sponsor and employer of several co-authors)

Based on: SKYWAY (Epilepsia, 2026)

Authors: Guerrini R, Marsh ED, Liao WP, ..., Murthy V

Citation: Guerrini R, et al. A phase 3, randomized clinical trial of soticlestat as adjunctive therapy for Lennox–Gastaut syndrome. Epilepsia. 2026. doi:10.1002/epi.70216

Content summarized and formatted by NeuroTrials.ai.

SKYWAY

(2026)

Objective

To investigate the efficacy and safety of soticlestat (TAK-935), a first-in-class cholesterol 24-hydroxylase (CH24H) inhibitor, as adjunctive therapy in children and adults with Lennox–Gastaut syndrome (LGS), a rare and treatment-resistant developmental epileptic encephalopathy.

Study Summary

• Soticlestat failed to reduce major motor drop (MMD) seizure frequency versus placebo: placebo-adjusted median difference −1.17% (95% CI −13.02 to 9.99, p = .785) over the full 16-week treatment period and 2.43% (95% CI −10.86 to 15.14, p = .778) over the 12-week maintenance period
• No meaningful difference was observed for most key secondary endpoints under prespecified hierarchical testing
• Numerical trends only favored soticlestat for CGI-I Disruptive Behaviors (OR 1.91, nominal p = .032) and CGI-I Seizure Intensity and Duration (OR 1.67, nominal p = .029)
• TEAEs occurred in 74.6% of soticlestat- vs 68.4% of placebo-treated participants; most frequent soticlestat TEAE was somnolence; serious TEAEs in 3 vs 1 participant
• Targeting CH24H does not appear to be an effective pharmacotherapeutic strategy for LGS

Intervention

Soticlestat (TAK-935) ≤300 mg twice daily (weight-adjusted, 20-mg minitablets for <45 kg; up to 300 mg for ≥45 kg) added to stable background antiseizure medications over 16 weeks (4-week titration + 12-week maintenance)

Inclusion Criteria

Age 2–55 years; weight ≥10 kg; adjudicated clinical diagnosis of LGS; ≥8 MMD seizures per month in the 3 months before screening; ≥8 MMD seizures per 28 days during the prospective baseline period

Study Design

Arms: Soticlestat ≤300 mg twice daily (n=134) vs Placebo (n=136), both added to stable background ASMs

Patients per Arm: Soticlestat: 134; Placebo: 136

Outcome

• Primary (not met): MMD seizure frequency difference −1.17% (95% CI −13.02 to 9.99, p = .785) over full treatment period; 2.43% (95% CI −10.86 to 15.14, p = .778) over maintenance — no meaningful difference
• Nominal secondary trends: CGI-I Non-seizure Symptoms Disruptive Behaviors OR 1.91 (95% CI 1.06–3.43, nominal p = .032); CGI-I Seizure Intensity and Duration OR 1.67 (95% CI 1.06–2.65, nominal p = .029)
• Safety: TEAEs 74.6% soticlestat vs 68.4% placebo; somnolence most common; serious TEAEs 3 vs 1

Bottom Line

Major Points

The primary endpoint was not met: soticlestat produced no meaningful reduction in major motor drop (MMD) seizure frequency compared to placebo (−1.17% difference, p = .785 over full treatment; 2.43% difference, p = .778 over maintenance).
Hierarchical testing was stopped at the primary endpoint; all key secondary efficacy endpoints are therefore interpreted as nominal only.
Nominal signals were seen for CGI-I Disruptive Behaviors (OR 1.91, p = .032) and CGI-I Seizure Intensity and Duration (OR 1.67, p = .029), but these were not adjusted for multiplicity.
TEAEs were slightly more frequent with soticlestat (74.6%) than placebo (68.4%), mostly mild to moderate; somnolence was the most frequent soticlestat-related TEAE.
Serious treatment-related TEAEs occurred in 3 soticlestat-treated vs 1 placebo-treated participant.
95% of participants were on ≥2 background ASMs, reflecting the refractory nature of the LGS population studied.
The results suggest that CH24H inhibition is not a suitable pharmacotherapeutic target for LGS, despite earlier phase 2 signals.

Study Design

Study Type: Phase 3, global, multicenter, randomized, double-blind, placebo-controlled, parallel-group
Randomization: Yes
Blinding: Double-blind (soticlestat and placebo tablets were indistinguishable in number and type)
Sample Size: 270
Follow-up: 16-week treatment period (4-week titration + 12-week maintenance); optional open-label extension thereafter; 1-week taper + 2-week safety follow-up for discontinuers
Centers: 84
Countries: China, Japan, United States, Italy, Serbia, Hungary, Canada, France, and 10 additional countries (18 total)

Primary Outcome

Definition: Percentage change from baseline in major motor drop (MMD) seizure frequency per 28 days during the full 16-week treatment period (titration + maintenance); MMD seizure defined as a seizure resulting in or likely to result in a fall due to involvement of major body areas

Control	Intervention	HR/OR	P-value
Not reported as standalone arm value	Not reported as standalone arm value	- (Full treatment: −13.02 to 9.99; Maintenance: −10.86 to 15.14)	Full treatment: p = .785; Maintenance: p = .778

Limitations & Criticisms

The primary endpoint selection was based on a post-hoc analysis of the phase 2 ELEKTRA study (drop seizures leading to or likely to lead to a fall), introducing risk of bias in endpoint definition.
Phase 2 efficacy signals were modest and did not reach statistical significance in the LGS subgroup (placebo-adjusted median reduction 17.08%, p = .1160), raising questions about whether the phase 3 was adequately powered for a drug with likely small effect sizes.
The heterogeneous LGS population (highly varied etiology, age range 2–55 years, polytherapy burden) may have increased outcome variability and diluted any treatment signal.
High background polytherapy (95% on ≥2 ASMs) limits the ability to detect incremental benefit from adjunctive soticlestat.
Nominal secondary signals (disruptive behaviors, seizure intensity) could reflect chance findings and should not be interpreted as evidence of efficacy given primary endpoint failure and lack of multiplicity adjustment.

Citation

Guerrini R, et al. A phase 3, randomized clinical trial of soticlestat as adjunctive therapy for Lennox–Gastaut syndrome. Epilepsia. 2026. doi:10.1002/epi.70216

View Original Publication →

SKYWAY

A phase 3, randomized clinical trial of soticlestat as adjunctive therapy for Lennox–Gastaut syndrome

Clinical Question

Bottom Line

Major Points

Design

Inclusion Criteria

Exclusion Criteria

Arms

Outcomes

Subgroup Analysis

Criticisms

Funding

Ask about SKYWAY