Dravet Syndrome and CBD Oil: Research, Dosing and Warnings

What does the research say?

In two large studies, children and young people with Dravet syndrome were randomly assigned to groups: one group took the CBD medicine on top of their existing medicines, the other took a placebo; neither families nor doctors knew who was taking which. In the first study, after 14 weeks, the number of convulsive seizures per month (median) fell from 12.4 to 5.9 in children taking CBD, while it fell from 14.9 to 14.1 in the placebo group. In the second study, seizures fell by 48.7% and 45.7% in the CBD groups and by 26.9% in the placebo group.

Becoming completely seizure-free was rare (5% in the CBD group and 0% in the placebo group in one study), and people on placebo also improved. The studies lasted 14 weeks, CBD was given as an add-on to treatment, and the studies were funded by the drug's manufacturer. The European Medicines Agency (EMA) states that the effect has not been proven in patients who do not take a medicine called clobazam. When the same studies were re-analyzed by clobazam use, CBD was better than placebo in people not taking clobazam too; the average reduction in seizure count was smaller in these patients, but the difference from placebo in the share of patients whose seizures fell by at least half was similar in the two groups. In 89 children and young people taking both stiripentol and clobazam, the result numerically favored CBD, but the difference was not significant. Because whether patients took these medicines was not decided at random, these comparisons are not conclusive.

In a study of 34 children aged 4–10 that mainly measured safety, different doses were compared with placebo. In some children, CBD raised the blood level of the active product of another epilepsy medicine (clobazam); liver values rose in 6 children who were also taking valproate, and all returned to normal. That is why monitoring with blood tests during treatment is important.

These results cannot be applied to CBD oils whose content and dose are unknown. In oils on the market, the amount on the label may not be reliable; the product may contain THC or contaminants, and there are reports of seizures getting worse with these products. There is no controlled study of unapproved cannabis oils in Dravet syndrome. In a retrospective review in the US based on parent reports, the share of children with Dravet whose seizures fell by more than half was 23%; the products' content was not verified, there was no comparison group, and at the same center, families with Dravet stopped these products sooner than other patients. In a study with no comparison group in which a cannabis oil produced under controlled conditions and also containing a small amount of THC (CBD:THC 50:1) was tried in 20 children with Dravet, seizures decreased; this small, uncontrolled study is not enough to show whether THC provides any benefit (evidence for THC-containing products is insufficient).

In an extension study that followed 315 patients who continued the medicine after the study ended for up to three years, the reduction in convulsive seizures continued; but there was no comparison group, only those who completed the first study and wanted to continue were enrolled, and about half of the patients left before completing the study. In everyday-practice data, mostly from Europe, the share of patients with Dravet whose seizures fell by at least half ranged from 22% to 51% across studies. In a study in France and in a multicenter international study, the response in patients with Dravet was no different from that in other severe epilepsies; in a two-center study in Italy, it was lower. In a study from Italy on long seizures (status epilepticus), an important problem in Dravet, fenfluramine use was linked to fewer episodes; for CBD, no significant link with episode duration was found. In a study of adults with Dravet in England, seizures fell by more than 30% in only 3 of the 17 patients assessed, and liver enzymes rose in more than half of the patients. None of these studies had a randomized comparison group; so they show the course of treatment in everyday practice, not the size of the effect.

There are other medicines approved for Dravet as well (fenfluramine, stiripentol). There is no study comparing them directly with each other; in two indirect comparisons that combined each one's placebo-controlled studies, the share of patients whose seizures fell by at least half was lower with CBD than with fenfluramine and stiripentol, and the result did not change when looking only at people using it together with clobazam, in line with the approval in Europe. Comparisons like these are not conclusive: the stiripentol studies were small and done years ago, some of the patients in the CBD studies were already taking stiripentol, and one of the analyses was funded by the maker of stiripentol. The picture for side effects was different: in one analysis, the share of patients with side effects was lower with CBD than with the other two medicines, but stopping treatment was more common with CBD than with stiripentol.

Strength of evidence: ●●●● Strong evidence. Finding: Proven effect. What does this mean?

Ongoing research on this topic

Trials testing CBD for this topic on ClinicalTrials.gov: 4 under way. A trial being under way does not show that CBD works.

NCT05044819: Phase 4, active, not recruiting, 154 participants
NCT06924827: Phase 4, not yet recruiting, target 25 participants
NCT06738732: Phase 1/2, not yet recruiting, target 25 participants

Registries in other countries also list 1 recruiting trial on this topic: other registries.

All trials on this topic and their history

Possible side effects and drug interactions

The most common side effects were sleepiness, decreased appetite, diarrhea, vomiting, fever and tiredness. Sleepiness occurred in 32% of people taking CBD and 11% of those taking placebo. In some patients, liver enzymes rose; this was more common at high doses and when used together with a medicine called valproate. In the extension study lasting up to three years, liver enzymes rose markedly in 22% of patients, and most of them (84%) were taking valproate.

CBD can slow down how some medicines are broken down in the body. For example, it can raise the level of the active substance that clobazam turns into in the body by about 3 times, and this increases sleepiness; levels of medicines such as stiripentol and everolimus can also rise. Because seizures can increase if CBD is stopped suddenly, treatment is stopped gradually.

Interactions with medicines used for this condition

Interactions reported for these medicines in peer-reviewed publications and official prescribing information. A medicine that is not listed can still interact with CBD; tell your doctor about every medicine you take.

Dravet Syndrome: medicines used and their interactions with CBD
Medicine	Importance	Warning	Source
Clobazam	High importance	Clobazam: If you take clobazam, CBD can raise the medicine's active metabolite about 3-fold and can increase drowsiness. Do not start without talking to your neurologist.	FDA Epidiolex label §7.2, §12.3 (05/2026), EMA prescribing information §4.5, PMID 26114620
Valproate (valproic acid) (Depakin, Convulex)	High importance	Valproate and the liver: Liver enzyme elevations were more common when CBD was used together with valproate. Your doctor should monitor your liver tests.	DailyMed Epidiolex label 5.1, PMID 28782097
Stiripentol, Topiramate, Rufinamide, Zonisamide, Eslicarbazepine	Caution	Other epilepsy medicines: With epilepsy medicines, CBD can change the blood levels of some of them. Your doctor may monitor your drug levels and side effects.	PMID 28782097, PMID 33541771, PMID 37593907, PMID 31211851, PMID 30791225
Carbamazepine (Tegretol)	Caution	Strong enzyme inducers: These medicines can lower the level of CBD in the blood: with rifampin it fell by about one third. There are no direct measurements with others such as carbamazepine; the prescribing information expects a similar drop. Tell your doctor.	DailyMed Epidiolex label 7.1, EMA Epidyolex SmPC 4.5

To check interactions with the medicines you take: CBD drug interaction checker.

How was CBD used in studies?

In studies and in the labeling, the dose is calculated as milligrams per day for each kilogram of body weight. Large studies used 10 or 20 mg per kilogram per day (study dose; not a personal recommendation), and the daily amount was split into two equal parts; the two doses had a similar effect, and side effects were fewer at the lower dose. Under the labeling, the dose is adjusted slowly over weeks under a doctor's supervision; the upper limit is 20 mg per kilogram per day (study dose; not a personal recommendation).

Because a fatty meal markedly increases the amount that reaches the blood, the medicine is always taken the same way in relation to food. Because it contains sesame oil, it is not used in people who are hypersensitive to sesame oil.

Study details (for health professionals)

In two double-blind RCTs (GWPCARE1 and GWPCARE2), prescription pure CBD added to existing medicines reduced convulsive seizures more than placebo over 14 weeks. 10 and 20 mg/kg/day had a similar effect; 10 mg/kg/day was better tolerated. Complete seizure freedom was rare (5% vs. 0% in GWPCARE1), and seizures also decreased in the placebo group. In a dose-ranging safety RCT with 5–20 mg/kg/day in 34 children aged 4–10, CBD levels rose in proportion to the dose, clobazam's active metabolite rose, and liver enzymes rose in 6 patients taking valproate (all returned to normal). The efficacy studies were all 14 weeks long, add-on treatment and manufacturer-funded; the EMA writes that efficacy has not been proven in patients not taking clobazam. The Stockings 2018 review in the table found CBD superior to placebo for a reduction in seizures of at least half, but rated the certainty of the evidence as low (GRADE); this assessment was based on a search up to October 2017 and, for this outcome, on two RCTs (291 patients). Later meta-analyses combining four RCTs (Lattanzi 2018; Lattanzi 2020, clobazam status) and the meta-analysis combining the three RCTs in Dravet only (Lattanzi 2020, Dravet meta-analysis: seizure reduction of at least half RR 1.69; 95% CI 1.21–2.36; stopping treatment 9.0% vs. 2.3%) gave results in the same direction; the guide's "strong" assessment rests on this whole body of evidence. The latest version of the Cochrane review of medicines in Dravet (2017) predates the CBD studies, so it does not include CBD. In analyses that split the same four RCTs by clobazam use, CBD was better than placebo in patients not taking clobazam too. The reduction in seizure count was smaller in patients without clobazam (both conditions and both doses combined, the ratio of seizure frequency vs. placebo was 0.59 with clobazam and 0.85 without clobazam; looking at the two doses separately, the difference was not significant without clobazam; manufacturer analysis, Devinsky 2020); in the share of patients whose seizures fell by at least half, the difference from placebo was similar in the two groups (RR 1.85 and 1.80 in an independent meta-analysis; Lattanzi 2020, clobazam status). In 89 patients with Dravet taking both stiripentol and clobazam, CBD's effect was numerically favorable but not significant (Devinsky 2020). Because whether patients took these medicines was not decided at random, these comparisons are not conclusive. In two network meta-analyses comparing Dravet medicines (there are no studies comparing the medicines directly; the comparisons are indirect), the share of patients whose seizures fell by at least half was lower with CBD than with fenfluramine and stiripentol (vs. fenfluramine OR 0.20; 95% CI 0.07–0.54; Lattanzi 2023); the result was the same when looking only at patients using it together with clobazam (Guerrini 2024). The certainty of these comparisons is limited: stiripentol rests on two small studies (64 children) done an estimated 15–20 years ago, 33–49% of patients in the CBD studies were already taking stiripentol, and one of the analyses was funded by the maker of stiripentol. The picture for side effects is different: in one analysis, the share of patients with side effects was lower with CBD than with fenfluramine and stiripentol, while stopping treatment was more common with CBD than with stiripentol (Lattanzi 2023). In the open-label extension of the two RCTs, the median reduction in convulsive seizures in 315 patients stayed between 45% and 74% in 12-week periods up to 156 weeks; according to the final results on ClinicalTrials.gov (registry, not a peer-reviewed publication), 170 of the patients completed the study and the median reduction in the last 12 weeks was 40%; liver enzymes rose markedly in 22% of patients, 84% of whom were taking valproate (Scheffer 2021). In everyday-practice studies, mostly from Europe, the share of patients with Dravet whose seizures fell by at least half ranged from 22% to 51%; in a study in adults in England, seizures fell by more than 30% in only 3 of 17 patients, and in the 15 patients using it without clobazam the rate was 13–21%; these studies have no comparison group, and in most of them some authors have financial ties to the manufacturer. There is no controlled study of unapproved cannabis products in Dravet. October 2026 search (PubMed, Cochrane, ClinicalTrials.gov): 3 placebo-controlled RCTs with cannabinoids on this topic (one a dose study mainly measuring safety and drug levels) and 6 systematic reviews (2 of them indirect comparisons with other Dravet medicines) were found; these are in the table below together with a manufacturer-supported pooled analysis of the same studies, 10 studies without randomized comparison (long-term extension, access programs, everyday practice, and unapproved or THC-containing products) and 3 drug-level studies measuring drug interactions (two placebo-controlled in patients with epilepsy, one in healthy volunteers). No controlled study with THC-containing products in Dravet was found.

Evidence assessment: Strong (double-blind, placebo-controlled RCTs + meta-analyses)

CBD alone

Approved, purified medicinal CBD (Epidiolex/Epidyolex) was found effective as an add-on treatment in Dravet syndrome (two double-blind RCTs and meta-analyses; strong evidence); the effect continued in long-term follow-up with no comparison group, and in indirect comparisons with limited certainty against the other medicines approved for Dravet (fenfluramine, stiripentol), its effect on reducing seizures looked smaller; this result cannot be applied to CBD oils whose content and dose are unknown.

Products containing THC

There is no controlled study with THC-containing products in Dravet syndrome. In the only prospective study specific to Dravet, a cannabis oil also containing a small amount of THC (CBD:THC 50:1) was tried in 20 children for 20 weeks and seizures decreased (median 70.6% for motor seizures); but because there was no placebo group and the contribution of THC could not be separated from that of CBD, this result does not show a benefit of THC. The study was largely funded by the company that makes the product.

Evidence: Insufficient: no controlled study with a THC-containing product in Dravet; there is only one small open-label study of 20 children with no comparison group

Approved drug regimen

FDA (≥1 year, no clobazam requirement): start at 2.5 mg/kg twice daily (5 mg/kg/day); after 1 week, maintenance dose 5 mg/kg twice daily (10 mg/kg/day). If further seizure reduction is needed, increase weekly by 2.5 mg/kg twice daily up to a maximum of 10 mg/kg twice daily (20 mg/kg/day); for faster titration, increase no more often than every other day. Label: 20 mg/kg/day gives somewhat greater seizure reduction but more side effects. EMA (≥2 years, add-on treatment only together with clobazam): 5 mg/kg/day for 1 week, then maintenance 10 mg/kg/day; increases above 10 mg/kg/day based on individual benefit-risk and with the full monitoring schedule, maximum 20 mg/kg/day. Hepatic impairment: moderate (Child-Pugh B) start at 1.25 mg/kg twice daily, severe (Child-Pugh C) 0.5 mg/kg twice daily. Stopping is done gradually. Source: FDA label (05/2026) §2.1–2.2, §2.5–2.6; EMA product information §4.1–4.2 (last updated in the EPAR on August 20, 2026). Turkey: Epidyolex/cannabidiol is not on the TİTCK (Turkish Medicines and Medical Devices Agency) list of approved products (October 16, 2026); on the TİTCK Foreign Active Ingredient List (September 28, 2026), cannabidiol oral solution (YDN02BG1099) is listed as a red-prescription drug and among the medicines that cannot be imported without TİTCK's written approval.

Dose range in studies

10–20 mg/kg/day in the labeling and RCTs (divided into two equal doses); up to a ceiling of 25–50 mg/kg/day in the open-label expanded access program; at most 30 mg/kg/day (mean 22 mg/kg/day) in the open-label extension study; in everyday-practice studies in Europe, the median or mean dose was mostly 11–16 mg/kg/day

Starting dose and titration

Labeling: start at 5 mg/kg/day, 10 mg/kg/day after 1 week; if needed, increase by 5 mg/kg/day weekly up to a maximum of 20 mg/kg/day. In the RCTs, a 2-week titration (10 mg/kg/day reached in 7 days, 20 mg/kg/day in 11 days), followed by 12 weeks of maintenance

These are doses researchers gave under medical supervision; they are not a recommended dose for you. "mg/kg/day" means milligrams per kilogram of body weight per day.

Dravet Syndrome and CBD: studies reviewed
Study	Participants	Product	Dose and titration	Duration	Result	Source
Devinsky 2017 (GWPCARE1) Double-blind, placebo-controlled RCT	120 Dravet syndrome, aged 2–18 (CT.gov eligibility criterion)	Oral CBD solution (GW; pure CBD)	20 mg/kg/day 2 weeks (20 mg/kg/day reached in 11 days), then 12 weeks of maintenance; preceded by a 4-week drug-free baseline observation (CT.gov NCT02091375; EMA product information §5.1)	14 weeks	Median monthly convulsive seizures 12.4 → 5.9 (placebo 14.9 → 14.1); adjusted median difference −22.8 points (p=0.01). ≥50% response 43% vs. 27% (p=0.08); seizure-free 5% vs. 0% (p=0.08). No significant reduction in non-convulsive seizures Side effects: Diarrhea, vomiting, tiredness, fever, sleepiness, abnormal liver tests; more study discontinuations in the CBD arm	PMID 28538134, doi:10.1056/NEJMoa1611618, NCT02091375, PMID 34265088, on day 12 at 20 mg/kg/day (nominal p = 0.02) and on day 13 at 10 mg/kg/day (p = 0.03), most of the most common side effects (sleepiness, decreased appetite, diarrhea) resolved within 4 weeks. Devinsky 2024, PMID 38677101, funded by the manufacturer): in 124 placebo patients, seizure counts fell from baseline, but seizure-free days and the caregiver's global impression barely changed
Miller 2020 (GWPCARE2) Double-blind, placebo-controlled, dose-comparison RCT	199 randomized (198 eligible) Dravet syndrome, aged 2–18	Pharmaceutical CBD solution (pure CBD)	10 or 20 mg/kg/day, 2 equal doses 2 weeks: started at 2.5 mg/kg/day; increased to 10 mg/kg/day on day 7 and to 20 mg/kg/day on day 11	14 weeks	Reduction in convulsive seizures 48.7% (10 mg) and 45.7% (20 mg), placebo 26.9%; vs. placebo 29.8% (95% CI 8.4–46.2; p=0.01) and 25.7% (95% CI 2.9–43.2; p=0.03). ≥50% reduction 43.9% and 49.3%, placebo 26.2% (OR 2.21 and 2.74); convulsive seizure-free throughout treatment: 2 patients at 10 mg, 2 at 20 mg, 1 on placebo. No clear difference between groups in sleep, daytime sleepiness, quality of life, adaptive behavior or cognitive function (few patients completed these measures). The manufacturer (GW Research) funded and ran it; all 10 authors have ties to the manufacturer (3 are employees) Side effects: Decreased appetite, diarrhea, sleepiness, fever, tiredness. Raised transaminases in 13 patients at 20 mg and 3 at 10 mg; all were using valproate. 5 discontinuations in the 20 mg arm	PMID 32119035, doi:10.1001/jamaneurol.2020.0073, PMID 32391881
Devinsky 2018 (dose-ranging safety study) Multicenter, double-blind, placebo-controlled dose-ranging RCT (4:1)	34 (5 mg/kg 10, 10 mg/kg 8, 20 mg/kg 9, placebo 7) Children aged 4–10 with Dravet syndrome	Pharmaceutical-grade pure CBD (GWP42003-P)	5, 10 or 20 mg/kg/day, divided into 2 doses a day 3 weeks of treatment including titration, tapering over 10 days	3 weeks	Safety and pharmacokinetics study: CBD and its metabolites rose in proportion to the dose; clobazam's active metabolite (N-desmethylclobazam) rose (except in those taking stiripentol); levels of other antiseizure medicines did not change Side effects: Most common fever, sleepiness, decreased appetite, sedation, vomiting, balance problems, behavior change; liver enzymes rose in 6 patients taking valproate, and all returned to normal	PMID 29540584, doi:10.1212/WNL.0000000000005254, NCT02091206
Lattanzi 2018 Systematic review and meta-analysis (4 RCTs)	550 LGS and Dravet syndrome	Medicinal CBD (pure)	10 and 20 mg/kg/day		Difference in seizure frequency vs. placebo 19.5 points (10 mg/kg/day) and 19.9 points (20 mg/kg/day). ≥50% reduction in all seizures 37.2% vs. 21.2% (20 mg; RR 1.76) Side effects: Stopping because of side effects 8.9% vs. 1.8% (RR 5.59); RR 6.89 at 20 mg, not significant at 10 mg	PMID 30390221, doi:10.1007/s40265-018-0992-5
Lattanzi 2020 (clobazam status) Systematic review and meta-analysis (4 RCTs)	714 LGS and Dravet syndrome	Medicinal CBD (pure)	not reported		≥50% response: without clobazam 29.1% vs. 15.7% (RR 1.80); with clobazam 52.9% vs. 27.8% (RR 1.85). Clobazam was not randomized; the authors caution about this	PMID 32452532, doi:10.1111/epi.16546
Stockings 2018 Systematic review (6 RCTs + 30 observational studies)	not reported Drug-resistant epilepsy; mean age 16.1	Medicinal CBD (RCTs) + various products in observational studies	20 mg/kg/day (RCT)		≥50% seizure reduction RR 1.74 (NNT 8); GRADE: low Side effects: Side effects RR 1.24; serious side effects RR 2.55	PMID 29511052, doi:10.1136/jnnp-2017-317168
Lattanzi 2020 (Dravet meta-analysis) Systematic review and meta-analysis (3 RCTs, Dravet only)	359 (CBD 228, placebo 131; numbers from the abstract, see source note) Dravet syndrome	Medicinal CBD (plant-derived purified oral solution)	not reported (in the abstract)		A reduction of at least half in convulsive seizures was more common with CBD than with placebo: RR 1.69 (95% CI 1.21–2.36; p=0.002). No difference in the risk of any side effect (RR 1.06; 95% CI 0.87–1.28); side effects significantly linked to CBD were sleepiness, decreased appetite, diarrhea and raised liver enzymes. A pooling of the same three RCTs; no new patient data. Side effects: Stopping treatment 9.0% with CBD, 2.3% with placebo (RR 3.12; 95% CI 1.07–9.10)	PMID 32040850, doi:10.1007/s40263-020-00708-6
Lattanzi 2023 (network meta-analysis) Systematic review and network meta-analysis (8 placebo-controlled RCTs in Dravet: stiripentol 2, CBD 3, fenfluramine 2, soticlestat 1)	680 (according to the article's text: CBD 228, fenfluramine 122, stiripentol 33, soticlestat 26, placebo 271); when the arms in the article's patient table are added up, CBD 221, total 673 Dravet syndrome	Pharmaceutical-grade CBD; comparators stiripentol, fenfluramine and soticlestat	CBD 10 and 20 mg/kg/day (5–20 in the dose-ranging study)	Search up to the last week of January 2023; studies 8–20 weeks	All the medicines were better than placebo for a reduction of at least half in convulsive seizures. In the indirect comparison, this rate was lower with CBD than with fenfluramine (OR 0.20; 95% CI 0.07–0.54) and higher with stiripentol than with CBD (OR 14.07; 95% CI 2.57–76.87; the stiripentol estimate rests on two small studies, and the confidence interval is very wide). No significant difference between the medicines for seizure freedom; the only medicine that significantly increased seizure freedom vs. placebo was stiripentol. There are no studies comparing the medicines directly. Of the 10 authors, 4 report receiving payments or support from the CBD manufacturer (GW), 1 from the maker of fenfluramine, Zogenix, and 5 from UCB, Zogenix's current owner (the people overlap; most are stated as unrelated to the study); the university paid the open-access fee. Side effects: With CBD, stopping treatment for any reason (OR 3.49; 95% CI 1.11–10.95) and stopping because of side effects (OR 5.18; 1.15–23.23) more common than with placebo; stopping treatment for any reason less common with stiripentol than with CBD (OR 0.08; 0.01–0.61 in the Results; the abstract says 0.45; 0.04–5.69). Any side effect less common with CBD than with fenfluramine (OR 0.22; 0.06–0.78) and stiripentol (stiripentol vs. CBD OR 75.72; 3.59–1598.58); no difference between the medicines in serious side effects	PMID 37695433, doi:10.1007/s40265-023-01936-y
Guerrini 2024 (network meta-analysis, approved doses) Systematic review and network meta-analysis (two placebo-controlled RCTs each for stiripentol, fenfluramine and CBD; approved doses and, for CBD, clobazam subgroups)	6 RCTs; the arms in the article's data table add up to 559 patients (64 in the stiripentol studies, 177 fenfluramine, 318 CBD) Dravet syndrome (as the first add-on medicine)	Pharmaceutical-grade CBD (GWPCARE1, GWPCARE2); comparators stiripentol and fenfluramine	CBD 10 and 20 mg/kg/day (all patients and those taking clobazam separately)	Studies 8–14 weeks	For a reduction of at least half in convulsive seizures, the absolute difference vs. placebo was 64% for stiripentol, 62% for fenfluramine 0.7 mg/kg/day, 19% for CBD 20 mg/kg/day (25% in those taking clobazam) and 16% for CBD 10 mg/kg/day (18% in those taking clobazam, not significant); both stiripentol and fenfluramine were significantly superior to the approved doses of CBD on this measure and for a 75% reduction. For seizure freedom, stiripentol was superior to fenfluramine and CBD. The stiripentol studies were done 15–20 years ago, were shorter, and one was stopped early because a clear benefit was seen at an interim analysis; 33–49% of patients in the CBD studies were already taking stiripentol; the comparisons are indirect. The maker of stiripentol (Biocodex) funded it; 1 of the 5 authors is a Biocodex employee; 2 are executives of consulting firms working for Biocodex and former employees of the maker of fenfluramine (Zogenix/UCB); 2 sit on advisory boards of companies including Biocodex and the CBD manufacturer and were investigators in the stiripentol studies. Side effects: No significant difference between the medicines in serious side effects; an increase in serious side effects vs. placebo only with CBD 20 mg/kg/day (RR 1.90; 95% CI 1.03–3.50). Stopping because of side effects lowest with stiripentol (the stiripentol studies were shorter)	PMID 38427284, doi:10.1002/epi4.12923
Devinsky 2020 (clobazam subgroups) Post hoc analysis in which the manufacturer split the 4 RCTs (GWPCARE1–4) by clobazam use and combined them (stratified meta-analysis)	714 (318 Dravet, 396 LGS); CBD 429 (240 with clobazam, 189 without clobazam), placebo 285 Dravet (aged 2–18) and LGS; in the Dravet studies, 64% of patients were taking clobazam and about 38% stiripentol	Epidiolex/Epidyolex (100 mg/mL oral solution, purified CBD)	10 and 20 mg/kg/day	14 weeks	With both doses and both conditions combined, the ratio of seizure frequency vs. placebo was 0.59 (95% CI 0.52–0.68) with clobazam and 0.85 (0.73–0.98; p=0.023) without clobazam: on this measure, the effect is smaller without clobazam and the dose groups are not significant individually. The OR for a reduction of at least half, however, is similar in the two groups: 2.51 (1.69–3.71) with clobazam, 2.40 (1.38–4.16) without clobazam. In 89 patients taking both stiripentol and clobazam (all from the Dravet studies; 58 CBD, 31 placebo), CBD's effect vs. placebo was numerically favorable but not significant (ratio 0.81; 0.57–1.17; OR 1.51; 0.36–6.41). Clobazam use was not randomized; about two-thirds of those not taking clobazam had tried and stopped clobazam before. All the authors (7/7) are employees (5) or consultants and investigators (2) of the manufacturer; the manufacturer funded it. Side effects: Sleepiness, rash, pneumonia and aggression more common together with clobazam; stopping because of side effects at 20 mg/kg/day 13.0% with clobazam, 8.0% without clobazam	PMID 32592183, doi:10.1111/ane.13305, Gunning 2021, PMID 32969022, in Dravet, ratio of convulsive seizures vs. placebo 0.71 (95% CI 0.60–0.83) in all patients, 0.63 (0.52–0.77) with clobazam, OR for a reduction of at least half 2.34 (1.51–3.63) and 2.78 (1.62–4.77), 3 of the 7 authors are employees of the manufacturer, 4 are consultants or investigators for the manufacturer
Scheffer 2021 (GWPCARE5, Dravet) Open-label extension study (those who completed GWPCARE1 and GWPCARE2); no control group	315 (95% of the 330 patients who completed the RCTs) Dravet syndrome (children and young people from the RCTs)	Epidiolex/Epidyolex (100 mg/mL oral solution, purified CBD)	Target 20 mg/kg/day; could be reduced or increased to at most 30 mg/kg/day based on response and tolerability; mean most frequently used dose 22 mg/kg/day From 2.5 to 20 mg/kg/day over 2 weeks	Median 444 days (18–1535 days); efficacy up to 156 weeks	In patients from GWPCARE1 and GWPCARE2, median reduction in convulsive seizures 45–74% and in total seizures 49–84% in 12-week periods up to 156 weeks; in each period, at least 83% of patients or caregivers reported improvement in overall condition. According to the final results on ClinicalTrials.gov (not a peer-reviewed publication), 170 of 315 patients completed the study (of those who left, 62 at the patient's or family's request, 26 because of side effects, 23 by investigator decision, 20 for lack of effect); median reduction in convulsive seizures in the last 12 weeks 40% (IQR −77.8 to +6.7). No control group, and only those who completed the first study and continued were included. Run by the manufacturer (GW Research); 2 of the 11 authors are employees of the manufacturer, and the conflict-of-interest statements of the others could not be read (in the interim analysis of the same study, all 7 authors had ties to the manufacturer). Side effects: Side effects 97% (mild 23%, moderate 50%, severe 25%); most common diarrhea 43%, fever 39%, decreased appetite 31%, sleepiness 28%. Stopping because of side effects 28 patients (9%). Liver enzymes (transaminases) >3×ULN in 69 patients (22%); 84% of them were taking valproate. In the final results in the registry, serious side effects in 133 patients (42%), 6 deaths (relationship to the drug not stated in the registry)	PMID 34406656, doi:10.1111/epi.17036, NCT02224573, separate results for Dravet in the registry, posted in February 2022). Interim analysis: Devinsky 2019, PMID 30582156, median 274 days, transaminases ≥3×ULN in 22 of 128 patients (17.2%) from GWPCARE1, all taking valproate, stopping because of side effects 6.4%, median reduction in convulsive seizures 38–44% up to week 48
Laux 2019 Open-label expanded access program (EAP), LGS/Dravet subgroup	152 (58 Dravet, 94 LGS) LGS and Dravet syndrome	Epidiolex 100 mg/ml (pure CBD)	Start at 2–10 mg/kg/day → ceiling of 25–50 mg/kg/day; median 21–25 mg/kg/day in weeks 12–96 Varied by center	Median 78.3 weeks	At week 12, major motor seizures reduced by 50% and all seizures by 44%; consistent up to 96 weeks (no control group; results given for LGS and Dravet together, no separate result for Dravet in the abstract) Side effects: Sleepiness 30%, diarrhea 24%; 28% stopped (mostly for lack of effect, 20%)	PMID 31022635, doi:10.1016/j.eplepsyres.2019.03.015
Ben-Menachem 2020 (stiripentol and valproate interaction, in patients) Phase 2, double-blind, placebo-controlled, randomized (4:1), two-arm drug interaction study; not designed to measure efficacy	35 randomized: stiripentol arm 14 (12 CBD, 2 placebo), valproate arm 21 (1 left before taking the drug; 16 CBD, 4 placebo); 15 in the drug-level analysis in the valproate arm Patients with epilepsy aged 16–55 taking a stable dose of stiripentol or valproate (Europe; mean age 29.5); syndrome not specified	Epidyolex/Epidiolex (100 mg/mL oral solution, purified CBD) or placebo	CBD 20 mg/kg/day To 20 mg/kg/day over 10 days	26 days (10 days of titration, 20 mg/kg/day on days 12–26)	CBD raised stiripentol levels somewhat: peak level 1.17-fold (90% CI 1.03–1.33), total exposure 1.30-fold (1.09–1.55); the authors recommend monitoring side effects in people using them together. There was no important change in the levels of valproate and its breakdown product 4-ene-VPA (valproate 0.87- and 0.83-fold; 4-ene-VPA 0.77- and 0.70-fold). Few patients, very small placebo groups (2 and 4 people), and efficacy was not measured. The manufacturer's study; 6 of the 10 authors are employees of the manufacturer, 4 are investigators for the manufacturer (3 of whom also received consulting fees). Side effects: In the stiripentol arm, side effects in 8 of 14 patients (all in the CBD group), most common diarrhea and tiredness; 1 patient stopped because of a serious rash, and liver enzymes rose in 2 patients. In the valproate arm, side effects 88% on CBD, 25% on placebo; 1 patient stopped because of seriously raised liver enzymes (ALT 5.5×ULN), 1 patient left because of diarrhea and nausea. Diarrhea in 42–69% of those taking CBD	PMID 32350749, doi:10.1007/s40263-020-00726-4, NCT02607891, NCT02607904
VanLandingham 2020 (clobazam interaction, in patients) Phase 2, double-blind, placebo-controlled, randomized (4:1) drug interaction study; not designed to measure efficacy	20 randomized (16 CBD, 4 placebo); 13 in the drug-level analysis (10 CBD, 3 placebo) Patients with epilepsy aged 18–65 whose seizures were not controlled and who were taking a stable dose of clobazam (at most 20 mg a day); syndrome not specified	Epidiolex (100 mg/mL oral solution, purified CBD) or placebo, added to clobazam	CBD 20 mg/kg/day, 2 equal doses; clobazam 5–20 mg a day (not changed) To 20 mg/kg/day over 10 days	33 days (10 days of titration, 21 days at a stable dose)	CBD did not change the level of clobazam itself (total exposure ratio 1.1; 90% CI 0.9–1.2); the peak level of clobazam's active breakdown product N-desmethylclobazam rose 2.2-fold (90% CI 1.4–3.5) and total exposure 2.6-fold (2.0–3.6); it did not change in those taking placebo (1.2 and 1.0). Few patients, and efficacy was not measured. The manufacturer's study; all 4 authors are employees of the manufacturer (one during the study). Side effects: Side effects in 13 of 16 patients on CBD and 2 of 4 on placebo; most common diarrhea (6/16), nausea and vomiting (3/16), sleepiness, lightheadedness and skin inflammation; 1 patient was withdrawn because of a serious side effect (seizure cluster), 1 patient left because of side effects; no deaths	PMID 32652616, doi:10.1002/jcph.1634, NCT02565108
Morrison 2019 (clobazam, stiripentol and valproate interaction) Phase 1, open-label, fixed-sequence drug interaction study (in the same person, first the single drug, then with CBD)	78 enrolled; 77 people took the drug, 70 completed (12 people in the group adding CBD to clobazam; the article's results section says 13) Healthy adult volunteers (aged 18–55), Netherlands	Epidiolex (100 mg/mL oral solution) + clobazam, stiripentol or valproate	CBD 750 mg twice daily (equivalent to the highest dose in the phase 3 studies); clobazam 5 mg, stiripentol 750 mg, valproate 500 mg twice daily After cases of rash, a 3–10-day titration was added to the protocol	Co-administration for 7–21 days depending on the group	CBD raised stiripentol exposure 1.3–1.6-fold; it raised the peak level and total exposure of clobazam's active breakdown product N-desmethylclobazam about 3.4-fold (clobazam itself 1.2-fold); valproate levels did not change. Clobazam raised CBD's active breakdown product 7-OH-CBD 1.5–1.7-fold; stiripentol, in contrast, lowered 7-OH-CBD by 29%. No placebo group. The manufacturer's study; 2 of the 4 authors are employees of the manufacturer and 2 are consultants to the manufacturer. Side effects: 6 people were withdrawn because of side effects (3 when clobazam was added, 3 when valproate was added); the study was temporarily halted because of cases of rash, and no link to the drug could be established; no serious side effects and no deaths	PMID 30791225, doi:10.1002/cpdd.665
D'Onofrio 2020 (France, slow titration) Prospective, open-label, multicenter study (7 rare epilepsy centers in France; under a temporary use authorization); no control group	125 (48 Dravet, 62 LGS, 5 TSC, 10 other) Children aged 2–18 with drug-resistant epilepsy	Purified CBD oral solution (Epidyolex; under a temporary use authorization)	Target 10 mg/kg/day over at least 1 month, then at most 20 mg/kg/day; median 10, 14 and 15.5 mg/kg/day at months 1, 2 and 6 (higher in Dravet than in LGS) Started at 2.5 mg/kg/day and increased by 2.5 mg/kg/day per week (10 mg/kg/day in 4 weeks)	6 months	At month 6, total seizure frequency fell by a mean of 41% (SD 37.5) in the whole group. In patients with Dravet (those with month-6 data; the article says 22 patients, but the percentages it reports appear to be calculated on 23 patients), seizures fell by a mean of 43.9% (SD 37.2); by at least half in 34.8% (8 patients) and by at least 90% in 8.7% (2 patients). One Dravet patient was seizure-free from month 1 on, and another at month 6. Efficacy and side effects did not differ between Dravet and LGS or between those taking and not taking clobazam. No control group; the conclusion that slow titration was better tolerated rests on an indirect comparison with earlier studies. The manufacturer is not in the funding section; 2 of the 16 authors report receiving meeting fees from the manufacturer. Two of the centers (Paris Necker, Lyon) are also in Nabbout 2023; patients may partly overlap. Side effects: Side effects 48.8% (sleepiness 26, fatigue 20, behavior problems 16 patients); no difference between Dravet and LGS; transaminase elevation ≥3-fold in 11 patients, all of whom were taking valproate and clobazam together; 26 patients (20.8%) stopped (mostly for lack of effect); sudden unexpected death in epilepsy in a 2-year-old child with Dravet 1 month after starting CBD	PMID 32903409, doi:10.3389/fneur.2020.00829
Nabbout 2023 (use without clobazam) Retrospective chart review (the manufacturer's study GWEP20052), European early access program; no control group	114 enrolled; 107 analyzed (15 Dravet, 92 LGS) Dravet and LGS, aged ≥2; those who used CBD for at least 3 months without taking clobazam (mean age 10.5 in Dravet); based on the authors' institutions, the centers are in France, England, Italy and Spain	Epidyolex (100 mg/mL oral solution)	Time-weighted mean dose 11.56 mg/kg/day in Dravet, 13.54 mg/kg/day in LGS	Up to 12 months	In Dravet, median change in 28-day seizure frequency in 3-month periods between 0% and −16.7%; reduction of at least half in convulsive seizures 21% (14 patients) at month 3, 13% (8 patients) at month 12. In LGS, this rate for drop seizures was 19% at month 3, 30% at month 12. Rates of staying on CBD without clobazam (LGS and Dravet together) at months 3, 6, 9 and 12 were 94%, 80%, 69% and 63%. The Dravet group is very small; no control group. Manufacturer-supported; 2 of the 12 authors are employees of the manufacturer (Jazz), and 7 report receiving consulting fees, speaker fees or travel support from the manufacturer. Some of the French centers are also in D'Onofrio 2020; patients may partly overlap. Side effects: Side effects 31% (most common sleepiness, seizures, diarrhea, decreased appetite); 2 patients stopped because of side effects; raised liver enzymes in 4 patients (all LGS)	PMID 37331197, doi:10.1016/j.seizure.2023.05.003
Cerulli Irelli 2025 (real world, SCN1A/Dravet subgroup) Retrospective, multicenter real-world study (27 centers, mostly in Europe, including Argentina); no control group	266 patients with monogenic epilepsy; 86 with a change in the SCN1A gene (a Dravet diagnosis also in 86 patients) Monogenic drug-resistant epilepsy; median age 12 when starting CBD	Epidiolex (Europe) or Convupidiol (Argentina); >99% purified CBD	Not given separately for Dravet	Median 17 months of follow-up (IQR 12–24); at least 3 months of CBD	At last follow-up, compared with baseline: in 86 patients with SCN1A, mean reduction in seizures 37.6% (SD 33), ≥50% response 50.6%, improvement on the doctor's assessment (CGI-I) 63.9%. There was no difference between diagnostic groups (Lennox–Gastaut, Dravet, TSC, other encephalopathies) in mean reduction, response rate, improvement or staying on the medicine (p > 0.4). Because only those who took CBD for at least 3 months were included, those who stopped early were left out; retrospective and uncontrolled. No funding statement was found in the text; 7 of the 40 authors report receiving payments or support unrelated to the study from the CBD manufacturer (Jazz). Three of the Italian centers (Rome Gemelli, Ancona Salesi, Bologna) are also in Perulli 2025 and Lelli 2026; patients may partly overlap. Side effects: Not given separately for Dravet; in the whole group, side effects during follow-up 38.5% (21.8% at last follow-up; most common sleepiness 12.9%)	PMID 40126049, doi:10.1111/epi.18378
Perulli 2025 (real world, Italy) Retrospective, two-center observational study (Rome Gemelli and Ancona Salesi); no control group	107 (18 Dravet, 59 LGS, 9 TSC, 21 other encephalopathies) Developmental and epileptic encephalopathies; median age 12.8 when starting CBD	Epidyolex (purified CBD)	At last follow-up, median 16.4 mg/kg/day (IQR 11.6–19.5); at most 25 mg/kg/day Started at 2.5 mg/kg/day and increased by 2.5 mg/kg/day per week	Median 20 months (IQR 11–27); at least 3 months of CBD	At last follow-up, seizures fell by at least half in 49 of 107 patients (46%) (the article gives this rate as 69% based on the 71 patients who continued CBD). Response in Dravet was lower than in the other groups: reduction of at least half in 7 of 18 patients (39%; 72% in the others, p=0.007) at month 3 and in 4 (22%; 51% in the others, p=0.028) at last follow-up; staying on the medicine 8/18 (44%) in Dravet, 71% in the others. Use together with valproate was linked to lower efficacy (p=0.006); clobazam had no effect. The Dravet group is small, and the study is uncontrolled and retrospective; a 27-center study published the same year (Cerulli Irelli 2025) found no difference between syndromes. No company support was received for the study; 4 of the 9 authors report receiving consulting or meeting fees unrelated to the study from the manufacturer (Jazz). Both centers (Rome Gemelli, Ancona Salesi) are also in Lelli 2026 and Cerulli Irelli 2025; patients may partly overlap. Side effects: Side effects 33.6%, mostly mild and temporary; 9% stopped because of side effects	PMID 41165013, doi:10.1002/epi4.70149
Silvennoinen 2021 (adult Dravet, England) Retrospective, two-center follow-up study (England; the manufacturer's early access program); no control group	18 (17 for efficacy) Adults with genetically confirmed Dravet syndrome (median age 27.5; 20–51)	Purified CBD (the manufacturer's early access program)	Up to 20 mg/kg/day as tolerated Increased up to 20 mg/kg/day (details not in the abstract)	Median 176 days	Convulsive seizures fell by more than 30% in 3 of 17 patients (17.6%) (in these three patients, the reduction was 87.5–100%); CBD was stopped for lack of effect in 7 patients, and at last follow-up 8 of the 18 patients were still on treatment. Improvement in overall condition 41.2% according to doctors, 47.1% according to families. Only 17.6% reached the seizure reduction required for the medicine to be reimbursed in England. Small, uncontrolled, two centers; conflict-of-interest statement and funding are not in the PubMed record. Side effects: Side effects in everyone; most common raised liver enzymes (52.9%). 3 patients (16.7%) stopped because of behavioral side effects, one of whom was seizure-free. Doses of other medicines had to be closely monitored and adjusted	PMID 34052628, doi:10.1016/j.seizure.2021.05.014
Lelli 2026 (status epilepticus, Italy) Retrospective, multicenter cohort (Italy); the link between exposure to new medicines and status epilepticus, with models adjusted for age and follow-up time	61 patients, 1033 episodes of status epilepticus Genetically confirmed Dravet syndrome (median age 13)	CBD (among 'newly approved medicines' in the abstract; no product name), examined together with stiripentol and fenfluramine	not reported (in the abstract)	not reported (in the abstract)	Fenfluramine use was linked to a decrease in the frequency of status epilepticus (IRR 0.19; p<0.001); the decrease with stiripentol was not significant (IRR 0.75; p=0.286). Stiripentol was linked to shorter episode duration (−5.0 min; p=0.003); CBD's link with episode duration was not significant (−2.7 min; p=0.506); CBD's link with episode frequency is not given separately in the abstract. Each additional new medicine was linked to an additional decrease in episode frequency and duration. Observational and retrospective, not randomized. The authors report no conflicts of interest. The centers (Ancona, Rome, Bologna) partly overlap with Perulli 2025 and Cerulli Irelli 2025. Side effects: not reported (in the abstract)	PMID 42636610, doi:10.1016/j.seizure.2026.08.011
Press 2015 (unapproved cannabis extracts) Retrospective chart review (single center, Colorado); based on parent reports; no control group	75 (number of patients with Dravet not given in the abstract) Children and adolescents with drug-resistant epilepsy (Dravet, Doose, LGS and others)	Oral cannabis extracts (unapproved, CBD-rich; content and THC ratio not verified)	not reported (in the abstract)	not reported (in the abstract)	In the whole group, 33% of families reported a reduction in seizures of more than half; in Dravet, this rate was 23% (LGS 88.9%, Doose 0%). Response was 47% in families who moved to Colorado for this treatment and 22% in those already living there. No improvement was seen on the EEG of responders. No control group, product content unknown; the authors point to confounding factors and call for controlled studies. Side effects: Side effects 44%: increased seizures 13%, sleepiness/tiredness 12%; developmental regression, abnormal movements, status epilepticus requiring intubation and death were reported rarely	PMID 25845492, doi:10.1016/j.yebeh.2015.02.043, Overlapping cohort from the same center: Treat 2017, PMID 27859038, a Dravet diagnosis was linked to shorter use (p=0.02), per parent report, response 24%, side effects 19%
McCoy 2018 (THC-containing cannabis oil) Prospective, open-label, single-center study (Toronto); no control group and no blinding	20 (19 completed; 1 child died of sudden unexpected death in epilepsy during the study and was excluded from the analysis) Children with drug-resistant Dravet syndrome (aged 2.1–17.8; mean 10.2); 70% were taking clobazam, 60% valproate, 50% stiripentol	TIL-TC150 (Tilray): 100 mg CBD and 2 mg THC per mL (50:1), in grapeseed oil; Health Canada-approved investigational product	Target 16 mg/kg/day CBD (0.32 mg/kg/day THC); mean reached 13.3 mg/kg/day CBD (7–16) and 0.27 mg/kg/day THC (0.14–0.32); mean 9.3 mg THC per day (2.4–16.8) Titration over 8 weeks; 8 of 19 patients reached the target dose (the others did not because of sleepiness, loss of appetite, diarrhea or weight loss)	20 weeks (including 8 weeks of dose titration)	Comparing baseline with weeks 17–20, median reduction in motor seizures 70.6% (p=0.006); ≥50% reduction 63%, 2 children seizure-free, seizures increased in 4 children. Seizure-free days and quality-of-life scores increased, and spike activity on EEG decreased. After week 20, 14 of the 19 children continued follow-up; the 5 families who stopped reported seeing no benefit. No control group and no blinding, 20 children; THC's contribution cannot be separated from CBD's. The company that makes the product (Tilray) funded most of the study; the researchers report receiving no payment from the company. Side effects: Side effects in 18 of 19 patients (95%), mostly during titration and temporary; most common sleepiness (17 patients), loss of appetite, diarrhea. Temporary raised liver enzymes in 8 patients (5 of whom also had high valproate levels), mild and temporary low platelets in 8 patients; resolved when the valproate or oil dose was adjusted. Weight loss of more than 5% in 5 patients (26%). No one stopped because of side effects; the dose was reduced in 6 of the 8 patients who reached the target dose. 1 child died of sudden unexpected death in epilepsy (SUDEP); the authors judged it unrelated to the product	PMID 30250864, doi:10.1002/acn3.621, NCT02983695, no results in the registry)

Reported side effects

Sleepiness/sedation: Epidiolex (LGS/Dravet) 32% (10 mg/kg/day 27%, 20 mg/kg/day 34%), placebo 11%
Diarrhea, decreased appetite, vomiting, fever, tiredness (common in RCTs)
Liver enzyme (ALT) >3×ULN: in controlled studies, LGS/Dravet 13%, TSC 12% (placebo 1%); 17% at 20 mg/kg/day, 1% at 10 mg/kg/day; >20×ULN <1%
ALT >3×ULN with valproate: valproate+clobazam 30%, valproate alone 21%, clobazam alone 4%, neither 3% (FDA)
Sleepiness/sedation with clobazam 46% (without clobazam 16%)
Stopping due to side effects 8.9% vs. 1.8% (Lattanzi 2018 meta-analysis)
Labeling warnings: suicidal thoughts and behavior (class warning for all antiseizure medicines), hypersensitivity (including to sesame oil), increased seizures on sudden stopping; EMA: seizure frequency may increase, weight loss/failure to gain weight
Postmarketing: cholestatic/mixed-type liver injury and (mostly in people taking valproate/clobazam) elevated ammonia were reported (FDA)
Long term (Dravet extension study GWPCARE5, median 444 days): side effects 97%, stopping due to side effects 9%, transaminases >3×ULN 22% (84% of whom were taking valproate); in the final results in the registry, serious side effects 42% (Scheffer 2021; NCT02224573)
Weight loss: in the LGS/Dravet controlled studies, a decrease of at least 5% in measured weight from baseline occurred in 16% of people taking CBD and 8% on placebo; the decrease appeared dose-related (18% at 20 mg/kg/day, 9% at 10 mg/kg/day). Weight loss reported as a side effect: 3% at 10 mg/kg/day, 5% at 20 mg/kg/day, 1% on placebo. In the TSC study (25 mg/kg/day), weight loss of at least 5% was 31% vs. 8% (FDA, product information 05/2026)

Monitoring done in studies

Before starting: ALT, AST and total bilirubin (FDA and EMA)
Liver tests at months 1, 3 and 6, then periodically or as clinically needed
FDA: within 1 month after a dose change or after adding a medicine that affects the liver; EMA: the monitoring schedule starts over when going above 10 mg/kg/day or when a medicine that affects the liver is changed
People using valproate and those with high baseline ALT/AST: per the EMA, at week 2 and months 1, 2, 3 and 6, then periodically
Stopping rule (FDA and EMA): CBD is stopped if transaminases are >3×ULN and bilirubin >2×ULN, or if transaminases are persistently >5×ULN
If liver symptoms occur (unexplained nausea, vomiting, right upper abdominal pain, tiredness, loss of appetite, jaundice, dark urine), check ALT/AST/bilirubin right away
Blood levels and side effects of co-administered antiseizure medicines (Gaston 2017, PMID 28782097)

Who should be careful

Epilepsy treatment is started, changed and stopped by a neurologist's decision; antiseizure medicines are not stopped on one's own
CBD is stopped by tapering the dose gradually; stopping suddenly carries a risk of increased seizures and status epilepticus (FDA §2.5, §5.5). EMA: in studies, stopping was done over ~10 days, reducing by ~10% per day
In moderate (Child-Pugh B) and severe (Child-Pugh C) hepatic impairment, the labeling specifies lower starting and maintenance doses; no adjustment is needed in mild impairment
Hypersensitivity to sesame oil is a contraindication (Epidyolex contains 736 mg of refined sesame oil and 79 mg of ethanol per ml)
People using valproate or clobazam need more frequent monitoring for liver effects and sedation
Age: FDA ≥1 year, EMA ≥2 years
The medicine is always taken the same way in relation to food; with a high-fat meal, Cmax rises 5-fold and AUC 4-fold. Measure with a calibrated syringe; a household spoon is not suitable
Unregulated (artisanal) CBD/cannabis oils are not a substitute for the approved medicine: the label is unreliable, they may contain THC and contaminants, and there are reports of increased seizures
In Turkey, fenfluramine (Fintepla, 2025) is approved for Dravet; medicinal CBD, however, is not approved (TİTCK, the Turkish Medicines and Medical Devices Agency, October 16, 2026)

Drug interactions

Clobazam: CBD raises the active metabolite N-desmethylclobazam ~3-fold (FDA) / 3–4-fold (EMA) (total exposure 2.6-fold in a placebo-controlled study in patients with epilepsy taking clobazam, VanLandingham 2020; 3.4-fold in healthy volunteers, Morrison 2019; this increase was not seen in 4 children with Dravet also taking stiripentol, Devinsky 2018); clobazam also raises 7-OH-CBD (AUC 47%). If sleepiness occurs, reducing the clobazam dose is considered
Valproate: raised liver enzymes increase markedly; diarrhea and decreased appetite also increase (EMA). If enzymes rise, CBD and/or valproate is reduced or stopped
Stiripentol: in healthy volunteers, Cmax 28%, AUC 55% (Morrison 2019); in a placebo-controlled study in patients with epilepsy, Cmax up 17%, AUC up 30% (Ben-Menachem 2020). Side effects are closely monitored
Everolimus: exposure rises ~2.5-fold; levels are monitored, and in patients using CBD, everolimus is started at a lower dose
Topiramate, rufinamide, zonisamide, eslicarbazepine: in an open-label study, serum levels rose as the CBD dose increased (most stayed within the therapeutic range)
Sirolimus, tacrolimus, digoxin (oral P-gp substrates): exposure may rise; level monitoring and dose reduction are considered
Strong CYP3A4/CYP2C19 inducers (rifampin; EMA: carbamazepine, enzalutamide, mitotane, St. John's wort): rifampin lowered CBD by ~32% and 7-OH-CBD by ~63% (FDA); FDA (05/2026): with a strong inducer, increasing the Epidiolex dose up to at most 2-fold may be considered based on response and tolerability
CYP1A2 substrates (theophylline, tizanidine, caffeine): caffeine AUC rose 95%; dose reduction is considered for narrow-range medicines
CYP2C19 prodrugs such as clopidogrel: the active metabolite may decrease
Alcohol and other CNS depressants: sleepiness and sedation increase; alcohol also increases CBD exposure (Cmax 93%, AUC 63%)

Abbreviations: RCT, randomized controlled trial (a study in which participants are assigned to groups by chance); placebo, an inactive substance; SMD, OR and RR, effect measures used in analyses that pool studies; CYP, liver enzymes that break down medicines.

If you are considering CBD

You can talk with your neurologist about whether the approved CBD medicine is right for your child.
Review all the medicines your child takes together for interactions, especially clobazam, valproate and stiripentol.
Plan the schedule of liver tests from the start; if unexplained nausea, abdominal pain, jaundice or dark urine occurs, tell the doctor right away.
Do not stop epilepsy medicines or CBD suddenly; make every change with the doctor.

Choose your age group, medicines and situation to prepare a short summary you can take to your doctor. Nothing you enter is sent anywhere.

Open the personal warnings tool Convert drops ↔ mg

Bottom line

In Dravet syndrome, the approved CBD medicine is an add-on treatment option supported by good studies. It does not help everyone to the same degree and requires regular monitoring; in indirect comparisons with limited certainty against the other medicines approved for Dravet, its effect looked smaller. There is no evidence that the CBD oils on the market will give the same result.

Frequently asked questions

Does CBD oil work for Dravet syndrome?

When added to existing medicines, the approved, purified CBD medicine reduced seizures more than an inactive drug (placebo); the evidence on this is strong. This result does not apply to CBD oils whose content and dose are unknown. In indirect comparisons with limited certainty against other medicines approved for Dravet, such as fenfluramine and stiripentol, its effect looked smaller; which medicine to try is decided together with the neurologist.

Can CBD oil be used in children?

The approved CBD medicine was studied in children with Dravet syndrome; it is approved from age 1 in the US and from age 2 in the European Union. It is used by decision of a pediatric neurologist and with regular liver tests.

What is the difference between CBD oil on the market and the approved medicine?

The approved medicine contains purified CBD, the same amount is in every bottle, and its effect has been shown in studies. In oils on the market, the amount on the label may not be reliable, and THC or contaminants may be present; that is why they are not a substitute for the medicine.

Can CBD be used together with other epilepsy medicines?

In the studies, CBD was already given on top of existing medicines. However, sleepiness is more common with clobazam, and raised liver enzymes are more common with valproate; CBD also raises the blood level of stiripentol somewhat (about 30% in a study in patients); that is why medicine doses and blood tests are monitored by the doctor.

Can CBD oil for Dravet syndrome be obtained with a prescription in Turkey?

In Turkey, products derived from cannabis are regulated by the Kenevirden Elde Edilen Ürünlere Dair Yönetmelik (Regulation on Products Derived from Hemp) dated January 31, 2026. These products, including CBD, can be sold only in pharmacies. Those in the medicine and health product category are dispensed with a prescription; the regulation does not state a prescription requirement for supplement products and cosmetics with a THC content of at most 0.3%. As of October 2026, there is no approved CBD medicine in Turkey; the active ingredient of the approved CBD medicine is on the Foreign Active Ingredient List of TİTCK (the Turkish Medicines and Medical Devices Agency), and bringing it in from abroad requires the agency's written approval. If you are thinking of using it, talk to your doctor first.

References

Show 23 references

Devinsky 2017 (GWPCARE1). Double-blind, placebo-controlled RCT. PMID 28538134, doi:10.1056/NEJMoa1611618, NCT02091375, PMID 34265088, on day 12 at 20 mg/kg/day (nominal p = 0.02) and on day 13 at 10 mg/kg/day (p = 0.03), most of the most common side effects (sleepiness, decreased appetite, diarrhea) resolved within 4 weeks. Devinsky 2024, PMID 38677101, funded by the manufacturer): in 124 placebo patients, seizure counts fell from baseline, but seizure-free days and the caregiver's global impression barely changed
Miller 2020 (GWPCARE2). Double-blind, placebo-controlled, dose-comparison RCT. PMID 32119035, doi:10.1001/jamaneurol.2020.0073, PMID 32391881
Devinsky 2018 (dose-ranging safety study). Multicenter, double-blind, placebo-controlled dose-ranging RCT (4:1). PMID 29540584, doi:10.1212/WNL.0000000000005254, NCT02091206
Lattanzi 2018. Systematic review and meta-analysis (4 RCTs). PMID 30390221, doi:10.1007/s40265-018-0992-5
Lattanzi 2020 (clobazam status). Systematic review and meta-analysis (4 RCTs). PMID 32452532, doi:10.1111/epi.16546
Stockings 2018. Systematic review (6 RCTs + 30 observational studies). PMID 29511052, doi:10.1136/jnnp-2017-317168
Lattanzi 2020 (Dravet meta-analysis). Systematic review and meta-analysis (3 RCTs, Dravet only). PMID 32040850, doi:10.1007/s40263-020-00708-6
Lattanzi 2023 (network meta-analysis). Systematic review and network meta-analysis (8 placebo-controlled RCTs in Dravet: stiripentol 2, CBD 3, fenfluramine 2, soticlestat 1). PMID 37695433, doi:10.1007/s40265-023-01936-y
Guerrini 2024 (network meta-analysis, approved doses). Systematic review and network meta-analysis (two placebo-controlled RCTs each for stiripentol, fenfluramine and CBD; approved doses and, for CBD, clobazam subgroups). PMID 38427284, doi:10.1002/epi4.12923
Devinsky 2020 (clobazam subgroups). Post hoc analysis in which the manufacturer split the 4 RCTs (GWPCARE1–4) by clobazam use and combined them (stratified meta-analysis). PMID 32592183, doi:10.1111/ane.13305, Gunning 2021, PMID 32969022, in Dravet, ratio of convulsive seizures vs. placebo 0.71 (95% CI 0.60–0.83) in all patients, 0.63 (0.52–0.77) with clobazam, OR for a reduction of at least half 2.34 (1.51–3.63) and 2.78 (1.62–4.77), 3 of the 7 authors are employees of the manufacturer, 4 are consultants or investigators for the manufacturer
Scheffer 2021 (GWPCARE5, Dravet). Open-label extension study (those who completed GWPCARE1 and GWPCARE2); no control group. PMID 34406656, doi:10.1111/epi.17036, NCT02224573, separate results for Dravet in the registry, posted in February 2022). Interim analysis: Devinsky 2019, PMID 30582156, median 274 days, transaminases ≥3×ULN in 22 of 128 patients (17.2%) from GWPCARE1, all taking valproate, stopping because of side effects 6.4%, median reduction in convulsive seizures 38–44% up to week 48
Laux 2019. Open-label expanded access program (EAP), LGS/Dravet subgroup. PMID 31022635, doi:10.1016/j.eplepsyres.2019.03.015
Ben-Menachem 2020 (stiripentol and valproate interaction, in patients). Phase 2, double-blind, placebo-controlled, randomized (4:1), two-arm drug interaction study; not designed to measure efficacy. PMID 32350749, doi:10.1007/s40263-020-00726-4, NCT02607891, NCT02607904
VanLandingham 2020 (clobazam interaction, in patients). Phase 2, double-blind, placebo-controlled, randomized (4:1) drug interaction study; not designed to measure efficacy. PMID 32652616, doi:10.1002/jcph.1634, NCT02565108
Morrison 2019 (clobazam, stiripentol and valproate interaction). Phase 1, open-label, fixed-sequence drug interaction study (in the same person, first the single drug, then with CBD). PMID 30791225, doi:10.1002/cpdd.665
D'Onofrio 2020 (France, slow titration). Prospective, open-label, multicenter study (7 rare epilepsy centers in France; under a temporary use authorization); no control group. PMID 32903409, doi:10.3389/fneur.2020.00829
Nabbout 2023 (use without clobazam). Retrospective chart review (the manufacturer's study GWEP20052), European early access program; no control group. PMID 37331197, doi:10.1016/j.seizure.2023.05.003
Cerulli Irelli 2025 (real world, SCN1A/Dravet subgroup). Retrospective, multicenter real-world study (27 centers, mostly in Europe, including Argentina); no control group. PMID 40126049, doi:10.1111/epi.18378
Perulli 2025 (real world, Italy). Retrospective, two-center observational study (Rome Gemelli and Ancona Salesi); no control group. PMID 41165013, doi:10.1002/epi4.70149
Silvennoinen 2021 (adult Dravet, England). Retrospective, two-center follow-up study (England; the manufacturer's early access program); no control group. PMID 34052628, doi:10.1016/j.seizure.2021.05.014
Lelli 2026 (status epilepticus, Italy). Retrospective, multicenter cohort (Italy); the link between exposure to new medicines and status epilepticus, with models adjusted for age and follow-up time. PMID 42636610, doi:10.1016/j.seizure.2026.08.011
Press 2015 (unapproved cannabis extracts). Retrospective chart review (single center, Colorado); based on parent reports; no control group. PMID 25845492, doi:10.1016/j.yebeh.2015.02.043, Overlapping cohort from the same center: Treat 2017, PMID 27859038, a Dravet diagnosis was linked to shorter use (p=0.02), per parent report, response 24%, side effects 19%
McCoy 2018 (THC-containing cannabis oil). Prospective, open-label, single-center study (Toronto); no control group and no blinding. PMID 30250864, doi:10.1002/acn3.621, NCT02983695, no results in the registry)

Points that could not be verified from the full text

The full texts of GWPCARE1, GWPCARE3 (NEJM) and GWPCARE4 (Lancet) could not be accessed; design, titration and results were verified with the ClinicalTrials.gov record and results, the PubMed abstract and the EMA product information. For Laux 2019, only the abstract was read
No official source was found on whether the route of importing for personal treatment with a doctor's report (TİTCK permission) has been used for Epidyolex in Turkey
Difference in efficacy: the expert opinion section of a fenfluramine review says CBD's efficacy appears 'lower' than fenfluramine's and 'similar' to stiripentol's (Schoonjans and Ceulemans 2021; PMID 33455486). This is not a direct comparison; the full text is paywalled, and only the abstract was read. In two network meta-analyses found in the October 2026 search (in the table; Lattanzi 2023, Guerrini 2024), the response rate with CBD was lower than with fenfluramine and stiripentol, so the 'similar to stiripentol' view does not agree with these analyses; but stiripentol rests on two small studies (64 children in total) done an estimated 15–20 years ago, the confidence intervals are very wide, and there is no study comparing the medicines directly
Of the studies added in the October 2026 search, only the PubMed abstract could be read for Scheffer 2021, Lattanzi 2020 (Dravet meta-analysis), Nabbout 2023, Silvennoinen 2021, Lelli 2026 and Press 2015. The centers of the three everyday-practice studies in Italy (Cerulli Irelli 2025, Perulli 2025, Lelli 2026) partly overlap; how many of the patients overlap is unknown. Some numbers do not add up within the articles themselves: D'Onofrio 2020 gives the Dravet subgroup as 22 patients, but the percentages it reports appear to be calculated on 23 patients; Lattanzi 2020 (Dravet meta-analysis) and Lattanzi 2023 give the number of patients taking CBD as 228, but the CBD arms of the same three studies add up to 221. The numbers were taken as given in the articles

Trusted sources from around the world

Show 5 references

Epidiolex drug label (DailyMed)US National Library of Medicine, English, May 2026
The official prescribing information for the approved CBD medicine: approved uses, side effects, liver monitoring and interactions.
Epidyolex (European Medicines Agency)European Medicines Agency (EMA), English
The European assessment report and product information for the approved CBD medicine.
Epilepsy Foundation: Medical cannabis and CBDEpilepsy Foundation (US), English, July 2025
The difference between the approved CBD medicine and unregulated CBD products; interactions with clobazam and valproate.
NICE TA614: CBD with clobazam for Dravet syndromeNICE (National Institute for Health and Care Excellence, UK), English, December 2019
The UK's health technology appraisal: approved CBD, together with clobazam, in patients aged 2 and over; convulsive seizures are reviewed every 6 months, and the medicine is stopped if there is not at least a 30% reduction.
International expert consensus on the diagnosis and treatment of Dravet syndrome (Wirrell et al., 2022)International panel of physicians and caregivers; supported by the Dravet Syndrome Foundation (Epilepsia journal), English, July 2022
A two-round consensus by an international panel of physicians and caregivers. Caregivers strongly supported using pharmaceutical-grade CBD as a first or second option, while physicians did not reach consensus on this; physicians did strongly agree on using it as a first, second or third option. Physicians strongly agreed that CBD products that are not pharmaceutical grade should not be recommended, while caregivers did not reach consensus (this is expert opinion).

These sources sell no products. All trusted sources

Research history

Here we record what changed on this page, when and why, with dates. The newest change is at the top.

October 28, 2026
On October 28, 2026, two sources found for this page during the searches for other topics were read from their primary sources. Weight loss numbers from the US product information of the approved CBD medicine were added to the side effects list: in the Dravet and Lennox–Gastaut controlled studies, the share of patients whose weight fell by at least 5% from baseline was 16% in the CBD group and 8% in the placebo group; this share was higher at the high dose. The 2021 position paper of the Brazilian Academy of Neurology was reviewed; it was not added to the page because it brought no new data and reported study results differently from their sources. The registry results of the extension study in the table were checked again against ClinicalTrials.gov and found correct. The conclusion and the strength of evidence did not change.
October 28, 2026
In October 2026, PubMed, Cochrane and ClinicalTrials.gov were systematically searched again (1295 PubMed and 46 ClinicalTrials.gov records reviewed; findings were reread from primary sources in an independent check). Studies added: a meta-analysis combining Dravet only (Lattanzi 2020), two network meta-analyses indirectly comparing Dravet medicines (Lattanzi 2023, Guerrini 2024), a manufacturer analysis by clobazam use (Devinsky 2020), the three-year extension of the two RCTs (Scheffer 2021), three drug-level studies measuring interactions with stiripentol, valproate and clobazam (Ben-Menachem 2020 and VanLandingham 2020 in patients, Morrison 2019 in healthy volunteers), everyday-practice studies from France, Italy and England (D'Onofrio 2020, Cerulli Irelli 2025, Perulli 2025, Lelli 2026; in adults, Silvennoinen 2021), use without clobazam (Nabbout 2023), unapproved cannabis extracts (Press 2015) and a THC-containing cannabis oil (McCoy 2018). Added to the evidence summary: that Cochrane's Dravet review does not include CBD, the clobazam and stiripentol subgroups, indirect comparisons with limited certainty against other Dravet medicines, long-term and everyday-practice data, and a sentence describing the search itself; the box and strength of evidence for THC-containing products were written for the first time (Insufficient). The two Lattanzi 2020 reviews in the table were distinguished by name; response rates, outcomes with no difference and company ties were added to the GWPCARE2 row. No new efficacy RCT was found. The conclusion and the strength of evidence did not change.
Related studies: Lattanzi 2020 (Dravet meta-analysis), published March 2020 (PMID 32040850, doi:10.1007/s40263-020-00708-6); Lattanzi 2023 (network meta-analysis), published October 2023 (PMID 37695433, doi:10.1007/s40265-023-01936-y); Guerrini 2024 (network meta-analysis, approved doses), published April 2024 (PMID 38427284, doi:10.1002/epi4.12923); Devinsky 2020 (clobazam subgroups), published December 2020 (PMID 32592183, doi:10.1111/ane.13305, Gunning 2021, PMID 32969022, in Dravet, ratio of convulsive seizures vs. placebo 0.71 (95% CI 0.60–0.83) in all patients, 0.63 (0.52–0.77) with clobazam, OR for a reduction of at least half 2.34 (1.51–3.63) and 2.78 (1.62–4.77), 3 of the 7 authors are employees of the manufacturer, 4 are consultants or investigators for the manufacturer); Scheffer 2021 (GWPCARE5, Dravet), published October 2021 (PMID 34406656, doi:10.1111/epi.17036, NCT02224573, separate results for Dravet in the registry, posted in February 2022). Interim analysis: Devinsky 2019, PMID 30582156, median 274 days, transaminases ≥3×ULN in 22 of 128 patients (17.2%) from GWPCARE1, all taking valproate, stopping because of side effects 6.4%, median reduction in convulsive seizures 38–44% up to week 48); Ben-Menachem 2020 (stiripentol and valproate interaction, in patients), published June 2020 (PMID 32350749, doi:10.1007/s40263-020-00726-4, NCT02607891, NCT02607904); VanLandingham 2020 (clobazam interaction, in patients), published October 2020 (PMID 32652616, doi:10.1002/jcph.1634, NCT02565108); Morrison 2019 (clobazam, stiripentol and valproate interaction), published November 2019 (PMID 30791225, doi:10.1002/cpdd.665); D'Onofrio 2020 (France, slow titration), published August 12, 2020 (PMID 32903409, doi:10.3389/fneur.2020.00829); Nabbout 2023 (use without clobazam), published August 2023 (PMID 37331197, doi:10.1016/j.seizure.2023.05.003); Cerulli Irelli 2025 (real world, SCN1A/Dravet subgroup), published July 2025 (PMID 40126049, doi:10.1111/epi.18378); Perulli 2025 (real world, Italy), published December 2025 (PMID 41165013, doi:10.1002/epi4.70149); Silvennoinen 2021 (adult Dravet, England), published October 2021 (PMID 34052628, doi:10.1016/j.seizure.2021.05.014); Lelli 2026 (status epilepticus, Italy), published November 2026 (online August 16, 2026) (PMID 42636610, doi:10.1016/j.seizure.2026.08.011); Press 2015 (unapproved cannabis extracts), published April 2015 (PMID 25845492, doi:10.1016/j.yebeh.2015.02.043, Overlapping cohort from the same center: Treat 2017, PMID 27859038, a Dravet diagnosis was linked to shorter use (p=0.02), per parent report, response 24%, side effects 19%); McCoy 2018 (THC-containing cannabis oil), published September 2018 (PMID 30250864, doi:10.1002/acn3.621, NCT02983695, no results in the registry))
October 27, 2026
The reason for the difference between one table review's (Stockings 2018) 'low certainty' rating and the guide's 'strong' rating was added to the evidence summary: the review was based on two studies up to 2017, and later meta-analyses point in the same direction. The conclusion and the strength of evidence did not change.

Older changes (5)

October 27, 2026
The Turkey line was corrected: the active ingredient of the approved CBD medicine (cannabidiol oral solution) is on the Foreign Active Ingredient List of TİTCK (the Turkish Medicines and Medical Devices Agency) and can be brought in from abroad with the agency's written approval; unsourced information about the SGK (Social Security Institution) list was removed. The conclusion and the strength of evidence did not change.
October 22, 2026
The names of the finding labels were simplified: the "Proven to work" label on this page became "Proven effect". The assessment did not change.
October 21, 2026
PubMed and ClinicalTrials.gov were searched for newly published and previously missed clinical studies. Study added: Devinsky 2018 (dose-ranging safety study). The conclusion and the strength of evidence did not change.
Related studies: Devinsky 2018 (dose-ranging safety study), published April 3, 2018 (PMID 29540584, doi:10.1212/WNL.0000000000005254, NCT02091206)
October 21, 2026
The full texts of the studies were reread and the information on the page was checked against its sources. Corrected: dose titration and participant details in the study table; drug interaction notes; the approved medicine's dosing schedule; the starting and titration summary; the "Who should be careful" list. Points that could not be verified from the full text went down from 4 items to 3. The conclusion and the strength of evidence did not change.
October 19, 2026
Page published. This first assessment reviewed 6 studies.

Evidence changes across all topics