What does the research say?
For drug-resistant epilepsies other than Dravet syndrome, Lennox–Gastaut syndrome and tuberous sclerosis, purified CBD medicine has been tried only in studies without a comparison group. These studies reported that seizures fell by between 36.5% and 51.4% at week 12. However, in well-designed studies, seizures also fell by 13–27% in people taking placebo; without a comparison group, it is impossible to know how much of the improvement came from the medicine.
Controlled studies are few and their results differ. In a study in Australia and New Zealand of 188 adults with drug-resistant focal epilepsy, a synthetic CBD gel rubbed onto the skin was tested at two different doses for 12 weeks; seizure frequency was no different from placebo at either dose. In a small study in Iran of 27 patients with frontal lobe epilepsy, CBD taken by mouth was reported to improve seizures and quality of life; however, the change in seizure frequency was not statistically different between the two groups, seizures were much more frequent in the CBD group at the start, and the publication is not clear about what the comparison group received. In a well-designed study of 162 adults with focal seizures using CBDV, a plant compound closely related to CBD, seizures also fell at a similar rate in both groups (40.5% vs. 37.7%). An expanded access program reported long-term reductions in seizures in 140 patients with focal epilepsy, but this program also had no comparison group.
Most of the data on unregulated oils sold on the market come from families' retrospective accounts. In one study, 47% of families who had moved to another state for this treatment reported a clear improvement, compared with 22% of families who already lived there; this suggests that expectations can influence the results. In a prospective study that also recorded seizures before treatment, the response rate was 24%; according to the authors, this rate was similar to the placebo groups in drug trials. In a pooled analysis of observational studies comparing CBD-rich extracts with purified CBD, the proportion of patients whose seizures fell by at least half was similar in both groups (37% vs. 42%); 'any improvement,' reported more often by extract users, was a loosely defined measure based mostly on family surveys. None of the pooled studies had a comparison group, and all three authors of the analysis had ties to cannabis product companies.
Some unregulated products contain THC, and the effect of THC-containing cannabis extracts has not been shown in controlled studies. In Australia, 29% of 58 product samples used by children with epilepsy had solvent residues above the accepted limit, and one contained pesticide residue. In the same research group's earlier study, families believed CBD was dominant in most of the extracts they used, but THC or the raw form of THC (THCA) was found in 98% of the samples; only 3 extracts came close to the lowest CBD dose used in studies. In Chile, homemade oils sent in by families contained CBD far below treatment levels, and 11 of 27 samples exceeded the 0.2% THC limit. In the US, the amount of CBD in 74% of 202 commercial CBD products differed from the label by at least 10% (an analysis funded by the maker of the approved medicine). In a study without a comparison group that prospectively followed 29 patients taking low-THC cannabis tablets, seizures did not change.
For these epilepsies, the largest body of data comes from real-world studies. In Germany, France, Denmark and Colombia, among patients using purified CBD off-label, the proportion whose seizures fell by at least half was reported as between 22% and 69%. In the comparisons in Germany and France, this rate was close to that seen in the approved diagnoses, while in the Danish study it was clearly lower. There are also small case series in epilepsy with myoclonic-atonic seizures, CDKL5 deficiency, Rett syndrome, 15q duplication, Angelman and Sturge–Weber syndromes, and FIRES. In a study of 20 children with childhood absence epilepsy, absence seizures counted on EEG did not decrease on average; in a pilot study of 9 babies with infantile spasms, the spasms did not stop in any of the babies. In epileptic spasms in children older than 2 years, more than half responded in a retrospective series. Because none of these studies had a comparison group, the evidence is weak.
Strength of evidence: ●●○○ Weak evidence. Finding: Mixed results. What does this mean?
Ongoing research on this topic
Trials testing CBD for this topic on ClinicalTrials.gov: 6 under way (2 recruiting); results are awaited from 1 that ended in the last two years. A trial being under way does not show that CBD works.
- NCT05324449: Phase 4, completed, results awaited, 20 participants
- NCT07723976: Phase 3, not yet recruiting, target 120 participants
- NCT07668856: Phase 3, recruiting, target 50 participants
Registries in other countries also list 2 recruiting trials on this topic: other registries.
Possible side effects and drug interactions
In studies of purified CBD, the most common side effects were sleepiness, decreased appetite, diarrhea and tiredness. In one study, serious events possibly related to the medicine occurred in 12% of patients; the most common was a long-lasting seizure that does not stop. In a study of children using unregulated oils, 13% of families reported an increase in seizures.
By slowing the breakdown of some medicines in the body, CBD can raise the blood levels of epilepsy medicines such as clobazam; together with valproate, it increases the risk of raised liver enzymes. Because the amount of CBD in unregulated products can be lower or higher than the label says, the size of these interactions is also unpredictable. In Europe, in about 39% of serious side effect reports involving unlicensed CBD products, the product was being used for epilepsy; the most commonly reported problems were mental health problems, liver problems and worsening of existing epilepsy, and the most common medicines used alongside were clobazam and valproate (spontaneous reports; frequency cannot be calculated).
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.
| 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 of purified CBD medicine, the dose was calculated in milligrams per kilogram of body weight per day. Treatment started at 2–10 mg per kilogram per day (study dose; not a personal recommendation) and could be increased, at rates that varied by center, up to 25–50 mg per kilogram per day (study dose; not a personal recommendation); there was no common schedule.
In children using unregulated oils, the highest dose reached was a median of 0.9 mg per kilogram per day (study dose; not a personal recommendation); in other words, far below the doses used in drug trials.
In the skin gel study, a total of 195 or 390 mg (study dose; not a personal recommendation) of CBD per day was applied, divided into two applications; in the frontal lobe epilepsy study, the dose taken by mouth was raised in weekly steps to 210 mg per day (study dose; not a personal recommendation).
Doses of products other than the approved medicine varied widely: in one US state's medical cannabis program, the median CBD dose among people getting products for seizures was 2.5 mg per kilogram per day (study dose; not a personal recommendation), about one quarter of the approved medicine's maintenance dose; 87% of patients had been given products that also contained THC. For CBDV, tested in focal seizures, the dose was raised up to 1,600 mg per day (study dose; not a personal recommendation).
Study details (for health professionals)
In adults with drug-resistant focal epilepsy, a synthetic CBD gel applied to the skin (a daily total of 195 or 390 mg, divided into two applications) did not change seizure frequency differently from placebo in a 188-person double-blind RCT. In a small 27-person controlled study in frontal lobe epilepsy, liposomal CBD by mouth (up to 210 mg/day) was reported to improve seizures; however, the change in seizure frequency over time was not statistically different between the two groups (p = 0.187), the difference in the proportion of patients who improved at week 8 was also not significant (p = 0.069), at baseline the CBD group's monthly seizure count was about 2.5 times that of the control group (86.5 vs. 34.0), and the publication is not clear about whether the control group received placebo. An expanded access program reported seizure reductions for up to 144 weeks in 140 patients with focal epilepsy, but there is no control group. In other drug-resistant epilepsies such as CDKL5 deficiency, Aicardi, Dup15q and Doose, there are only open-label data without a placebo arm; seizure reductions were reported, but because reductions of 13–27% were also seen in the placebo arms of the approved medicine's RCTs, these findings cannot be counted as a drug effect. There is no approval for these epilepsies. Data on unregulated CBD/cannabis oils are mostly retrospective parent reports; in a prospective study that recorded seizures for 4 weeks before treatment, the response rate was at placebo level, and increased seizures have been reported. The manufacturer's studies in epilepsy with myoclonic-atonic seizures (NCT05288283; 2 patients randomized, ended by a business decision) and in children under 2 years (NCT04485104; 3 patients, ended due to low enrollment) produced no interpretable efficacy data. In a pilot study in infantile spasms (9 babies, 40 mg/kg/day, 2 weeks), the spasms did not stop in any baby; the randomized phase was not started. In a double-blind RCT in 162 adults with focal seizures using CBDV, a plant compound closely related to CBD, seizures fell similarly in both groups (40.5% vs. 37.7%; p = 0.648; manufacturer funding). Old controlled trials of pure CBD by mouth are very small: the 15-patient placebo-controlled study from 1980 has no statistical comparison; together with three small trials from the same period, one published as a letter to the editor and one as a conference abstract (48 patients in total), a Cochrane review (search 2013) wrote that no reliable conclusions can be drawn from these reports. In childhood absence epilepsy, in an uncontrolled study of 20 children by the maker of a synthetic CBD (results only on ClinicalTrials.gov, no peer-reviewed publication), the mean change in absence seizures counted on EEG, one of the primary outcomes, was not in the direction of a decrease (+41% and +35% in the two dose groups; highly variable from patient to patient), while the other primary outcome, time to first absence with hyperventilation, became longer; because the two outcomes pointed in different directions, the result is unclear. In real-world studies (Germany, France, Denmark, Colombia), the proportion of patients with off-label diagnoses whose seizures fell by at least half was reported as between 22% and 69%; in the comparisons in Germany and France, this rate was close to that in approved diagnoses, while in the Danish study it was clearly lower (22% vs. 70% at month 3). None of these studies has a control group; seizure outcomes rely on doctors' notes, seizure diaries and family reports, and some of those who stopped early are left out of the analysis. Two systematic reviews pooling epilepsies other than Dravet syndrome and LGS (2021 and 2026; the second was funded by the manufacturer) found only open-label data apart from TSC; 33 of the 57 studies in the 2026 review were case reports or small series. The UK's national guideline (NICE 2019) made no practice recommendation for cannabis-based products in these epilepsies, nor a recommendation against their use; it called for research into the efficacy of CBD in these epilepsies and into the effect of adding THC to CBD compared with CBD alone and with placebo. There are no controlled studies of CBD-rich products that contain THC or have an unknown THC content: in the only structured study, which prospectively followed 29 patients on a low-THC product, seizures did not change; the high response rates reported in open-label series in Israel and Colombia and the improvement reported in a national cohort in Thailand have no comparison group (evidence for THC is very weak). In unregulated products: in a pooled analysis of observational studies comparing CBD-rich extracts with purified CBD, the proportion of patients whose seizures fell by at least half was similar in both groups (37% vs. 42%; uncontrolled studies, and all three authors of the analysis had ties to cannabis companies); in product analyses, content often differed from the label, and THC, solvents, heavy metals and pesticides were found; in spontaneous side effect reports in Europe, in about 39% of serious reports involving unlicensed CBD, the product was being used for epilepsy. October 2026 search (PubMed, Cochrane, ClinicalTrials.gov): 4 controlled studies with cannabinoids on this topic (one with CBDV instead of CBD) and more than 30 systematic reviews of cannabinoid use in epilepsy were found (most rely on studies in Dravet syndrome, LGS and TSC); all the controlled studies, the 4 reviews most relevant to these epilepsies (including Cochrane) and the UK's national guideline are in the table below. The three small trials from 1978–1990 in the Cochrane review are not separate rows; they are mentioned in the review's row. The other studies in the table are open-label or observational studies without a comparison group, and product analyses; later publications from the same programs and studies with overlapping patients are in the other sources section of the relevant row.
Evidence assessment: Weak (mostly uncontrolled, open-label; Class III); in adult focal epilepsy, two large RCTs (CBD applied to the skin and CBDV, a compound closely related to CBD) were negative; in real-world studies, response rates reported in off-label diagnoses were mostly close to those in approved diagnoses (clearly lower in one study), but all are uncontrolled. Unregulated ('artisanal') CBD/cannabis oils: Very low
CBD alone
Purified medicinal CBD has mostly been tested in uncontrolled studies in other drug-resistant epilepsies; in adult focal epilepsy, CBD applied to the skin in a 188-person RCT, and CBDV, a compound closely related to CBD, in a 162-person RCT, were no different from placebo; in real-world studies, seizures fell by at least half in 22% to 69% of patients using it off-label, but none of these studies has a comparison group; with unregulated CBD oils, the effect stayed at placebo level, and dose and content are unreliable.
Products containing THC
Some unregulated products contain THC; the effect of THC-containing cannabis extracts has not been shown in controlled studies: in the only structured study, which prospectively followed 29 patients taking low-THC tablets, seizures did not change; the high response rates reported in open-label series in Israel and Colombia with CBD-rich oils containing small amounts of THC have no comparison group. In Australia, families believed CBD was dominant in most of the extracts they used, but THC or the raw form of THC (THCA) was found in 98% of the samples.
Evidence: Very weak (no controlled studies; no effect was seen in the only prospective structured study, and the other data come from uncontrolled series and unregulated oils)
Approved drug regimen
Not approved. CBD's approval as a medicine covers only seizures in Dravet syndrome, Lennox–Gastaut syndrome and TSC, and covers only purified, standard-dose medicinal CBD (Epidiolex/Epidyolex, 100 mg/ml). In Turkey, medicinal CBD is not approved (list of TİTCK, the Turkish Medicines and Medical Devices Agency, October 16, 2026). On the TİTCK Foreign Active Ingredient List (September 28, 2026), cannabidiol oral solution (YDN02BG1099) is listed as a red-prescription medicine and among the medicines that cannot be imported without TİTCK's written approval.
Dose range in studies
In open-label studies, starting at 2–10 mg/kg/day, up to a ceiling of 25–50 mg/kg/day; in children using unregulated oils, median highest dose 0.9 mg/kg/day; CBD gel on the skin, a daily total of 195 or 390 mg, divided into two applications (focal epilepsy RCT); in frontal lobe epilepsy, 70–210 mg/day by mouth; CBDV in focal seizures up to 1,600 mg per day (RCT); synthetic CBD in childhood absence epilepsy 20 or 30 mg/kg/day
Starting dose and titration
In open-label studies, gradual increase varying by center; no common schedule reported
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.
| Study | Participants | Product | Dose and titration | Duration | Result | Source |
|---|---|---|---|---|---|---|
| O'Brien 2022 (STAR 1) Multicenter, double-blind, placebo-controlled RCT, 3 arms | 188 (195 mg 63, 390 mg 62, placebo 63) Adults with drug-resistant focal epilepsy (Australia, New Zealand) | Synthetic CBD gel, applied to the skin (transdermal); pharmaceutical grade | Daily total of 195 mg or 390 mg, divided into two applications (per the journal's correction) | 12 weeks (followed by an open-label extension of up to 2 years) | Seizure frequency no different from placebo at either dose (2.51 and 2.59 vs. 2.49 per 28 days); ≥50% reduction in 60.8% at month 6 of the open-label extension (uncontrolled) Side effects: Treatment-related side effects 33.9% at 390 mg, 17.5% at 195 mg, 12.7% with placebo; stopped because of side effects 3–5% | PMID 35802375, doi:10.1001/jamanetworkopen.2022.20189, ACTRN12616000510448 |
| Ebadi 2023 Triple-blind, controlled study (Iran) | 27 (12 CBD / 15 control) Drug-resistant frontal lobe epilepsy (adults) | Liposomal pure CBD (99.95%, contains no THC), oral solution (40 mg/mL) | 70 mg/day (week 1) → 140 mg/day (week 2) → 210 mg/day, once daily Weekly increase: 70 → 140 → 210 mg/day | 8 weeks | At week 4, seizure improvement 66.7% with CBD and 20% in controls (difference not significant at week 8, p = 0.069); the between-group difference in the change in seizure frequency over time was not significant (p = 0.187); baseline monthly seizure count 86.5 with CBD and 34.0 in controls; seizure severity did not change; at week 8, more patients improved in quality of life with CBD Side effects: Found to be safe; side effect numbers not in the publication | PMID 37470002, doi:10.3389/fneur.2023.1143783, IRCT20210608051515N1 |
| Devinsky 2016 Open-label, uncontrolled; 11 US centers | 214 enrolled; 162 safety, 137 efficacy analysis Mixed drug-resistant epilepsies, ages 1–30 | Oral CBD (pure CBD) | Starting at 2–5 mg/kg/day, up to intolerance or a ceiling of 25 or 50 mg/kg/day By center | 12 weeks | Median motor seizures 30.0 → 15.8/month; median reduction 36.5% (no control group) Side effects: Side effects 79%: sleepiness 25%, decreased appetite 19%, diarrhea 19%, tiredness 13%, convulsion 11%. Serious side effects 30%; severe events possibly related to the medicine 12% (most often status epilepticus, 6%). 1 SUDEP (unrelated) | PMID 26724101, doi:10.1016/S1474-4422(15)00379-8, Single-center reports from centers that took part in the same program (some patients shared): Sands 2019, PMID 30460546, UCSF, 26 children under individual patient and expanded access permits between April 2013 and December 2014, mean 21 months): at month 24, 9 of the 26 children were still on treatment, and 7 (26.9%) had a lasting reduction of more than half in motor seizures, 57.7% stopped due to insufficient effect, side effects 80.8%, serious side effects 23.1% (3 status epilepticus, 2 catatonia). Patel 2021, PMID 34050682, MGH, 54 patients, median 45.5 months): responder rate similar at year 1 and at the last visit (41.7–42.6%), no control group |
| Szaflarski 2018 Interim analysis of an open-label expanded access program, 25 US centers | 607 (safety) Drug-resistant epilepsies; mean age 13 (0.4–62) | Oral CBD (pure CBD) | Starting at 2–10 mg/kg/day → ceiling of 25–50 mg/kg/day; median 25 mg/kg/day By center | Median 48 weeks | At week 12, 51% reduction in convulsive seizures and 48% in all seizures; similar up to 96 weeks. ≥50% / ≥75% / 100% response: 52% / 31% / 11% (no control group) Side effects: Diarrhea 29%, sleepiness 22%. 24% stopped (lack of effect 15%, side effects 5%) | PMID 29998598, doi:10.1111/epi.14477, Final pooled analysis of the same program: Szaflarski 2023, PMID 36537757, 35 US centers, 892 patients, up to January 2019, median 694 days): 36% stopped (insufficient effect 19%, side effects 7%), median reduction in convulsive seizures 50–67%, at least 50% reduction 51–59%, side effects 88%, serious side effects 41%, 20 deaths judged by investigators as unrelated to treatment, no control group. Funding also includes GW/Jazz and Alabama's Carly's Law fund, so the UAB patients in Gaston 2021 may be included in this pool, PMID 38420790, PMID 37243404, absence and myoclonic seizures and epileptic spasms, at least 50% reduction in about half of patients, PMID 38572913 |
| Patel 2025 (expanded access, focal epilepsies) Subanalysis of an open-label expanded access program (the same program as Devinsky 2016 and Szaflarski 2018; no control group) | 140 efficacy (TSC 33, other focal epilepsies 107); 148 safety Drug-resistant focal epilepsies (ages 2–73) | Epidiolex (100 mg/mL oral solution) | Starting at 2–10 mg/kg/day, up to a ceiling of 25–50 mg/kg/day; median 23–25 mg/kg/day Gradual increase up to the level the patient tolerated | Up to 144 weeks | In focal epilepsies other than TSC, median reduction in focal seizures 46–75%; 38% stopped (lack of effect 18%, side effects 9%); no control group Side effects: Side effects 95% (most often diarrhea 48%); serious side effects 40%; stopped because of side effects 8% | PMID 40673944, doi:10.1111/epi.18496 |
| Devinsky 2018 (CDKL5, Aicardi, Dup15q, Doose) Open-label, compassionate use/EAP | 55 safety (CDKL5 20, Aicardi 19, Dup15q 8, Doose 8) Ages 1–30 | Epidiolex (pure CBD) | Not reported (in the abstract) not reported | 48 weeks (efficacy) | Convulsive seizure reduction: week 12 51.4% (n=35), week 48 59.1% (n=27). Authors: Class III evidence; RCTs needed Side effects: 27% stopped by week 144 (4 side effects, 9 lack of effect) | PMID 30006259, doi:10.1016/j.yebeh.2018.05.013 |
| Gaston 2021 Open-label (UAB EAP), 2 years | 169 (89 children, 80 adults) Drug-resistant epilepsy | Epidiolex 100 mg/ml (pure CBD) | Starting at 5 mg/kg/day → ceiling of 50 mg/kg/day According to seizure control and tolerability | Up to 2 years | Seizure reduction in children 52% (month 1), 51% (year 1), 75% (year 2); in adults 60%, 81%, 85% (no control group) Side effects: Diarrhea, sedation, decreased appetite | PMID 33667843, doi:10.1016/j.yebeh.2021.107862, Drug interaction analyses from the same UAB program: Gaston 2017, PMID 28782097, 39 adults, 42 children): as the CBD dose increased, topiramate, rufinamide and N-desmethylclobazam levels rose and the clobazam level fell, in adults, zonisamide and eslicarbazepine levels also rose, changes other than clobazam and its metabolite stayed within the therapeutic range, AST/ALT were higher in those taking valproate. Final analysis Gaston 2023, PMID 37593907, 169 people, 80 adults): with CBD dose, levels of clobazam, N-desmethylclobazam, free valproate, felbamate and topiramate rose (levetiracetam in children only, perampanel in adults only), according to the authors, the clinical importance is probably small. The program was funded by the state of Alabama (Carly's Law), and the medicine was donated by GW. This UAB program is the Alabama arm of the US expanded access program in row 1768 (Carly's Law funding), some patients are very likely also counted in that row |
| Knupp 2019 Prospective observation | 21 (32 consented; 11 excluded from the analysis) Children with epilepsy using unregulated oral cannabis extracts | Unregulated cannabis extract (CBD and THC-COOH measured) | Median highest CBD dose used 0.9 mg/kg/day not reported | 12 weeks (after a 4-week baseline observation) | Response rate 24%; according to the authors, similar to the placebo rates in RCTs of medicinal CBD. Blood CBD/THC-COOH levels were not related to response Side effects: Stopped early in 3 children (14%) because of perceived increase in seizures | PMID 31550641, doi:10.1016/j.seizure.2019.09.007 |
| Press 2015 Retrospective parent report (Colorado) | 75 Children with drug-resistant epilepsy | Oral cannabis extracts (unregulated) | not reported not reported | not reported | 33% reported ≥50% reduction; 47% among families who moved to Colorado for this treatment and 22% among those who already lived there (a sign of an expectation effect). No improvement in EEG background Side effects: Side effects 44%; increased seizures 13%, sleepiness/tiredness 12%. Rare: developmental regression, abnormal movements, status epilepticus requiring intubation, death | PMID 25845492, doi:10.1016/j.yebeh.2015.02.043, Overlapping cohort from the same center: Treat 2017, PMID 27859038, moving to Colorado was associated with perceived benefit (65% vs. 38%, p = 0.01) but was not independently associated with longer use, a Dravet diagnosis was associated with shorter use (p = 0.02), according to parent reports, response 24%, side effects 19% (most often sleepiness and increased seizures) |
| Porcari 2018 Retrospective | 108 Children with epilepsy | Unregulated (artisanal) CBD preparations; content not verified | not reported not reported | not reported | ≥50% seizure reduction reported in 39% and seizure freedom in 10%; uncontrolled and open to bias Side effects: The most commonly reported side effect was sedation (<4%; all were also taking clobazam) | PMID 29429908, doi:10.1016/j.yebeh.2018.01.026 |
| Sulak 2017 Retrospective patient data (Washington and California) and 4 case reports; no control group | 272 Patients with epilepsy | Unregulated (artisanal) cannabis products; content not verified | not reported not reported | not reported | Among 272 patients from Washington and California, more than 50% seizure reduction was reported in 146 (54%), complete response in 26 (10%) and no effect in 37 (14%) (two percentages in the abstract do not match the numbers; the numbers were used); most patients used CBD-rich handmade formulations, and some used products with added THC or THCA; uncontrolled, retrospective and open to bias Side effects: According to the authors, mild and infrequent | PMID 28254350, doi:10.1016/j.yebeh.2016.12.032 |
| Suraev 2022 (PELICAN subanalysis) Product analysis (contaminant screening) | 78 samples (tested for solvents in 58, heavy metals in 51, pesticides in 31 samples) Unregulated cannabis products used in children with epilepsy in Australia | Unregulated cannabis products | Ethanol or isopropanol above the accepted limit (5000 ppm) in 17/58 samples (29%); bifenthrin (a pesticide) at 4.9 times the limit in one sample. No samples exceeded heavy metal limits (51 samples). Most pesticide tests were inconclusive | PMID 34954507, doi:10.1016/j.yebeh.2021.108496, Epilepsy Behav 2022, 127:108496 (online December 2021). Earlier study in which the samples were collected (same research group, Suraev 2022 describes its samples as collected in an earlier study, 78 samples in 2022, 61 in 2018): Suraev 2018, PMID 29977078, PMID 30068968, 61 samples from 38 families (51 extracts when combined) were analyzed. Families believed CBD was the main component in 71% of the extracts, but 98% of the samples contained THC and/or its raw form THCA, CBD was the dominant cannabinoid in only 23%, was absent in 4 samples, and only 3 extracts (6%) came close to the lowest CBD dose in studies (10 mg/kg/day). Families rated 75% of the extracts as effective, the cannabinoid content of extracts rated effective and ineffective did not differ. No control group, results rely on family reports | ||
| GWPCARE7 pilot + extension (infantile spasms) Open-label pilot phase and open-label extension (uncontrolled); results from the ClinicalTrials.gov record | 9 Infantile spasms, 1–24 months | Oral CBD solution (GW; pure CBD) | Target 40 mg/kg/day not reported | Pilot 2 weeks; extension up to 12 months | In the pilot, spasms did not stop and hypsarrhythmia did not resolve in any of the 9 babies; the study met the no-go criterion. In the extension, at the day 379 measurement in the registry, 3 of the 7 babies who could be assessed had no spasms and their hypsarrhythmia had resolved; however, only 2 of the 9 babies completed the extension (3 left due to lack of effect), this measurement may also include babies who stopped treatment; no control group Side effects: In the pilot, 1 status epilepticus (serious). In the extension, serious side effects in 2 of 9; 3 babies stopped due to lack of effect | NCT02953548, NCT02954887 |
| Cunha 1980 Double-blind, placebo-controlled randomized study (phase 2; phase 1 in healthy volunteers) | 15 patients (CBD 8, placebo 7; the wording in the abstract can also be read as a placebo group of 8 patients); also 16 healthy volunteers in phase 1 Epilepsy with secondary generalization from a temporal focus that could not be controlled with medicines | Pure CBD by mouth (capsules) | 200–300 mg/day not reported (in the abstract) | Up to 4.5 months | Of the 8 patients taking CBD, 4 had almost no major seizures throughout the study, 3 had partial improvement and 1 showed no effect; in the placebo group, 1 patient had clear improvement and the others had no change. The abstract has no statistical comparison, and the size of the placebo group is inconsistent (15 patients were divided into two groups, the CBD group 8; for placebo it says 'no change in 7 patients, improvement in 1 patient'). No randomization details are given; the Cochrane review (Gloss 2014) rated this report as low quality. Patients continued their previous epilepsy medicines. Side effects: No serious side effects or signs of toxicity were reported (monitored with physical exams, blood and urine tests, EEG and ECG) | PMID 7413719, doi:10.1159/000137430, A review that also describes the same trial: Carlini and Cunha 1981, PMID 7028792, improvement in 7 of the 8 patients taking CBD and in 1 patient taking placebo) |
| Brodie 2021 (CBDV, focal seizures) Phase 2, multicenter, double-blind, placebo-controlled randomized study (centers in Czechia, Hungary, Poland, Spain and the UK) | 162 randomized (CBDV 81, placebo 81); 162 analyzed Adults (aged 18–65; mean age 36) with at least 1 focal seizure per week despite at least 2 medicines | Cannabidivarin (CBDV; GWP42006), oral solution 50 mg/mL | 400 mg twice a day (week 1) → 600 mg twice a day (week 2) → 800 mg twice a day (1,600 mg per day) for 6 weeks; mean daily dose 1,480 mg Weekly increase: 400 → 600 → 800 mg, twice a day | 4 weeks of baseline observation, 8 weeks of treatment (2 weeks of dose increase + 6 weeks at a fixed dose), followed by 12 days of dose reduction | Focal seizure frequency fell by 40.5% in the CBDV group and 37.7% in the placebo group; the difference was not significant (treatment ratio 0.95; 95% CI 0.78–1.17; p = 0.648). At least 50% reduction: 36% with CBDV, 33% with placebo. No difference in any seizure type or in secondary outcomes (seizure-free days, rescue medication use, doctor and patient global impression, cognitive tests). The authors note that the placebo response was higher than in previous epilepsy studies. Funded by the manufacturer (GW Research); 3 of the 8 authors are GW employees and hold share options. Side effects: Side effects 73% with CBDV, 48% with placebo (most often diarrhea 25% vs. 7%, sleepiness 15% vs. 3%); stopped because of side effects 14% vs. 3%; serious side effects 4% vs. 1% (in the CBDV group, one patient with a picture of severe drug toxicity and one with kidney failure together with diarrhea; both resolved when the medicine was stopped). Liver enzymes (ALT or AST) >3×ULN in 3 patients with CBDV (2 stopped) and 1 patient with placebo; all were also taking another medicine that can affect the liver | PMID 33998885, doi:10.1089/can.2020.0075, NCT02365610, no results in the registry or PubMed). The journal added a correction note for Figure 2 (April 15, 2021) |
| Gloss 2014 (Cochrane) Cochrane systematic review (randomized studies only) | 4 randomized studies, 48 patients in total Any type of epilepsy (add-on treatment or monotherapy); before the approved CBD medicine's Dravet and LGS studies | Pure CBD, by mouth (in all four trials) | 200–300 mg/day | Search up to September 9, 2013; the trials were short | All four reports provided data only on the side effect question; the review's main outcome (seizure freedom for at least one year) and the response rate could not be measured in any of them. The review wrote that no reliable conclusion can be drawn on whether cannabinoids are effective in epilepsy. All the reports were of low quality: one is only a conference abstract and one a letter to the editor; none gives randomization details, and whether the groups were similar at baseline was not checked. Because the search was done in September 2013, it does not include the later studies of the approved CBD medicine; there is no newer version on PubMed. The authors have no known conflicts of interest. Side effects: According to the review, no side effects were reported in patients taking CBD; because 200–300 mg/day of CBD was given to a small number of patients for a short time, long-term safety could not be assessed | PMID 24595491, doi:10.1002/14651858.CD009270.pub3, PMID 22696383, PMID 7413719, a separate row in the table), PMID 3941934, letter to the editor), Trembly 1990 (conference abstract only) |
| Lattanzi 2021 (epilepsies other than Dravet and LGS) Systematic review (no meta-analysis); 42 studies: 3 RCTs (one an efficacy study in TSC, GWPCARE6, two drug interaction studies), open-label studies and their subanalyses, retrospective chart reviews, clinical series and case reports | 42 studies; pooled number of patients not given Epilepsies other than Dravet and LGS (including TSC); children and adults | Only the approved medicinal CBD (plant-derived purified, >98%, oral solution in sesame oil) | Up to 50 mg/kg/day | Search up to October 2020 | Placebo-controlled evidence exists only for TSC (GWPCARE6); data for other conditions are open-label and uncontrolled. 18 of the 42 studies report results separately by condition (CDKL5, Aicardi, Doose, Dup15q, epilepsy with myoclonic absences, Sturge–Weber, SYNGAP1, migrating focal seizures, SCN8A, epileptic spasms, cortical dysplasia, FIRES, status epilepticus), and 24 report drug-resistant epilepsies of mixed causes together. In infantile spasms, the placebo-controlled study met the 'no-go' criterion in its pilot phase and was stopped. The authors write that open-label studies provide Class III (weak) evidence, that the lack of a control group and expectations create bias, and that the results cannot be applied to other cannabis products. 3 of the 7 authors received fees or research support from the CBD manufacturer GW; the review had no funding. Side effects: Most often sleepiness (more common with clobazam), decreased appetite, diarrhea and raised liver enzymes (mostly in those taking valproate); one center reported low platelets together with valproate | PMID 33754312, doi:10.1007/s40263-021-00807-y |
| Coppola 2026 (DEE and complex drug-resistant epilepsies) Systematic review (narrative synthesis; no meta-analysis) | 57 studies reporting seizure outcomes, 37 different epilepsy types, 971 patients; 33 of the studies were case reports or small series Developmental and epileptic encephalopathies and complex drug-resistant epilepsies other than Dravet, LGS and TSC; most often focal/multifocal epilepsy (401 patients) and Angelman syndrome (188 patients) | Approved medicinal CBD (plant-derived purified oral solution) | not reported (in the abstract) | Search up to March 2024 | 47 of 57 studies reported a seizure reduction in at least one patient; definitions of reduction differed from study to study (reduction in 20–100% of patients; mean or median reduction 12–99%). In 22 studies, at least one patient was seizure-free for 48 days or longer. The authors write that CBD may reduce seizures in these epilepsies and support new studies; most of the data come from small series without a control group. The review was funded by the manufacturer (Jazz Pharmaceuticals); 2 of the 4 authors are Jazz employees and shareholders, and 2 have received consulting fees or research support from Jazz. Side effects: Mostly mild or moderate; most often digestive: diarrhea 17–50%, decreased appetite 7–45%, vomiting 5–86% (by study) | PMID 41558068, doi:10.1016/j.eplepsyres.2026.107731, PMID 41820159, content not on PubMed, could not be read). Non-seizure outcomes from the same systematic search: Samanta 2026, PMID 42520581, 32 studies, 1, 343 patients, most studies reporting improvements in cognitive function, behavior, communication and quality of life were observational and at moderate-to-high risk of bias, same authors and same company ties) |
| Pamplona 2018 (CBD-rich extracts and pure CBD) Meta-analysis of observational studies (11 studies; categorical comparison with Fisher's test) | 670 patients (11 studies: 5 pure CBD, 6 CBD-rich cannabis extract) Drug-resistant epilepsy (mostly children; mixed, including Dravet and LGS) | Purified CBD and CBD-rich cannabis extracts | 1–50 mg/kg/day; mean 25.3 for pure CBD and 6.0 mg/kg/day for extracts (CBD equivalent) | Search February and December 2017; treatment durations 3–12 months (mean 6.2 months) | Reports of 'any improvement' were 71% (318/447) with extracts and 46% (81/175; p < 0.0001) with pure CBD; on the standard outcome of at least 50% reduction, there was no difference: 37% (122/330) with extracts and 42% (94/223; p = 0.52) with pure CBD; response rate across all patients 39%. None of the studies had a control group; the authors write that the placebo effect could not be measured and that the interpretation that extracts may be more effective needs to be confirmed in a controlled study. All the extract studies were family surveys or retrospective chart reviews (two publications from the same Colorado cohort, Press 2015 and Treat 2017, were counted separately); the pure CBD studies were prospective drug programs. All 3 authors have ties to cannabis product companies (2 to Entourage Phytolab, 1 to Bedrocan). Numbers are per the journal's correction. Side effects: Mild side effects reported in 76% (158/216) of pure CBD studies and 33% (148/447) of extract studies; severe side effects 26% (41/155) and 7% (23/328) (p < 0.0001); side effects were recorded in different ways in studies with different designs | PMID 30258398, doi:10.3389/fneur.2018.00759, Correction: Front Neurol 2018, PMID 30687209, PMC6335388, numbers in Tables 2 and 4 and the abstract were corrected, according to the authors, the conclusion did not change) |
| NICE 2019 (NG144 guideline) National guideline (based on an evidence review) | Evidence review D: 4 RCTs (2 Dravet, 2 LGS; as indirect evidence) + 11 single-arm observational studies (8 prospective, 2 retrospective, 1 unclear) Severe, drug-resistant epilepsy (children and adults); United Kingdom | Cannabis-based medicinal products: pure CBD and THC:CBD extracts | Search December 19, 2018 – January 21, 2019; guideline November 2019 (last updated on the page March 22, 2021) | No practice recommendation was made for cannabis-based products other than approved CBD: the committee stated that the observational evidence was of very low quality (all single-arm, at high risk of bias) and that, because the disease mechanisms differ, Dravet/LGS studies cannot serve as a basis for recommendations in other epilepsies. No recommendation against use was made either: this would have stopped treatment for patients thought to be benefiting and would have blocked research; the committee was also concerned that a recommendation against might steer families toward products without a prescription (online or over the counter). It asked that people seeking treatment for severe epilepsy be told there is no clear evidence on the safety and efficacy of these products. The committee wrote that the amounts of CBD and THC may vary in products that are not pharmaceutical grade, especially whole-plant extracts. Research recommendations: the efficacy of CBD in different epilepsies and the effect of adding THC compared with CBD alone and with placebo. For approved CBD, the guideline refers to the technology appraisals for Dravet, LGS and TSC. Side effects: In the observational studies, side effects were reported in up to 98% of patients; because there was no control group, how much was due to the product could not be determined | PMID 31909928, Evidence review D: epilepsy (NICE, November 2019, PMID 35107910, NBK577085). Guideline: nice.org.uk/guidance/ng144 | |
| Absence epilepsy, synthetic CBD (NCT03336242) Phase 2, open-label, non-randomized dose-finding study (two dose groups); results from the ClinicalTrials.gov record, no peer-reviewed publication | 20 (10 + 10); 19 completed (1 patient left because of side effects); EEG absence count analyzed in 19 patients, hyperventilation outcome in 17 Drug-resistant childhood absence epilepsy with at least 2 medicines tried; ages 3–17 (mean 8.4), 11 centers in the US | Synthetic (not plant-derived) pharmaceutical-grade CBD oral solution; manufacturer INSYS Therapeutics (current sponsor in the registry Radius Pharmaceuticals) | 20 mg/kg/day or 30 mg/kg/day, divided into two doses per day; the planned 10 mg/kg/day group was not opened In the 30 mg/kg/day group, started at 20 mg/kg/day for 5 days | 4 weeks of treatment (after a 5–10-day dose increase); up to about 54 weeks in the extension study | On one of the primary outcomes, the mean change in the number of absence seizures on 4-hour video-EEG pointed toward an increase, not a decrease: +40.7% (SD 180.9; 10 patients) in the 20 mg/kg/day group and +35.1% (SD 121.5; 9 patients) in the 30 mg/kg/day group; in the registry, a negative value means improvement. The standard deviations are several times the mean: the response is highly variable from patient to patient; the median and how many patients had a reduction are not in the registry. On the other primary outcome, time to first absence with hyperventilation lengthened by a mean of 162% (SD 344; 9 patients) and 65% (SD 116; 8 patients); a longer time is read as improvement, meaning the two outcomes point in different directions. Seizure-free at week 4 according to diaries: 1 of 10 patients and 0 of 9 patients. Family and doctor global impression scores averaged 2.9 and 3.1 (1–7 scale; 3 = minimally improved). According to the sponsor's rationale, more patients showed a clinically meaningful reduction in seizure count in the 20 mg/kg/day group than in the 30 mg/kg/day group, and the planned third group (10 mg/kg/day) was not opened; the extension study was terminated because the manufacturer filed for bankruptcy protection. No control group, no statistical comparison was made. Side effects: Side effects in 6 of 10 patients in the 20 mg/kg/day group and 7 of 10 patients in the 30 mg/kg/day group; no serious side effects. In the 30 mg/kg/day group, 1 patient left because of side effects; in the same group, raised liver enzymes were recorded once each under four separate terms (ALT, AST, liver enzyme, liver test) (whether it was the same patient is not stated); seizures were recorded as a side effect in 2 patients. In the extension (11 patients), no serious side effects; 3 patients left due to lack of effect | NCT03336242, 2023), NCT03355300, results June 18, 2023). No peer-reviewed publication of this study was found in PubMed (October 28, 2026) |
| Scheffer 2021 (CBD gel on the skin, developmental and epileptic encephalopathy) Open-label, two-center, non-randomized before-and-after study (Melbourne and Wellington); no control group | 48 (seizure analysis 46; 33 patients with focal impaired awareness and tonic-clonic seizures analyzed separately) Children and adolescents aged 3–18 with developmental and epileptic encephalopathy (8 Dravet, 5 Lennox–Gastaut, 6 myoclonic-atonic epilepsy, 3 West, 26 other); they were taking 1–4 seizure medicines | Synthetic CBD gel, applied to the skin (4.2%; ZYN002, Zynerba) | Applied twice a day, total 250–1000 mg/day (by weight) Start 250 mg/day (≤25 kg) or 500 mg/day (>25 kg); could be increased to 500/750 at month 1 and to 750/1000 mg/day at month 2.5 | 6.5 months (1 month baseline observation, 1 month dose increase, 5.5 months maintenance at a flexible dose) | Median reduction in all countable seizures over 6.5 months was only 12.3% (IQR −18.6 to 46.3). Focal impaired awareness and tonic-clonic seizures, where the best response was seen, were analyzed separately afterward: in these 33 patients, median reduction 43.5% (IQR −23.8 to 57.5), 57.7% at month 5; at least 50% reduction 42.4% at month 2 and 62.5% at month 5. Improvement was smaller in those taking clobazam (26.8% vs. 45.5%). Sleep scale scores improved. No control group; the authors note that dropout of non-responders, expectation effects and regression to the mean may have affected the result. In the controlled study in adult focal epilepsy (O'Brien 2022), the same gel was no different from placebo. The study was designed and funded by the manufacturer (Zynerba); 7 of the 9 authors have ties to the company (2 are employees and hold patents and shares). Side effects: Any side effect 96% (29% in the baseline observation month); treatment-related side effects 60% (29/48), 96% mild-moderate: dryness and pain at the application site, sleepiness (8% each; all 4 patients with sleepiness were taking clobazam); treatment-related diarrhea in 1 patient. 15 serious side effects in 10 patients (21%) (10 infections, 4 seizure-related); 2 may be drug-related: nonconvulsive status epilepticus and lower respiratory tract infection. 8 patients (17%) stopped (7 lack of effect, 1 skin reaction) | PMID 34477852, doi:10.1001/jamanetworkopen.2021.23930, ACTRN12618000516280 |
| Kaplan 2017 (Sturge-Weber syndrome) Open-label expanded access study (under an FDA investigational permit, single center, Kennedy Krieger); no control group; 8-week baseline observation | 5 started; data for 4 patients at week 14 (according to the registry, 1 patient was withdrawn at week 9 due to lack of effect) Patients with Sturge-Weber syndrome with brain involvement and drug-resistant epilepsy; according to the registry, mean age 8.8 (4 under 18, 1 adult) | Epidiolex (purified CBD, oral solution) | According to the registry, at most 25 mg/kg/day According to the registry, started at 2 mg/kg/day; after 7 days, increased by 3 mg/kg/day to 5 mg/kg/day, then increased by 5 mg/kg/day each week up to at most 25 mg/kg/day | Week 14 assessment; those remaining were on CBD for 63–80 weeks | Motor seizures fell by more than half in 2 of 4 patients at week 14 and in 3 patients (those with brain involvement on both sides) at the last check; these patients reported improved quality of life. In the registry, the seizure reductions of the 4 patients at the last check were 12%, 100%, 83% and 64%. One of the 4 patients in the analysis first left due to lack of effect and then restarted at a lower dose; according to the registry, the fifth patient was withdrawn at week 9 due to lack of effect (side effects are also mentioned in one place) and was not included in the analysis. Very small and uncontrolled; the medicine's manufacturer (today Jazz) is a collaborator in the registry. Side effects: According to the article, mild drug-related side effects in 3 patients. In the registry, serious events in 3 of 5 patients (eye twitching, pneumonia, facial skin infection; relationship to the medicine not stated); temporary increase in seizures 3, behavior problems 2, raised AST 1 patient | PMID 28454984, doi:10.1016/j.pediatrneurol.2017.02.009, NCT02332655, collaborators Jazz Pharmaceuticals and the Faneca 66 Foundation). Later study by the same team: Smegal 2023, PMID 36508880, NCT04447846, no seizures were seen, and improvements in neurological score, quality of life and anxiety were reported, on the working memory test, the registry's primary outcome, the mean T score changed from 37.1 to 39.1 in 7 patients (no statistics given), 1 patient stopped because of a moderate side effect, open-label and uncontrolled |
| Gofshteyn 2017 (FIRES) Multicenter case series; open-label use under emergency or expanded access protocols (5 centers); no control group | 7 Children with febrile infection-related epilepsy syndrome (FIRES) who had not responded to medicines and other treatments; in the acute (non-stop seizures) or chronic phase of the disease | Epidiolex (purified CBD) | Not reported (in the abstract) not reported | Not reported (in the abstract) | After CBD was started, the frequency and duration of seizures fell in 6 of 7 children; on average, 4 seizure medicines could be stopped. At the latest assessment, 5 children could walk, 1 walked with help and 4 could talk. Because other treatments were given at the same time and there was no control group, it is impossible to know how much of the improvement came from CBD; the authors also stress that this is an open-label case series. Side effects: 1 child died of multiple organ failure related to isoflurane anesthesia (not attributed to CBD); other side effects not in the abstract | PMID 27655472, doi:10.1177/0883073816669450 |
| Kühne 2023 (Germany, 16 centers) Retrospective, multicenter real-world study (Germany, 16 epilepsy centers, December 2015–December 2021); no control group | 311 (235 children, 76 adults); 206 outside LGS, Dravet and TSC Children and adults with epilepsy from various causes (structural, genetic, unknown; including syndromes such as Angelman and Rett); 74.9% drug-resistant. 91.3% of treatment was off-label (epilepsy type, age, use without clobazam or above the approved dose) | Purified CBD: approved oral solution (70.7%) or a standard 100 mg/mL CBD solution prepared in pharmacies in Germany (NRF 22.10; 29.3%) | Median starting dose 3.2 mg/kg/day; median final dose 17.8 mg/kg/day (children 19.0, adults 13.3; at most 45 mg/kg/day) Slower than recommended: median increase of 2.8 mg/kg per week | Median follow-up 15.8 months (3 days–5.9 years) | At the last follow-up, 6.4% of patients were seizure-free, 30.5% had more than 50% seizure reduction, 31.2% had a smaller reduction, 23.2% had no change and 8.7% had an increase. The authors give seizure freedom or more than 50% reduction as 40.8% in LGS/Dravet/TSC and 36.9% in other epilepsies and do not find the difference significant; however, 36.9% also appears in the article as the rate for the whole group (115 of 311 patients): if 40.8% of the 105 patients with approved diagnoses is taken, the rate in the other 206 patients comes out at about 35% (the detailed supplementary table could not be read). Use together with clobazam did not improve the result. In 21.5% of those who benefited in the first 3 months, the effect later decreased. No control group; retrospective, and seizure outcomes rely on doctors' notes, seizure diaries and family reports. GW Pharma paid the salary of the doctor who collected data in the study; 10 of the 32 authors reported consulting, speaker fees or institutional support from GW Side effects: Side effects 46.9%: sleepiness and reduced attention 24.1%, digestive complaints 25.1% (diarrhea in 38 patients), psychiatric symptoms 6.1%, raised liver enzymes 1.6% (4 of the 5 patients were taking valproate). Stopped because of side effects 9.3%, stopped because of insufficient effect 19.6% | PMID 36693811, doi:10.1002/epi4.12699, Later off-label use study from some of the same centers (some patients may be shared): Hollander 2025, PMID 40696489, 6 centers in Germany, 2019–2023, the Frankfurt, Freiburg, Greifswald and Lingen centers and 5 of the 10 authors are also in Kühne 2023, LGS, Dravet and TSC patients were not included): 108 patients (74 focal onset, 14 genetic generalized, 20 other epilepsies, 76 adults), at month 3, seizures fell by at least half in 42 of the 108 patients who started (38.9%), retention 61% at month 12, side effects 38%, discontinuation in which side effects played a role 19.5%, seizure counts rely on diaries kept by caregivers, no control group, the study was not funded, and 9 of the 10 authors reported fees from pharmaceutical companies, including Jazz, the maker of the approved medicine |
| Calonge 2024 (Paris, adults) Retrospective, single center (Paris, Pitié-Salpêtrière; 2018–2024); approved-indication and off-label groups compared; no control group without CBD | 91 (32 approved indication: 21 LGS, 8 Dravet, 3 TSC; 59 off-label) Adults with drug-resistant epilepsy (mean age about 30). In the off-label group, genetic or chromosomal (24), brain malformation (7), other causes such as lack of oxygen at birth, post-encephalitis and FIRES (12), and unknown (16) causes | Approved purified CBD (oral solution) | Highest dose median 11.1 mg/kg/day (off-label) and 10.7 mg/kg/day (approved) not reported | At least 1 year; median follow-up 24 months | More than 50% seizure reduction 35.6% (21/59) in the off-label group and 31.3% (10/32) in the approved group; the difference was not significant (p = 0.85). Retention at 1 year 74.6% and 75.0%. In the off-label group, clobazam was used more often among responders (71.4% vs. 28.9%; p = 0.002). No control group; response relies on categories in doctors' notes. Jazz supported data collection; 4 of the 8 authors reported fees from Jazz Side effects: In the off-label and approved groups: any side effect 40.7% and 62.5%; sleepiness 22.0% and 50.0% (p = 0.01); increased seizures 8.5% and 6.3%; abnormal liver tests 1.7% and 6.3%. Of the 23 patients who stopped in the first year, 13 stopped due to insufficient effect, 4 due to increased seizures and 6 due to tolerability problems | PMID 39501537, doi:10.1002/brb3.70122 |
| Espinosa-Jovel 2023 (Colombia, other than LGS, Dravet and TSC) Retrospective, multicenter (Bogotá, Colombia); no control group | 78 Drug-resistant epilepsies other than LGS, Dravet and TSC (mostly structural and genetic causes), over age 2; median age 24, median 3 medicines, median 30 seizures per month at baseline | Purified CBD (product name not in the abstract) | not reported (in the abstract) not reported (in the abstract) | Median treatment duration 14 months (IQR 10–17) | At the last visit, mean seizure reduction 67.8%; at least 50% reduction in 68.8%, 11.5% seizure-free. Use together with clobazam did not change efficacy or safety. Retention 78.2%. The abstract does not say whom the analysis covered at the last visit or how early dropouts were counted. No control group; retrospective. The authors reported no conflicts of interest Side effects: CBD-related side effects 28.2% (22 patients); types not in the abstract | PMID 37769547, doi:10.1016/j.seizure.2023.09.015 |
| Zilmer 2021 (Denmark) Retrospective cohort, single center (Filadelfia Epilepsy Hospital, Denmark; December 2016–December 2019); no control group | 78 (seizure effect in 77; numerical analysis in the 51 patients who kept a seizure diary); 15 Dravet/LGS, 62 other epilepsies Children and young adults over age 2 with various epilepsies and at least one severe seizure per week (most with motor seizures) | Pharmacy-prepared (compounded) CBD oil, 300 mg/mL; derived from cannabis, contains no THC | Starting at 5 mg/kg/day, at most 20 mg/kg/day Increase of 5 mg/kg/day every 3–5 days (faster schedule after 2018) | Median follow-up 15.2 months (3–36) | In the 51 patients who kept a seizure diary (15 had no pre-treatment measurement, and the first month of treatment was used as the baseline), at least 50% reduction was 31.4% at month 3, 28.1% at month 12 and 20.0% at month 24. At month 3, 70% in Dravet/LGS and 22% in other epilepsies (p < 0.05; subgroup numbers not in the text); according to the authors, the rate in other epilepsies is not higher than some of the placebo groups in the approved medicine's studies. Response tended to be higher in those using it together with clobazam (40% vs. 25.8% at month 3; not significant); the clobazam dose was reduced in 67.9% of those taking clobazam. In the 26 patients without a diary, families reported benefit at a rate of 53.8% at month 3 and 11.8% at month 12. No control group; no funding, and the authors reported no conflicts of interest Side effects: Most often tiredness (especially during dose increases); increased seizures more common at 10 mg/kg/day and above; treatment stopped in 2 patients (2.6%) because of status epilepticus attributed to CBD. Discontinuation 38.8% in year 1, 66.7% in year 2 | PMID 33913416, Dan Med J 2021, 68(5):A07200527 |
| Cerulli Irelli 2026 (15q duplication and Angelman syndrome) Retrospective, multicenter (9 centers in Europe); no control group | 22 (12 15q duplication syndrome, 10 Angelman syndrome) Drug-resistant epilepsy due to 15q11.2-q13.1 duplication or deletion; median age 14.5 at the start of CBD; all patients with 15q duplication and 2 patients with Angelman had a Lennox–Gastaut-like picture | Epidyolex (purified CBD) | Initial maintenance dose median 10 mg/kg/day (IQR 10–14); 14.5 mg/kg/day (11–18) at the last check | Median 21 months of follow-up (IQR 14–33); those with at least 6 months of follow-up were included | At the last check, mean seizure reduction 55.7% (95% CI 38.7–72.7); in 14 of 22 patients (63.6%), at least 50% reduction, more than 75% reduction in 9, and 4 (18.2%) seizure-free. There was no difference between Angelman and 15q duplication (p > 0.2). Tonic seizures fell by a median of 60% and myoclonic seizures by 70%; there was no reduction in spasms (median 0%). There was no relationship between dose and response. In the doctors' assessment, 16/22 improved, and 2 Angelman patients got worse (behavior; in one, myoclonic seizures became more severe). Retrospective, small and uncontrolled; it shares the same ethics approval as the same team's monogenic epilepsy study (Cerulli Irelli 2025). No funding for the study; 5 of the 15 authors report receiving fees or support from the manufacturer (Jazz) unrelated to the study. Side effects: Side effects in 6 patients (27.3%): sleepiness/sedation 5, digestive complaints 3, raised liver enzymes 1, skin finding 1; still present in 2 at the last check. 4 patients stopped due to lack of effect (retention 18/22) | PMID 41992447, doi:10.1002/epi4.70241 |
| Vignoli 2025 (typical Rett syndrome and CDKL5 deficiency) Retrospective, multicenter (Italian national study group); no control group | 27 (14 Rett/MECP2, 13 CDKL5) Genetically confirmed typical Rett syndrome or CDKL5 deficiency with drug-resistant seizures, median age 10.5 (26 girls) | Epidiolex (purified CBD) | From 5 to 20 mg/kg/day; median 15 mg/kg/day at the last check (IQR 11.1–18.8) From 5 to 20 mg/kg/day according to tolerability and response | Median 14 months (IQR 8.5–20); those who took CBD for at least 6 months were included | Seizures fell by at least half in 18 of 27 patients (66.6%; 95% CI 46.0–83.5) and by more than 75% in 7 (25.9%); they did not change in 9. Rett 9/14, CDKL5 9/13 (no difference, p = 1.0). No increase in seizures was seen. 12 families (44.4%) reported improvement in attention and responsiveness. The authors write in the abstract that the effect did not reach statistical significance (the text has no corresponding test for a single-arm study). Because the requirement of having taken CBD for at least 6 months leaves out those who stopped early, the rates may look high; in a study partly from the same centers with a requirement of at least 3 months of CBD, the response in patients with CDKL5 and MECP2 was much lower (other source). Retrospective and uncontrolled. Supported by government research funding; 2 of the 23 authors report consulting or advisory board ties with the manufacturer (GW). Side effects: Sleepiness 3, restlessness/agitation 2, loss of appetite 2, insomnia 1 patient; 2 patients stopped because of side effects | PMID 40543048, doi:10.1002/epi4.70078, Monogenic epilepsy study partly from the same centers (6 shared centers and authors, periods overlap): Cerulli Irelli 2025, PMID 40126049, 27 centers, at least 3 months of CBD): in the CDKL5 (12 patients) and MECP2 (12 patients) groups, the proportion of patients whose seizures fell by at least half was 33.3%, some of the patients are very likely shared with those in this row. The same study is a separate row in the Dravet and TSC tables |
| Caraballo 2023 (epilepsy with myoclonic-atonic seizures) Retrospective, multicenter (first author's center Buenos Aires, Garrahan Hospital; other centers not in the abstract); no control group | 26 (22 epilepsy with myoclonic-atonic seizures, 4 Sturge-Weber syndrome with myoclonic-atonic seizures) Children with drug-resistant epilepsy with myoclonic-atonic seizures (Doose syndrome) | CBD; product type (purified CBD or CBD-rich cannabis oil) not stated in the abstract | 8–40 mg/kg/day not reported | Mean 19 months of follow-up | In 15 of 26 patients (57.7%), seizures fell by more than half, and 3 (11.5%) became seizure-free; a 25–50% reduction was reported in the remaining 11 (42.3%). Drop seizures, atypical absences and nonconvulsive status epilepticus responded well. In 3 of the patients with Sturge-Weber, focal seizures fell by 30%. Retrospective and uncontrolled; the authors report no company ties. Side effects: Mild; CBD was not stopped in any patient | PMID 37182500, doi:10.1016/j.yebeh.2023.109245 |
| González-Alguacil 2026 (epileptic spasms over age 2) Retrospective, single-center (Madrid, Niño Jesús Hospital, 2020–2024); no control group | 53 Children older than 2 years with drug-resistant epilepsy whose main seizure type is epileptic spasms | Epidyolex (purified CBD) | Not reported (in the abstract) not reported | Not reported (in the abstract) | In 58.5% of 53 patients, spasms fell by at least half; according to the abstract, spasms stopped completely in 15% of these. The highest response was seen in those with malformations of cortical development and Down syndrome. 77.3% of patients were also taking clobazam. CBD was stopped in 17% due to lack of effect. Retrospective, single-center and uncontrolled. All 6 of the 6 authors report receiving research or educational support from pharmaceutical companies, including the manufacturer (Jazz). Side effects: Side effects 62.2%: most often sleepiness, high blood ammonia and behavior disorders; 11.3% stopped because of side effects | PMID 41197417, doi:10.1016/j.pediatrneurol.2025.10.013 |
| Devinsky 2022 (medical cannabis containing THC) Prospective observational study (structured follow-up in a residential care center, New York); no control group and no blinding | 29 (15 focal or multifocal, 3 generalized, 11 LGS) People aged 12–46 with drug-resistant epilepsy that began in childhood (an average of 10 medicines tried) | Medical cannabis tablets, THC:CBD 1:20 and/or 1:50 (Columbia Care) | At most 6 mg THC per day; at the highest dose, a mean of 5.9 mg THC and 117.2 mg CBD per day (0.11 and 2.3 mg/kg/day) THC increased by 2 mg per week (1:20 tablet); if the response was inadequate, patients switched to the 1:50 product | At least 24 weeks (at most 9 months) after a 90-day baseline period | At the stable highest dose, convulsive seizure frequency did not change significantly from baseline (median −5.3%; p = 0.99); more than 50% reduction in 6 patients (21%) (4 of whom had no convulsive seizures during this period), at least 50% increase in 4 patients (13.8%); the authors write that the change in individual patients could not be told apart from spontaneous fluctuation. No difference between the focal/multifocal, generalized and LGS groups in seizure frequency, seizure duration or post-seizure time; a trend toward shorter post-seizure time in the focal and generalized epilepsy groups (trend only). Rescue medication, behavior and sleep did not change. The CBD dose was much lower than in the approved medicine's studies. The company that makes the product (Columbia Care) partly funded the study; the researchers did not receive payment from the company. Side effects: 10 side effects in 6 patients (vomiting, gum or nose bleeding, respiratory infection, tiredness, dizziness and 1 status epilepticus), all temporary and most unrelated to the medicine; 16 serious side effects in 7 patients (5 of them pneumonia), all judged unrelated to the medicine; 2 patients were withdrawn from the study because of serious side effects (one for increased seizures, one for increased behavior problems; the authors discuss that the behavior increase may have been due to THC, and behavior returned to its previous level when the medicine was stopped) | PMID 35267245, doi:10.1002/acn3.51537 |
| Navarro 2023 (adult focal epilepsy, compounded CBD+THC) Open-label, prospective, single-center cohort (Bogotá, Colombia); no control group | 44 (followed for at least 3 months; 38.6% of those enrolled) Adults with drug-resistant focal epilepsy | Cannabis-based compounded (pharmacy-prepared) formula: 100 mg CBD and less than 1.9 mg THC per mL, sublingual | Median 200 mg CBD and 4 mg THC per day (CBD 3.7 mg/kg/day) Started at 0.1 mL twice a day, with weekly increases | August 2020–July 2022; primary outcome at week 12 | At week 12, more than 50% seizure reduction in 79.5% of patients; median monthly seizures fell from 11 to 2.5 (p < 0.001; median reduction 84.1%). Only 38.6% of those enrolled could be followed for 3 months, and the analysis covers only them; no control group; single-author study, and the author also works in a medical cannabis and drug-resistant epilepsy program. Side effects: Drug-related adverse effects reported in 5 patients (type not given in the abstract) | PMID 36129615, doi:10.1007/s10072-022-06393-1 |
| Uliel-Sibony 2021 (Israel, 20:1 oil, tolerance) Prospective, single-center follow-up of consecutive patients (Tel Aviv Sourasky, Dana-Dwek Children's Hospital); no control group | 92 (tolerance analysis 84) Children and adults aged 1–37 with treatment-resistant epilepsy (mean age 11.8) | Cannabis oil extract, CBD:THC 20:1 | Mean CBD 11.3 mg/kg/day (4–38) not reported (in the abstract) | Mean 19.8 ± 12.5 months (3–45 months) | More than 50% seizure reduction 54%, seizure freedom 9% (8 patients); seizure counts reported by patients or families at monthly visits. Tolerance developed in 25% of patients (21 of 84) after a mean of 7.3 months (the dose had to be raised by at least 30% for the same effect, or seizures increased by at least 30%); tolerance decreased as epilepsy duration increased (p = 0.038). 31% stopped due to lack of effect or side effects. No control group. Side effects: Side effects in 47 of 87 patients (51%); one of the reasons for stopping | PMID 32713661, doi:10.1016/j.braindev.2020.06.018, Earlier study from the same center: Hausman-Kedem 2018, PMID 29674131, 20:1 oil, 57 patients, 46 in the efficacy analysis, mean CBD 11.4 mg/kg/day, median follow-up 18 months, side effects 46% and the main reason for stopping, better response in children under 10 years of age and with doses above 11 mg/kg/day, the response rate sentence in the abstract ('26 patients (56%) ≤50% reduction') is unclear). Retrospective study of 5 Israeli pediatric neurology clinics, including this center: Tzadok 2016, PMID 26800377, 74 children, 20:1 CBD:THC oil, CBD 1–20 mg/kg/day, mean 6 months, according to family reports, more than 50% seizure reduction in 38 children (52%), stopped in 5 children (7%) due to increased seizures, one of the authors works at a cannabis company and one at the Ministry of Health's medical cannabis unit). Some of the patients in the three studies are very likely shared |
| Lusawat 2025 (Thailand, CBD-rich oil) Prospective observational study, 19 public hospitals (Thailand); no control group | 101 Children with drug-resistant epilepsy from various causes (median age 10; an average of 7 medicines tried; median 75 seizures per month) | Medical-grade CBD-rich oil made in Thailand (THC ratio not in the abstract) | Median CBD 6 mg/kg/day; effective dose in the range of 1–15 mg/kg/day not reported (in the abstract) | Median 15 months of follow-up (2021–2023) | The rate of more than 50% seizure reduction and the median reduction in monthly seizure count pointed toward improvement at months 3, 6, 9 and 12 and at the last visit (rates not in the abstract); seizure types other than complex motor seizures responded. 33 patients stopped (57% due to side effects, 30% due to lack of effect, 12% due to non-adherence). No control group; because response rates are not given in the abstract, the size of the effect is unknown. Side effects: Side effects in 92% of patients, most (95%) mild: sleepiness, raised liver enzymes, loss of appetite, restlessness | PMID 40460512, doi:10.1016/j.pediatrneurol.2025.04.015 |
| Lyu 2026 (Minnesota medical cannabis program) Dispensing records and per-visit patient reports from a state medical cannabis program (retrospective registry analysis); no control group | 1,940 (1,284 certified for seizures only) People in Minnesota certified for medical cannabis because of seizures (68.4% adults) | Cannabis products dispensed in the state program; 86.8% of patients were given products containing both CBD and THC | Median CBD 2.5 mg/kg/day (about 75% lower than the approved medicine's maintenance dose); in regular users, CBD 0.001–33.1 and THC 0.002–12.6 mg/kg/day; 87% high CBD:THC ratio products not reported (in the abstract) | July 2015–December 2023 | According to patient reports, seizure frequency improved during follow-up in 67.6%; at the last visit, 76.2% were also using at least one epilepsy medicine. Doses varied widely; people aged 60 and over reported side effects more often. Patients found the product helpful but expensive and mostly chose products containing THC. Reports are self-reported; no control group. Side effects: No rate in the abstract; side effects more common at age 60 and over | PMID 42612572, doi:10.1016/j.yebeh.2026.111230 |
| Ríos-Pohl 2024 (Chile, product analysis) Product analysis (high-performance liquid chromatography of samples sent in voluntarily by families) | 35 samples (27 homemade, 6 unlicensed commercial, 2 products equivalent to the approved medicine) CBD oils given to children with epilepsy in Chile to treat seizures | Homemade oils, unlicensed commercial CBD products and bioequivalent CBD products | March 2020–September 2021 | The CBD in the two products bioequivalent to the approved medicine was close to the labeled 100 mg/mL (88.3 and 90 mg/mL), and THC was below the limit (0 and 0.03 mg/mL). In unlicensed commercial products, CBD ranged from 0.1 to 78.5 mg/mL (mean 25.4); 4 of 6 products were far below the labeled amount (25–100 mg/mL on the label, 0.1–3.3 mg/mL measured); of the two samples with high CBD (two batches of the same brand: 67 and 78.5 mg/mL), one exceeded the THC limit (2.43 mg/mL). In homemade oils, CBD was 0–6.6 mg/mL, far below treatment levels; THC ranged from 0 to 388 mg/mL and was above the 0.2% (2 mg/mL) limit in 11 of 27 samples (in two, 6.3 and 388 mg/mL, at a level the authors wrote carries a risk of poisoning); the solvent was unknown in two thirds of the homemade samples. The samples came from families who responded to a call on social media; patients' diagnoses, seizures and side effects were not asked about. | PMID 39503571, doi:10.1002/epi4.13081 | |
| Gidal 2024 (US commercial CBD products) Product analysis (label accuracy and contaminant screening; accredited laboratory, one sample per product) | 202 products (100 drops/tinctures, 48 gummies, 34 vaping products, 20 topical) CBD products sold over the counter in the US (not FDA-approved); bought from brand websites and online marketplaces | Unlicensed commercial CBD products (full-spectrum, broad-spectrum, isolate or type not stated) | Products bought October–December 2021 | In 74% of products, the amount of CBD deviated from the label by at least 10%: 46% contained more than the label and 28% less; one product contained 565% of the labeled CBD; no CBD was detected in 5 products. The share of products that did not match the product type stated on the label (full-spectrum, broad-spectrum, isolate) was 26% in the abstract and 32–43% by type in the findings; 3 broad-spectrum products that should have been THC-free contained THC, 2 of them above 0.3%. Heavy metals were found in 44 products (most often lead; 5 products exceeded the limit), solvent residues in 181 products and pesticides in 30 products; there were 53 violations of regulatory limits in total. The study was funded by the maker of the approved CBD medicine (Jazz), which took part in the design, analysis and writing; all 7 of the 7 authors had ties to the company (2 employees, 1 former employee, 4 consultants or company-funded researchers). | PMID 38562466, doi:10.3389/fphar.2024.1335441 | |
| Calapai 2024 (unlicensed CBD, European side effect reports) Pharmacovigilance analysis: spontaneous side effect reports in the European Medicines Agency's EudraVigilance database | not reported (in the abstract) People in Europe with a side effect reported after using an unlicensed CBD product | Unlicensed CBD products (not approved as medicines) | not reported (in the abstract) | 18.9% of side effect reports involving unlicensed CBD were serious; serious reports were more common in males and adults, and less common in children aged 3–11. In 38.8% of reports, the product had been used to treat epilepsy. The most common serious side effects were psychiatric disorders, liver disorders and worsening of existing epilepsy; the medicines most often seen together with the side effects were clobazam and valproic acid, followed by marijuana. Based on spontaneous reports; causality and frequency cannot be calculated. | PMID 37800192, doi:10.1002/ptr.8028 |
Reported side effects
- Devinsky 2016: side effects 79% (sleepiness 25%, decreased appetite 19%, diarrhea 19%, tiredness 13%, convulsion 11%); serious side effects 30%, status epilepticus 6%
- Szaflarski 2018: diarrhea 29%, sleepiness 22%
- Unregulated oils: increased seizures 13% (Press 2015) and early discontinuation 14% (Knupp 2019)
- 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)
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 restarts when the dose goes above 10 mg/kg/day or when a medicine affecting the liver is changed
- People taking valproate and those with high baseline ALT/AST: per 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 stay >5×ULN
- If liver symptoms occur (unexplained nausea, vomiting, upper right abdominal pain, tiredness, loss of appetite, jaundice, dark urine), check ALT/AST/bilirubin right away
- Blood levels and side effects of antiseizure medicines used at the same time (Gaston 2017)
Who should be careful
- Epilepsy treatment is started, changed and stopped by the neurologist's decision; antiseizure medicines are not stopped on one's own
- CBD is stopped by gradually reducing the dose; stopping suddenly carries a risk of increased seizures and status epilepticus (FDA §2.5, §5.5). EMA: in the studies, CBD was stopped over ~10 days, reducing by ~10% per day
- In moderate (Child-Pugh B) and severe (Child-Pugh C) liver impairment, the label specifies lower starting and maintenance doses; no adjustment is needed in mild impairment
- Hypersensitivity to sesame oil is a contraindication (Epidyolex contains 736 mg refined sesame oil and 79 mg ethanol per ml)
- People taking 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. Doses are measured with a calibrated syringe; a household spoon is not suitable
- Unregulated (artisanal) CBD/cannabis oils do not replace the approved medicine: labels are unreliable, they may contain THC and contaminants, and increased seizures have been reported
- The approval covers only Dravet syndrome, LGS and TSC; use in other epilepsies is the doctor's decision
- In unregulated products, CBD may be lower than the label (ineffective dose) or higher (unexpected side effects and clobazam/valproate interaction), and they may contain THC
Drug interactions
- Clobazam: CBD increases the active metabolite N-desmethylclobazam ~3-fold (FDA) / 3–4-fold (EMA); clobazam also increases 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 up 28%, AUC 55%; in patients, Cmax up 17%, AUC 30%. Side effects are monitored closely
- Everolimus: exposure increases ~2.5-fold; levels are monitored, and in patients taking 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 increase; 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
- The approved CBD medicine is approved only for seizures in Dravet syndrome, Lennox–Gastaut syndrome and tuberous sclerosis; using it for another type of epilepsy is the neurologist's decision.
- All epilepsy medicines being used, especially clobazam and valproate, should be reviewed for interactions, and liver tests should be planned.
- If you are considering a product sold on the market, review its label, THC content and independent laboratory certificate of analysis with your doctor.
- Do not stop epilepsy medicines or CBD suddenly; if you notice an increase in seizures, tell your doctor right away.
Choose your age group, medicines and situation to prepare a short summary you can take to your doctor. Nothing you enter is sent anywhere.
Bottom line
Early findings on CBD in drug-resistant epilepsies are striking, but controlled studies are few, and the two largest in focal epilepsy were negative. Comparative studies are needed for a firm conclusion; the effect of oils sold on the market has not been shown. Taking every step together with your neurologist is the safest way.
Frequently asked questions
Does CBD oil work for other drug-resistant epilepsies?
Fewer seizures have been reported with purified CBD medicine, but most of these studies had no comparison group, so the evidence is weak. In two controlled studies in adults with focal epilepsy (CBD gel applied to the skin in 188 people, and CBDV, a compound closely related to CBD, in 162 people), results were no different from placebo. In real-world studies following off-label use, seizures fell in some patients, but these studies had no comparison group. In a prospective study with oils sold on the market that also recorded seizures before treatment, the results were at the level seen in people taking an inactive medicine (placebo) in drug trials. There is no approved CBD medicine for these epilepsies.
Can CBD oil be used in children?
Most studies in these epilepsies were done in children, and none were comparative. Increased seizures have been reported in some children using unregulated oils. If you are considering any CBD product for your child, it is important to make the decision together with a pediatric neurologist.
Why aren't the positive results reported by families considered enough?
Without a comparison group, it is impossible to tell whether an improvement came from CBD, from the passage of time or from expectations. In one study, families who had moved to another state for this treatment reported improvement much more often than families who already lived there; this difference shows the role of expectations. In well-designed studies, seizures can also fall in people taking an inactive medicine (placebo): in a placebo-controlled study of adults with focal seizures, seizures fell by about 38% in the placebo group. In a study without a comparison group, this improvement would have been credited to the treatment.
What problems can CBD oils sold on the market have?
The actual amount of CBD may be lower or higher than the label says; this can lead to no effect or to unexpected side effects and drug interactions. Some products may contain THC, solvents, heavy metals or pesticide residues. For example, in the US, the CBD content of 74% of 202 commercial products differed from the label by at least 10%; in Australia, THC or the raw form of THC (THCA) was found in almost all of the extracts used for children with epilepsy.
Can CBD oil be obtained with a prescription for drug-resistant epilepsy 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 classified as medicines and health products 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 (Turkish Medicines and Medical Devices Agency), and importing it from abroad requires the agency's written approval. If you are thinking of using it, talk to your doctor first.
References
Show 40 references
- O'Brien 2022 (STAR 1). Multicenter, double-blind, placebo-controlled RCT, 3 arms. PMID 35802375, doi:10.1001/jamanetworkopen.2022.20189, ACTRN12616000510448
- Ebadi 2023. Triple-blind, controlled study (Iran). PMID 37470002, doi:10.3389/fneur.2023.1143783, IRCT20210608051515N1
- Devinsky 2016. Open-label, uncontrolled; 11 US centers. PMID 26724101, doi:10.1016/S1474-4422(15)00379-8, Single-center reports from centers that took part in the same program (some patients shared): Sands 2019, PMID 30460546, UCSF, 26 children under individual patient and expanded access permits between April 2013 and December 2014, mean 21 months): at month 24, 9 of the 26 children were still on treatment, and 7 (26.9%) had a lasting reduction of more than half in motor seizures, 57.7% stopped due to insufficient effect, side effects 80.8%, serious side effects 23.1% (3 status epilepticus, 2 catatonia). Patel 2021, PMID 34050682, MGH, 54 patients, median 45.5 months): responder rate similar at year 1 and at the last visit (41.7–42.6%), no control group
- Szaflarski 2018. Interim analysis of an open-label expanded access program, 25 US centers. PMID 29998598, doi:10.1111/epi.14477, Final pooled analysis of the same program: Szaflarski 2023, PMID 36537757, 35 US centers, 892 patients, up to January 2019, median 694 days): 36% stopped (insufficient effect 19%, side effects 7%), median reduction in convulsive seizures 50–67%, at least 50% reduction 51–59%, side effects 88%, serious side effects 41%, 20 deaths judged by investigators as unrelated to treatment, no control group. Funding also includes GW/Jazz and Alabama's Carly's Law fund, so the UAB patients in Gaston 2021 may be included in this pool, PMID 38420790, PMID 37243404, absence and myoclonic seizures and epileptic spasms, at least 50% reduction in about half of patients, PMID 38572913
- Patel 2025 (expanded access, focal epilepsies). Subanalysis of an open-label expanded access program (the same program as Devinsky 2016 and Szaflarski 2018; no control group). PMID 40673944, doi:10.1111/epi.18496
- Devinsky 2018 (CDKL5, Aicardi, Dup15q, Doose). Open-label, compassionate use/EAP. PMID 30006259, doi:10.1016/j.yebeh.2018.05.013
- Gaston 2021. Open-label (UAB EAP), 2 years. PMID 33667843, doi:10.1016/j.yebeh.2021.107862, Drug interaction analyses from the same UAB program: Gaston 2017, PMID 28782097, 39 adults, 42 children): as the CBD dose increased, topiramate, rufinamide and N-desmethylclobazam levels rose and the clobazam level fell, in adults, zonisamide and eslicarbazepine levels also rose, changes other than clobazam and its metabolite stayed within the therapeutic range, AST/ALT were higher in those taking valproate. Final analysis Gaston 2023, PMID 37593907, 169 people, 80 adults): with CBD dose, levels of clobazam, N-desmethylclobazam, free valproate, felbamate and topiramate rose (levetiracetam in children only, perampanel in adults only), according to the authors, the clinical importance is probably small. The program was funded by the state of Alabama (Carly's Law), and the medicine was donated by GW. This UAB program is the Alabama arm of the US expanded access program in row 1768 (Carly's Law funding), some patients are very likely also counted in that row
- Knupp 2019. Prospective observation. PMID 31550641, doi:10.1016/j.seizure.2019.09.007
- Press 2015. Retrospective parent report (Colorado). PMID 25845492, doi:10.1016/j.yebeh.2015.02.043, Overlapping cohort from the same center: Treat 2017, PMID 27859038, moving to Colorado was associated with perceived benefit (65% vs. 38%, p = 0.01) but was not independently associated with longer use, a Dravet diagnosis was associated with shorter use (p = 0.02), according to parent reports, response 24%, side effects 19% (most often sleepiness and increased seizures)
- Porcari 2018. Retrospective. PMID 29429908, doi:10.1016/j.yebeh.2018.01.026
- Sulak 2017. Retrospective patient data (Washington and California) and 4 case reports; no control group. PMID 28254350, doi:10.1016/j.yebeh.2016.12.032
- Suraev 2022 (PELICAN subanalysis). Product analysis (contaminant screening). PMID 34954507, doi:10.1016/j.yebeh.2021.108496, Epilepsy Behav 2022, 127:108496 (online December 2021). Earlier study in which the samples were collected (same research group, Suraev 2022 describes its samples as collected in an earlier study, 78 samples in 2022, 61 in 2018): Suraev 2018, PMID 29977078, PMID 30068968, 61 samples from 38 families (51 extracts when combined) were analyzed. Families believed CBD was the main component in 71% of the extracts, but 98% of the samples contained THC and/or its raw form THCA, CBD was the dominant cannabinoid in only 23%, was absent in 4 samples, and only 3 extracts (6%) came close to the lowest CBD dose in studies (10 mg/kg/day). Families rated 75% of the extracts as effective, the cannabinoid content of extracts rated effective and ineffective did not differ. No control group, results rely on family reports
- GWPCARE7 pilot + extension (infantile spasms). Open-label pilot phase and open-label extension (uncontrolled); results from the ClinicalTrials.gov record. NCT02953548, NCT02954887
- Cunha 1980. Double-blind, placebo-controlled randomized study (phase 2; phase 1 in healthy volunteers). PMID 7413719, doi:10.1159/000137430, A review that also describes the same trial: Carlini and Cunha 1981, PMID 7028792, improvement in 7 of the 8 patients taking CBD and in 1 patient taking placebo)
- Brodie 2021 (CBDV, focal seizures). Phase 2, multicenter, double-blind, placebo-controlled randomized study (centers in Czechia, Hungary, Poland, Spain and the UK). PMID 33998885, doi:10.1089/can.2020.0075, NCT02365610, no results in the registry or PubMed). The journal added a correction note for Figure 2 (April 15, 2021)
- Gloss 2014 (Cochrane). Cochrane systematic review (randomized studies only). PMID 24595491, doi:10.1002/14651858.CD009270.pub3, PMID 22696383, PMID 7413719, a separate row in the table), PMID 3941934, letter to the editor), Trembly 1990 (conference abstract only)
- Lattanzi 2021 (epilepsies other than Dravet and LGS). Systematic review (no meta-analysis); 42 studies: 3 RCTs (one an efficacy study in TSC, GWPCARE6, two drug interaction studies), open-label studies and their subanalyses, retrospective chart reviews, clinical series and case reports. PMID 33754312, doi:10.1007/s40263-021-00807-y
- Coppola 2026 (DEE and complex drug-resistant epilepsies). Systematic review (narrative synthesis; no meta-analysis). PMID 41558068, doi:10.1016/j.eplepsyres.2026.107731, PMID 41820159, content not on PubMed, could not be read). Non-seizure outcomes from the same systematic search: Samanta 2026, PMID 42520581, 32 studies, 1, 343 patients, most studies reporting improvements in cognitive function, behavior, communication and quality of life were observational and at moderate-to-high risk of bias, same authors and same company ties)
- Pamplona 2018 (CBD-rich extracts and pure CBD). Meta-analysis of observational studies (11 studies; categorical comparison with Fisher's test). PMID 30258398, doi:10.3389/fneur.2018.00759, Correction: Front Neurol 2018, PMID 30687209, PMC6335388, numbers in Tables 2 and 4 and the abstract were corrected, according to the authors, the conclusion did not change)
- NICE 2019 (NG144 guideline). National guideline (based on an evidence review). PMID 31909928, Evidence review D: epilepsy (NICE, November 2019, PMID 35107910, NBK577085). Guideline: nice.org.uk/guidance/ng144
- Absence epilepsy, synthetic CBD (NCT03336242). Phase 2, open-label, non-randomized dose-finding study (two dose groups); results from the ClinicalTrials.gov record, no peer-reviewed publication. NCT03336242, 2023), NCT03355300, results June 18, 2023). No peer-reviewed publication of this study was found in PubMed (October 28, 2026)
- Scheffer 2021 (CBD gel on the skin, developmental and epileptic encephalopathy). Open-label, two-center, non-randomized before-and-after study (Melbourne and Wellington); no control group. PMID 34477852, doi:10.1001/jamanetworkopen.2021.23930, ACTRN12618000516280
- Kaplan 2017 (Sturge-Weber syndrome). Open-label expanded access study (under an FDA investigational permit, single center, Kennedy Krieger); no control group; 8-week baseline observation. PMID 28454984, doi:10.1016/j.pediatrneurol.2017.02.009, NCT02332655, collaborators Jazz Pharmaceuticals and the Faneca 66 Foundation). Later study by the same team: Smegal 2023, PMID 36508880, NCT04447846, no seizures were seen, and improvements in neurological score, quality of life and anxiety were reported, on the working memory test, the registry's primary outcome, the mean T score changed from 37.1 to 39.1 in 7 patients (no statistics given), 1 patient stopped because of a moderate side effect, open-label and uncontrolled
- Gofshteyn 2017 (FIRES). Multicenter case series; open-label use under emergency or expanded access protocols (5 centers); no control group. PMID 27655472, doi:10.1177/0883073816669450
- Kühne 2023 (Germany, 16 centers). Retrospective, multicenter real-world study (Germany, 16 epilepsy centers, December 2015–December 2021); no control group. PMID 36693811, doi:10.1002/epi4.12699, Later off-label use study from some of the same centers (some patients may be shared): Hollander 2025, PMID 40696489, 6 centers in Germany, 2019–2023, the Frankfurt, Freiburg, Greifswald and Lingen centers and 5 of the 10 authors are also in Kühne 2023, LGS, Dravet and TSC patients were not included): 108 patients (74 focal onset, 14 genetic generalized, 20 other epilepsies, 76 adults), at month 3, seizures fell by at least half in 42 of the 108 patients who started (38.9%), retention 61% at month 12, side effects 38%, discontinuation in which side effects played a role 19.5%, seizure counts rely on diaries kept by caregivers, no control group, the study was not funded, and 9 of the 10 authors reported fees from pharmaceutical companies, including Jazz, the maker of the approved medicine
- Calonge 2024 (Paris, adults). Retrospective, single center (Paris, Pitié-Salpêtrière; 2018–2024); approved-indication and off-label groups compared; no control group without CBD. PMID 39501537, doi:10.1002/brb3.70122
- Espinosa-Jovel 2023 (Colombia, other than LGS, Dravet and TSC). Retrospective, multicenter (Bogotá, Colombia); no control group. PMID 37769547, doi:10.1016/j.seizure.2023.09.015
- Zilmer 2021 (Denmark). Retrospective cohort, single center (Filadelfia Epilepsy Hospital, Denmark; December 2016–December 2019); no control group. PMID 33913416, Dan Med J 2021, 68(5):A07200527
- Cerulli Irelli 2026 (15q duplication and Angelman syndrome). Retrospective, multicenter (9 centers in Europe); no control group. PMID 41992447, doi:10.1002/epi4.70241
- Vignoli 2025 (typical Rett syndrome and CDKL5 deficiency). Retrospective, multicenter (Italian national study group); no control group. PMID 40543048, doi:10.1002/epi4.70078, Monogenic epilepsy study partly from the same centers (6 shared centers and authors, periods overlap): Cerulli Irelli 2025, PMID 40126049, 27 centers, at least 3 months of CBD): in the CDKL5 (12 patients) and MECP2 (12 patients) groups, the proportion of patients whose seizures fell by at least half was 33.3%, some of the patients are very likely shared with those in this row. The same study is a separate row in the Dravet and TSC tables
- Caraballo 2023 (epilepsy with myoclonic-atonic seizures). Retrospective, multicenter (first author's center Buenos Aires, Garrahan Hospital; other centers not in the abstract); no control group. PMID 37182500, doi:10.1016/j.yebeh.2023.109245
- González-Alguacil 2026 (epileptic spasms over age 2). Retrospective, single-center (Madrid, Niño Jesús Hospital, 2020–2024); no control group. PMID 41197417, doi:10.1016/j.pediatrneurol.2025.10.013
- Devinsky 2022 (medical cannabis containing THC). Prospective observational study (structured follow-up in a residential care center, New York); no control group and no blinding. PMID 35267245, doi:10.1002/acn3.51537
- Navarro 2023 (adult focal epilepsy, compounded CBD+THC). Open-label, prospective, single-center cohort (Bogotá, Colombia); no control group. PMID 36129615, doi:10.1007/s10072-022-06393-1
- Uliel-Sibony 2021 (Israel, 20:1 oil, tolerance). Prospective, single-center follow-up of consecutive patients (Tel Aviv Sourasky, Dana-Dwek Children's Hospital); no control group. PMID 32713661, doi:10.1016/j.braindev.2020.06.018, Earlier study from the same center: Hausman-Kedem 2018, PMID 29674131, 20:1 oil, 57 patients, 46 in the efficacy analysis, mean CBD 11.4 mg/kg/day, median follow-up 18 months, side effects 46% and the main reason for stopping, better response in children under 10 years of age and with doses above 11 mg/kg/day, the response rate sentence in the abstract ('26 patients (56%) ≤50% reduction') is unclear). Retrospective study of 5 Israeli pediatric neurology clinics, including this center: Tzadok 2016, PMID 26800377, 74 children, 20:1 CBD:THC oil, CBD 1–20 mg/kg/day, mean 6 months, according to family reports, more than 50% seizure reduction in 38 children (52%), stopped in 5 children (7%) due to increased seizures, one of the authors works at a cannabis company and one at the Ministry of Health's medical cannabis unit). Some of the patients in the three studies are very likely shared
- Lusawat 2025 (Thailand, CBD-rich oil). Prospective observational study, 19 public hospitals (Thailand); no control group. PMID 40460512, doi:10.1016/j.pediatrneurol.2025.04.015
- Lyu 2026 (Minnesota medical cannabis program). Dispensing records and per-visit patient reports from a state medical cannabis program (retrospective registry analysis); no control group. PMID 42612572, doi:10.1016/j.yebeh.2026.111230
- Ríos-Pohl 2024 (Chile, product analysis). Product analysis (high-performance liquid chromatography of samples sent in voluntarily by families). PMID 39503571, doi:10.1002/epi4.13081
- Gidal 2024 (US commercial CBD products). Product analysis (label accuracy and contaminant screening; accredited laboratory, one sample per product). PMID 38562466, doi:10.3389/fphar.2024.1335441
- Calapai 2024 (unlicensed CBD, European side effect reports). Pharmacovigilance analysis: spontaneous side effect reports in the European Medicines Agency's EudraVigilance database. PMID 37800192, doi:10.1002/ptr.8028
Points that could not be verified from the full text
- For Devinsky 2016, Devinsky 2018 (CDKL5/Aicardi/Dup15q/Doose) and Gaston 2021, only the PubMed abstract was read; the values match the abstract
- No official source was found on whether the route of importing from abroad for personal treatment with a doctor's report (permission from TİTCK, the Turkish Medicines and Medical Devices Agency) has been used for Epidyolex in Turkey
- A study of switching from unregulated CBD to Epidiolex (NCT06924827; Toronto, 25 children planned) is not yet recruiting; estimated start January 2027, no results (ClinicalTrials.gov, last updated September 11, 2026)
- Ebadi 2023: the abstract says the control group took only their existing medicines, while the methods section says placebo was given; what the control group received remains unclear
- Controlled studies with unknown results (ClinicalTrials.gov, October 28, 2026): there are no results in the registry or peer-reviewed publications for a placebo-controlled study of CBD in children and adolescents in Brazil (NCT02783092; 126 patients planned, status unknown), a crossover study of the approved CBD in electrical status epilepticus in sleep (NCT04721691; completed in January 2024), and a placebo-controlled study of CBD+THC in adults (NCT03808935; terminated with 17 patients).
- Ongoing or planned placebo-controlled studies (ClinicalTrials.gov, October 28, 2026): studies by the maker of the approved CBD, a phase 3 study with oral CBD solution in developmental and epileptic encephalopathies (NCT07723976; not yet recruiting) and a phase 2/3 study with a capsule form of the same active ingredient in juvenile myoclonic epilepsy (NCT07723963; recruiting); a Canadian study that will compare pure CBD and a CBD-rich plant extract that also contains a small amount of THC with placebo (NCT07023744; not yet recruiting); individualized crossover studies in children in the Netherlands with a full-spectrum cannabis extract that also contains a small amount of THC (NCT07668856; recruiting). None had results as of this date.
- The original texts of the 1978, 1986 and 1990 trials in the Cochrane review could not be read (one is only a conference abstract); the details of these trials could not be verified from the primary source.
Trusted sources from around the world
Show 4 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 NG144: Cannabis-based medicinal products guidelineNICE (National Institute for Health and Care Excellence, UK), English, November 2019 (last updated March 2021)
For approved CBD in severe treatment-resistant epilepsy, the UK guideline refers only to the appraisals for tuberous sclerosis, Lennox–Gastaut and Dravet syndrome; it lists the effectiveness of CBD in the epilepsies, and the effect of adding THC to CBD, as questions that need research.
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.
The evidence level text now uses the same level name as the top of the page; the level and the result did not change.
The last sentence of the page's short summary was rewritten in a calmer, more guiding tone: instead of "Talk to your doctor before you start.", it now says whom you can talk with, and about what, if you are considering CBD. The conclusion and the strength of evidence did not change.
In October 2026, PubMed, Cochrane and ClinicalTrials.gov were systematically searched again (3,670 PubMed and 119 ClinicalTrials.gov records were screened; findings were reread from primary sources in an independent check). 27 studies added: a placebo-controlled study in focal seizures with CBDV, a compound closely related to CBD (Brodie 2021; seizures fell similarly to placebo), and the small 1980 CBD study (Cunha 1980); the Cochrane review (Gloss 2014), two systematic reviews specific to these epilepsies (Lattanzi 2021, Coppola 2026), a meta-analysis comparing CBD-rich extracts with pure CBD (Pamplona 2018) and the UK's national guideline (NICE 2019); childhood absence epilepsy (a study with results only in the registry), epileptic spasms over 2 years of age, CDKL5 deficiency and Rett, epilepsy with myoclonic-atonic seizures, 15q duplication and Angelman, Sturge–Weber and FIRES series, and CBD gel on the skin in developmental and epileptic encephalopathies; real-world data on off-label use from Germany, France, Denmark and Colombia; studies with THC-containing or CBD-rich products from the US (a prospective study and a state program), Israel, Colombia and Thailand; product content analyses from Chile and the US, and European side effect reports. Later publications from the same programs or studies with overlapping patients were added to the 'other sources' section of the relevant rows; one row's result was written out with numbers. A sentence describing the search itself was added to the evidence summary. No new efficacy RCT with purified oral CBD was found in these epilepsies. The conclusion and the strength of evidence did not change.
Related studies: Cunha 1980, published 1980 (PMID 7413719, doi:10.1159/000137430, A review that also describes the same trial: Carlini and Cunha 1981, PMID 7028792, improvement in 7 of the 8 patients taking CBD and in 1 patient taking placebo)); Brodie 2021 (CBDV, focal seizures), published December 2021 (PMID 33998885, doi:10.1089/can.2020.0075, NCT02365610, no results in the registry or PubMed). The journal added a correction note for Figure 2 (April 15, 2021)); Gloss 2014 (Cochrane), published March 5, 2014 (PMID 24595491, doi:10.1002/14651858.CD009270.pub3, PMID 22696383, PMID 7413719, a separate row in the table), PMID 3941934, letter to the editor), Trembly 1990 (conference abstract only)); Lattanzi 2021 (epilepsies other than Dravet and LGS), published March 2021 (PMID 33754312, doi:10.1007/s40263-021-00807-y); Coppola 2026 (DEE and complex drug-resistant epilepsies), published February 2026 (PMID 41558068, doi:10.1016/j.eplepsyres.2026.107731, PMID 41820159, content not on PubMed, could not be read). Non-seizure outcomes from the same systematic search: Samanta 2026, PMID 42520581, 32 studies, 1, 343 patients, most studies reporting improvements in cognitive function, behavior, communication and quality of life were observational and at moderate-to-high risk of bias, same authors and same company ties)); Pamplona 2018 (CBD-rich extracts and pure CBD), published September 2018 (PMID 30258398, doi:10.3389/fneur.2018.00759, Correction: Front Neurol 2018, PMID 30687209, PMC6335388, numbers in Tables 2 and 4 and the abstract were corrected, according to the authors, the conclusion did not change)); NICE 2019 (NG144 guideline), published November 11, 2019 (PMID 31909928, Evidence review D: epilepsy (NICE, November 2019, PMID 35107910, NBK577085). Guideline: nice.org.uk/guidance/ng144); Absence epilepsy, synthetic CBD (NCT03336242) (NCT03336242, 2023), NCT03355300, results June 18, 2023). No peer-reviewed publication of this study was found in PubMed (October 28, 2026)); Scheffer 2021 (CBD gel on the skin, developmental and epileptic encephalopathy), published September 2021 (PMID 34477852, doi:10.1001/jamanetworkopen.2021.23930, ACTRN12618000516280); Kaplan 2017 (Sturge-Weber syndrome), published June 2017 (PMID 28454984, doi:10.1016/j.pediatrneurol.2017.02.009, NCT02332655, collaborators Jazz Pharmaceuticals and the Faneca 66 Foundation). Later study by the same team: Smegal 2023, PMID 36508880, NCT04447846, no seizures were seen, and improvements in neurological score, quality of life and anxiety were reported, on the working memory test, the registry's primary outcome, the mean T score changed from 37.1 to 39.1 in 7 patients (no statistics given), 1 patient stopped because of a moderate side effect, open-label and uncontrolled); Gofshteyn 2017 (FIRES), published January 2017 (PMID 27655472, doi:10.1177/0883073816669450); Kühne 2023 (Germany, 16 centers), published June 2023 (PMID 36693811, doi:10.1002/epi4.12699, Later off-label use study from some of the same centers (some patients may be shared): Hollander 2025, PMID 40696489, 6 centers in Germany, 2019–2023, the Frankfurt, Freiburg, Greifswald and Lingen centers and 5 of the 10 authors are also in Kühne 2023, LGS, Dravet and TSC patients were not included): 108 patients (74 focal onset, 14 genetic generalized, 20 other epilepsies, 76 adults), at month 3, seizures fell by at least half in 42 of the 108 patients who started (38.9%), retention 61% at month 12, side effects 38%, discontinuation in which side effects played a role 19.5%, seizure counts rely on diaries kept by caregivers, no control group, the study was not funded, and 9 of the 10 authors reported fees from pharmaceutical companies, including Jazz, the maker of the approved medicine); Calonge 2024 (Paris, adults), published November 2024 (PMID 39501537, doi:10.1002/brb3.70122); Espinosa-Jovel 2023 (Colombia, other than LGS, Dravet and TSC), published November 2023 (PMID 37769547, doi:10.1016/j.seizure.2023.09.015); Zilmer 2021 (Denmark), published April 28, 2021 (PMID 33913416, Dan Med J 2021, 68(5):A07200527); Cerulli Irelli 2026 (15q duplication and Angelman syndrome), published June 2026 (PMID 41992447, doi:10.1002/epi4.70241); Vignoli 2025 (typical Rett syndrome and CDKL5 deficiency), published August 2025 (PMID 40543048, doi:10.1002/epi4.70078, Monogenic epilepsy study partly from the same centers (6 shared centers and authors, periods overlap): Cerulli Irelli 2025, PMID 40126049, 27 centers, at least 3 months of CBD): in the CDKL5 (12 patients) and MECP2 (12 patients) groups, the proportion of patients whose seizures fell by at least half was 33.3%, some of the patients are very likely shared with those in this row. The same study is a separate row in the Dravet and TSC tables); Caraballo 2023 (epilepsy with myoclonic-atonic seizures), published June 2023 (PMID 37182500, doi:10.1016/j.yebeh.2023.109245); González-Alguacil 2026 (epileptic spasms over age 2), published January 2026 (PMID 41197417, doi:10.1016/j.pediatrneurol.2025.10.013); Devinsky 2022 (medical cannabis containing THC), published April 2022 (PMID 35267245, doi:10.1002/acn3.51537); Navarro 2023 (adult focal epilepsy, compounded CBD+THC), published January 2023 (PMID 36129615, doi:10.1007/s10072-022-06393-1); Uliel-Sibony 2021 (Israel, 20:1 oil, tolerance), published January 2021 (PMID 32713661, doi:10.1016/j.braindev.2020.06.018, Earlier study from the same center: Hausman-Kedem 2018, PMID 29674131, 20:1 oil, 57 patients, 46 in the efficacy analysis, mean CBD 11.4 mg/kg/day, median follow-up 18 months, side effects 46% and the main reason for stopping, better response in children under 10 years of age and with doses above 11 mg/kg/day, the response rate sentence in the abstract ('26 patients (56%) ≤50% reduction') is unclear). Retrospective study of 5 Israeli pediatric neurology clinics, including this center: Tzadok 2016, PMID 26800377, 74 children, 20:1 CBD:THC oil, CBD 1–20 mg/kg/day, mean 6 months, according to family reports, more than 50% seizure reduction in 38 children (52%), stopped in 5 children (7%) due to increased seizures, one of the authors works at a cannabis company and one at the Ministry of Health's medical cannabis unit). Some of the patients in the three studies are very likely shared); Lusawat 2025 (Thailand, CBD-rich oil), published August 2025 (PMID 40460512, doi:10.1016/j.pediatrneurol.2025.04.015); Lyu 2026 (Minnesota medical cannabis program), published December 2026 (online August 2026) (PMID 42612572, doi:10.1016/j.yebeh.2026.111230); Ríos-Pohl 2024 (Chile, product analysis), published December 2024 (PMID 39503571, doi:10.1002/epi4.13081); Gidal 2024 (US commercial CBD products), published March 2024 (PMID 38562466, doi:10.3389/fphar.2024.1335441); Calapai 2024 (unlicensed CBD, European side effect reports), published January 2024 (PMID 37800192, doi:10.1002/ptr.8028)
Older changes (4)
The studies' sources were reread. Corrected: the dose of the CBD gel applied to the skin was written as a daily total (the journal's correction); the lower bound of seizure reduction in people taking placebo in the approved medicine's studies was corrected (13%); it was added that the between-group difference in the frontal lobe study was not statistically significant and that the groups were unbalanced at baseline; the denominator of the rate in the infantile spasms extension was corrected and the limitation of this measurement was added; the description of the prospective study with unregulated oils as the 'only' one was corrected; information on Turkey's list of medicines from abroad was added. The conclusion and the strength of evidence did not change.
PubMed and ClinicalTrials.gov were searched for newly published and previously missed clinical studies. 3 studies added: O'Brien 2022 (STAR 1), Ebadi 2023 and Patel 2025 (expanded access, focal epilepsies). In the 188-person randomized study testing CBD applied to the skin (O'Brien 2022), seizures did not fall differently from placebo; the text was updated accordingly. The conclusion and the strength of evidence did not change.
Related studies: O'Brien 2022 (STAR 1), published July 1, 2022 (PMID 35802375, doi:10.1001/jamanetworkopen.2022.20189, ACTRN12616000510448); Ebadi 2023, published July 3, 2023 (PMID 37470002, doi:10.3389/fneur.2023.1143783, IRCT20210608051515N1); Patel 2025 (expanded access, focal epilepsies), published October 2025 (PMID 40673944, doi:10.1111/epi.18496)
The full texts of the studies were reread, and the information on the page was checked against its sources. Corrected: participant numbers, side effects, results and duration in the study table; drug interaction notes; the approved medicine schedule; the evidence summary. Study added: GWPCARE7 pilot + extension (infantile spasms). Points that could not be verified from the full text went down from 6 items to 3. The year of one publication named on the page was corrected to the journal's print year instead of the online publication year. The conclusion and the strength of evidence did not change.
Related studies: GWPCARE7 pilot + extension (infantile spasms) (NCT02953548, NCT02954887)
Page published. This first review examined 9 studies.