Almost everything we know about cannabidiol as an epilepsy drug comes from children. The clinical trials that got it approved recruited mostly young patients with rare, severe syndromes that begin in early childhood. But those children grow up, and their seizures usually come with them. A new retrospective study from three hospitals in southern Spain asks a question that has been hanging in the air for years: does cannabidiol keep working once the patient is an adult, and does the body tolerate it the same way?
The paper, published in September 2026 in Epilepsia Open under the title “Effectiveness and safety of cannabidiol in adult patients with epilepsy: A multicenter, retrospective study” (DOI: 10.1002/epi4.70331), follows 37 adults with Lennox–Gastaut syndrome, Dravet syndrome or tuberous sclerosis complex who took pharmaceutical-grade cannabidiol for at least a year. The lead author is Sara Sánchez-Gamino of Hospital Regional Universitario de Málaga; the work was funded by Jazz Pharmaceuticals, the company that sells the drug. Both of those facts matter for how the results should be read, and we will come back to them.
Key Takeaways
After 12 months of treatment, 45.9% of the adults in this cohort had at least halved their seizure frequency, and 59.5% had cut it by a quarter or more. Roughly four in five patients were still taking cannabidiol a year in, a retention rate higher than most comparable real-world studies have reported. Side effects were common, affecting 64.9% of patients, but were mostly mild; the most frequent were drowsiness, abnormal liver blood tests and irritability, and six patients stopped treatment because of them. Patients were also able to reduce the number of other seizure medicines they took, from a median of four to three. Caregivers reported better alertness or behaviour in about three in ten patients. The study has no control group, a small sample, and industry funding, so it shows an association in everyday clinical practice rather than proof that cannabidiol caused the improvement.
Who Was in the Study
The 37 patients were treated at Hospital Regional Universitario de Málaga, Hospital Universitario Virgen de la Victoria (also in Málaga) and Complejo Hospitalario de Jaén. To be included, a patient had to be 18 or older, carry a diagnosis of Dravet syndrome, Lennox–Gastaut syndrome or tuberous sclerosis complex, have started cannabidiol before August 2024, and have at least 12 months of follow-up data in their medical chart.
The median age was 25, with the middle half of patients falling between 22 and 39. Two thirds (67.5%) were male, and median body weight was 60 kg. Almost everyone, 34 of 37, had Lennox–Gastaut syndrome, a form of epilepsy that begins in childhood, brings several seizure types at once and is notoriously hard to control. One patient had Dravet syndrome and two had epilepsy linked to tuberous sclerosis complex, a genetic condition that causes benign growths in the brain and other organs.
These were heavily treated patients. Before cannabidiol they had already been through a median of four anti-seizure medications. By the definition used by the International League Against Epilepsy, epilepsy counts as drug-resistant once two appropriate medicines have failed; these patients were well past that threshold. Most also had moderate-to-severe intellectual disability, which, as the authors note, made formal cognitive testing impractical.
Cannabidiol here means Epidyolex, the highly purified oral solution approved by the European Medicines Agency for these three conditions in patients aged two and older, given alongside clobazam for Dravet and Lennox–Gastaut (EMA product information). It is not the CBD oil sold in shops. The patients started at a median of 4.41 mg per kilogram of body weight per day, settled at a maintenance dose of 10.45 mg/kg/day, and reached a median maximum dose of 11 mg/kg/day. That sits at the lower end of what the label allows, a point the authors flag when comparing their results with other cohorts.
How the Researchers Measured Seizures
This was a retrospective chart review, meaning the researchers went back through existing medical records rather than designing a trial and recruiting patients in advance. The seizure data themselves came from daily diaries kept by patients or, more often, their caregivers, which clinicians reviewed at scheduled visits before treatment and again at 6 and 12 months.
Baseline seizure frequency was defined as the total number of clinically relevant seizures in the six months before cannabidiol was started. Response was then sorted into three tiers: a reduction of more than 25%, at least 50%, and at least 75% compared with baseline. The 50% threshold is the conventional bar for calling someone a “responder” in epilepsy research.
One limitation the authors state plainly: they could not analyse results by seizure type. Many patients had several kinds of seizure at once, and caregivers could not reliably tell them apart in a way that would hold up statistically. So the study measures overall seizure burden, not, say, drop attacks specifically.
Cognitive and behavioural changes were tracked by asking families at each visit whether they had noticed any change in alertness, interaction, behaviour, sleep or daily functioning. No standardised questionnaire was used.
Seizure Reduction: The Main Results
At 6 months, 24 of 37 patients (64.9%) had reduced their seizures by at least a quarter, 18 (48.6%) by at least half, and 9 (24.3%) by at least three quarters. At 12 months, the numbers were 22 (59.5%), 17 (45.9%) and 3 (8.1%) respectively.
Read those figures carefully. Roughly half of these adults were responders at the one-year mark, which is in the same range as paediatric trials. But the share of very strong responders, those with a 75% or greater reduction, dropped noticeably between 6 and 12 months, from about a quarter of patients to under a tenth. The paper does not dwell on this, but it is a reminder that early gains do not always hold at the same intensity.
A small number of patients also had shorter seizures even when the count did not fall: two patients (5.4%) at 6 months and four (10.8%) at 12 months.
The researchers checked whether dose explained who responded and who did not. It did not: median maintenance dose was statistically indistinguishable across the response tiers. Nor did taking clobazam alongside cannabidiol predict a 50% response, a finding worth noting because clobazam and cannabidiol interact chemically, and some researchers have argued that part of cannabidiol’s effect works through raising clobazam levels in the blood. This small dataset does not settle that debate, but it does not support the idea that clobazam is doing the heavy lifting.
Fewer Other Medicines
One of the more practical findings is what happened to the rest of the patients’ drug regimens. Before cannabidiol, the median patient took four other anti-seizure medications. By the end of follow-up, that had fallen to three, a change the authors report as statistically significant (p = 0.022).
For people with severe epilepsy, every additional drug carries its own side effects, interactions and cognitive fog. Shaving one medicine off a four-drug regimen is not trivial. Similar patterns have appeared elsewhere: the authors cite a US expanded access programme in which doses of concomitant medicines were reduced in 46% of patients taking clobazam and 52% of those on valproic acid (Laux et al., 2019).
There is a catch, though, and the authors acknowledge it. If doctors adjusted other medicines during follow-up, some of the seizure improvement might come from those adjustments rather than from cannabidiol itself. In a retrospective study without a control group, the two cannot be cleanly separated.
What Caregivers Noticed
Families reported improvements in alertness, behaviour, interaction or everyday functioning in 11 of 37 patients (29.7%). These improvements were not linked to dose, to how much seizures fell, or to how many other drugs were dropped.
The authors treat this result with visible caution, and rightly so. It is entirely subjective, it was collected without a validated instrument, and families who hope a new treatment will work are prone to seeing improvement. At the same time, the researchers make a fair defence of caregiver reports in this population: most of these patients cannot complete standard neuropsychological tests, and the people who live with them see changes that a 20-minute clinic visit would miss. Caregiver-reported outcomes are increasingly accepted as meaningful endpoints in developmental and epileptic encephalopathies, and this study follows that trend without overclaiming.
Safety: Common but Mostly Mild
Twenty-four of 37 patients (64.9%) experienced at least one adverse event. The most common were drowsiness (10 patients, 27%), abnormal liver function tests (7 patients, 18.9%) and irritability (4 patients, 10.8%). Less frequent were gastrointestinal symptoms, insomnia and paradoxical worsening of seizures (2 patients each, 5.4%), and single cases of increased appetite, loss of appetite and urinary retention.
The liver results deserve a closer look because they are the best-known safety concern with pharmaceutical cannabidiol. In all seven cases the enzyme increase stayed below twice the upper limit of normal, a level generally considered mild. Six resolved without any change to treatment, and one resolved after the dose of valproic acid, a medicine known to stress the liver, was tapered. No patient stopped cannabidiol because of liver results, and the researchers found no correlation between liver abnormalities and either the maintenance or the maximum dose.
Six patients (16.2%) withdrew because of side effects: drowsiness in three, and one each for insomnia, irritability, gastrointestinal symptoms and loss of appetite. Overall, seven patients discontinued: one for lack of effect alone, one for side effects alone, and five for both.
An interesting detail emerges when dose is compared between those with and without side effects. Patients who had adverse events were on a significantly lower maintenance dose, 7 mg/kg/day versus 12.5 mg/kg/day. That sounds backwards until you consider the likely explanation: clinicians probably kept the dose low in patients who were already showing side effects, or those patients never reached a higher dose because the side effects arrived first. It illustrates why retrospective data can describe what happened but struggle to explain why.
Compared with other real-world cohorts, the 65% adverse event rate was a little higher than a large German chart review (42%) and an Italian expanded access programme (52%), but well below a US expanded access programme (91%) and the long-term extension trials (96–97%). The authors attribute the spread to differences in patient populations, how fast doses were increased, and how other medicines were managed.
Retention: Four in Five Still Taking It
At 12 months, 30 of 37 patients (81.1%) were still on cannabidiol. Follow-up for those who continued was long, a median of 34 months, and some patients had been on the drug for nearly six years by the time records were reviewed. Everyone who stopped did so within the first 20 months, which the authors read as reassuring: the window is long enough to catch early side effects or confirm that the drug is not helping, and beyond it, people tended to stay.
Retention of 81% compares favourably with the 61–76% reported in other real-world studies and with the 50–62% seen in long-term extension trials. The authors suggest that everyday clinical practice, where doctors can adjust doses and other medicines freely, may simply be a more forgiving environment than a rigid trial protocol.
How This Fits With Other Evidence
The Spanish results land squarely within the range other groups have reported. A German multicentre chart review of 202 patients with the same three conditions found a 44% responder rate at 12 months, with adults aged 18 and over responding at 37% (Strzelczyk et al., 2025). The Italian expanded access programme reported 49% responders at 12 months (Iannone et al., 2021). An earlier Spanish expanded access programme covering both children and adults had already pointed in the same direction (Villanueva et al., 2022). And the original randomised trials in Dravet syndrome (Devinsky et al., 2017) and Lennox–Gastaut syndrome established the effect under controlled conditions, mostly in younger patients.
What the new paper adds is not a dramatic new number but a confirmation in an underrepresented group: adults, followed for a long time, in ordinary hospital practice, on relatively modest doses. That is genuinely useful information for neurologists deciding whether to continue or start cannabidiol in patients who have aged out of paediatric care. For readers interested in how cannabidiol behaves differently across age groups, our piece on CBD pharmacokinetics in older adults covers the other end of the lifespan.
The Limitations, Stated Plainly
The authors list them honestly, and so should we. The study is retrospective, so it can only describe what happened to patients who were already being treated. There is no control group, so there is no way to know how these patients would have fared without cannabidiol; some seizure patterns fluctuate on their own. Thirty-seven patients is a small sample, and the fact that 34 had Lennox–Gastaut syndrome means the study says almost nothing about Dravet syndrome or tuberous sclerosis in adults specifically.
Seizure type could not be analysed. Cognitive and behavioural improvement was measured by asking families, with no standardised tool, and cannot be reproduced. Changes to other medicines during follow-up may have contributed to the results. The study was funded by Jazz Pharmaceuticals, which markets Epidyolex, and the authors disclose research funding, consultancy fees or speaker honoraria from Jazz and a dozen other pharmaceutical companies. Medical writing support was also funded by Jazz. None of this means the data are wrong, but it is the kind of context readers should weigh, especially when the conclusions are favourable to the sponsor’s product.
Finally, a point that applies to every article about cannabidiol and epilepsy: these results concern a prescription medicine with a defined purity, a defined dose and specialist monitoring, including blood tests. They do not transfer to over-the-counter CBD products, whose actual cannabidiol content often bears little relation to the label, as we reported in our analysis of mislabelled CBD oils in Europe. Our complete guide to CBD explains that distinction in more depth, and a recent six-country survey shows how far the reasons people actually use CBD have drifted from the conditions for which it is approved.
What Happens Next
The authors call for prospective studies with larger samples, control groups, structured caregiver training, standardised seizure diaries and pre-planned subgroup analyses by cause of epilepsy. Those are the studies that would turn this association into something closer to proof. Until then, this paper is best read as what it is: a careful description of how 37 adults with severe, drug-resistant epilepsy fared over several years on a relatively low dose of pharmaceutical cannabidiol, with results that look a lot like what has been seen in children.
FAQ
The responder rates were in the same range. About 46% of adults had at least halved their seizures at 12 months, compared with 44–49% in large paediatric and mixed-age real-world cohorts and, in the German study cited by the authors, 37% among adults specifically. The study cannot prove cannabidiol caused the reduction, but it found no sign that adults respond worse.
No. The patients took Epidyolex, a highly purified prescription medicine dosed by body weight and monitored with blood tests. Consumer CBD products are not standardised, are not approved to treat epilepsy, and independent testing has repeatedly found their actual cannabidiol content differs from what the label says.
Drowsiness affected 27% of patients, abnormal liver tests 18.9% and irritability 10.8%. The liver changes were all mild, below twice the normal range, and resolved without stopping the drug. Six patients (16.2%) did stop because of side effects, most often drowsiness. The authors found no safety problems beyond the drug’s already known profile.
Yes, by prescription. In the United States, Epidiolex is approved by the FDA for seizures associated with Lennox–Gastaut syndrome, Dravet syndrome and tuberous sclerosis complex in patients aged one year and older, and it was removed from the Controlled Substances Act schedules in 2020. In the European Union, Epidyolex holds a centralised marketing authorisation from the European Medicines Agency for the same three conditions in patients aged two and older. In both jurisdictions it is a prescription-only medicine typically started by an epilepsy specialist; general CBD products sold as supplements or cosmetics are not approved for epilepsy and are regulated separately.
Legal Disclaimer
This article is a journalistic summary of a published scientific study and is provided for general information only. It is not medical advice and does not substitute for consultation with a qualified neurologist or other healthcare professional. Cannabidiol as discussed here refers to a prescription medicine used under specialist supervision; people with epilepsy should never start, stop or change any anti-seizure medication without medical guidance. The legal status of cannabidiol products varies by country and by product category; readers should verify the rules in their own jurisdiction. The Cannex does not endorse any product or manufacturer mentioned in this article.
