CBD and clinical trials: what is proven, and at what dose

Epidyolex, an oral solution of purified cannabidiol, was authorised by the European Medicines Agency on 19 September 2019. Its indications, from the age of 2, are Lennox-Gastaut syndrome and Dravet syndrome, in combination with clobazam, and tuberous sclerosis complex.

Randomised, placebo-controlled trials support the authorisation. In Dravet syndrome, 120 patients took 20 mg/kg/day of CBD or a placebo for 14 weeks. Over that period, the median monthly count of convulsive seizures went from 12.4 to 5.9 on CBD, against 14.9 to 14.1 on placebo. A Lennox-Gastaut trial of 171 patients aged 2 to 55 measured a median fall in drop seizures of 43.9%, against 21.8% on placebo.

The results came with side effects: diarrhoea, drowsiness, loss of appetite and raised liver enzymes are more frequent on CBD. In the Lennox-Gastaut trial, 14% of patients on CBD stopped because of an adverse effect.

A French study has since looked at everyday practice. At the Necker-Enfants malades hospital in Paris, 103 children with drug-resistant epilepsy were followed for 18 months after they started CBD. Their parents reported that 54% of the children had fewer seizures after one month. At 18 months, 55% were still on it, and 20% had lost appetite. These trials are covered in detail in our article on Epidyolex.

The anxiety studies cited most often are short and small. In 2011 a Brazilian team gave 600 mg of CBD or a placebo to 24 people with social anxiety, 12 per group, before a simulated public-speaking test. CBD reduced self-reported anxiety and discomfort during the test, in a study its own authors call preliminary.

An imaging study from the same year gave 400 mg to 10 patients with social anxiety, whose self-reported anxiety fell. In 2018, a trial in 32 volunteers with marked paranoid traits found no effect of 600 mg of CBD on anxiety or on paranoid thinking.

A meta-analysis published in The Lancet Psychiatry in 2026 sums up the field. It pools 54 randomised trials of cannabinoids (CBD, THC or both) in 2,477 participants with mental or substance use disorders. The certainty of evidence is rated low for most outcomes, 44% of the trials carry a high risk of bias, and no significant effect appears on anxiety. For now, the authors conclude, routine use is "rarely justified".

Online, the sleep study quoted most is a case series published in The Permanente Journal in 2019. American doctors went through the records of 72 adults they had offered CBD, mostly 25 mg a day, alongside their usual care. Sleep scores improved during the first month for 66.7% of them, then fluctuated over time. There was no control group and no placebo, so the improvement may reflect patients' expectations, the medical follow-up or simply time.

A controlled trial in 2018 compared 300 mg of CBD with placebo in 27 healthy volunteers, recording 8 hours of sleep. CBD changed neither sleep stages nor sleep duration. Our article on CBD and sleep goes into the question in detail.

In 2024, four researchers writing in the Journal of Pain listed every randomised CBD trial with pain as an outcome. They found 16, covering 12 types of pain at doses from 6 to 1,600 mg, and 15 of the 16 trials show no benefit of CBD over placebo.

Sativex sometimes comes up as an example. This medicine combines THC and CBD in roughly equal parts and is authorised for muscle stiffness in multiple sclerosis once other options have failed, so it tells us nothing about CBD alone. What French law allows anyone to say on the subject is described in our article on CBD and pain.

The trials with positive results used high doses, given once or under medical supervision. The chart sets these doses on a single scale beside the provisional level EFSA set for foods.

Several orders of magnitude separate them. On 9 February 2026 EFSA published a provisional safe level of 0.0275 mg per kilo per day, about 2 mg for an adult of 70 kg. The level does not cover people under 25, pregnant or breastfeeding women, or anyone on medication. Our article on how much CBD per day explains the calculation.

In 2022 EFSA concluded that the safety of CBD as a novel food could not yet be established. Data were missing on the liver, the digestive tract, the hormonal and nervous systems, psychological effects and reproduction. In 2025 the French agency ANSES proposed to classify CBD as a presumed reproductive toxicant, category 1B.

These uncertainties have legal consequences. Article 10 of Regulation (EC) No 1924/2006 prohibits any health claim on a food unless that claim has been authorised and entered on the Union list. In France, foods containing CBD are not authorised under the novel food regulation, as our overview of regulation in 2026 explains. Medical cannabis is another matter, covered in our article on its general roll-out in France.

When you see a headline such as "CBD improves sleep", five questions tell you how much weight the study can bear.

1. Is there a randomised placebo group? If not, the effect may come from expectation or from time.
2. How many people took part? With 12 per group, a result is only a lead.
3. What dose was given, and for how long? A single dose of 600 mg tells you nothing about 20 mg a day over months.
4. Was CBD tested alone or with THC? Sativex and many meta-analyses combine the two.
5. Has another team reproduced the result? On anxiety, the 2018 trial contradicts the one from 2011.