Cannabidiol (CBD)

Nobody was randomised to clobazam — patients arrived on it or not — so the two groups differ in ways beyond the drug. The response was much higher in the clobazam group in both arms, placebo included, which the authors flag rather than resolve. Only Dravet and Lennox-Gastaut trials are pooled.

The trials ran one to six months and 64% concerned nerve pain, so this does not cover long-term use or the everyday aches cannabidiol is sold for. The cannabidiol-dominant result is a null across a whole product category, not a tested null for any particular product or dose.

One trial does not fit cleanly: in Lennox-Gastaut syndrome the 20mg/kg/day arm ran about five percentage points ahead of the 10mg/kg/day arm on drop seizures. No trial here powered its two active doses against each other, so every dose comparison is descriptive rather than a tested difference.

Eighty-four separate comparisons across 23 conditions means some significant findings are expected by chance, and most were graded low or very low quality. The missing cannabidiol effect on pain is an absence of evidence inside these trials, not a tested null.

The placebo groups improved too — 21.8% fewer drop seizures in Lennox-Gastaut. In Dravet, the share of patients whose seizures halved did not separate from placebo (43% against 27%), so the finding rests on the seizure count alone. Both trials were manufacturer-funded.

The search closed in April 2018, before several cannabidiol-specific trials were published. The one anxiety signal the review did find came from pharmaceutical THC, in people whose main problem was chronic pain or multiple sclerosis, and was graded very low quality.

Twenty people, run by the maker of the cannabidiol formulation, and built to measure blood levels rather than outcomes — it establishes the exposure change, not its consequences. The dose is a prescription epilepsy one; whether the same happens at consumer amounts was not tested.

The prescription figures are set per kilogram of body weight; the milligram equivalents here use the label's own 75kg reference, and many of the patients in these trials are children. The consumer average has a standard deviation of 40.7mg, nearly as wide as the mean itself.

Single doses in a residential unit, about six people per formulation, so this is not a test of daily use. Cannabidiol did not convert to THC by any route tried here, but only the routes and formulations tried are covered.

This is the agency's stated position under US federal law, not a court ruling and not a description of other countries or of individual states. The page was last updated in July 2024 and the law in this area moves.

Urine, not blood, so these concentrations cannot be set beside the plasma figures on the prescription label. About six people took each formulation, and each took a single dose in a residential unit rather than using the product over time.

A pilot of thirty people cannot rule out a small effect. One objective measure did move — sleep efficiency favoured cannabidiol at week two — while every measure the sleepers reported themselves did not. And 150mg nightly is about three times what people typically take.

Each study sets its own tolerance — a tenth either way here, a stricter band for the drinks — so the percentages describe different tests and should not be pooled. The samples are snapshots too: one online channel in 2016, one US state later, with single-figure counts inside some categories.

These rates come from patients with severe epilepsy on cannabidiol at 10 to 25mg/kg/day, far above what consumers take. NIH's LiverTox reference gives 21% for the valproate-and-clobazam group where the label gives 30%, attributing its figure to extension studies rather than to these trials.

Both are dated regional snapshots — products bought online in 2016 and in one US state in 2020 — and neither describes what is on shelves now. The synthetic cannabinoids turned up in vaping products specifically, so nothing here says the same of oils or edibles.

The pooled studies mix one-off experimental stress in healthy volunteers with treating diagnosed anxiety disorders — different questions, answered as one. A large pooled effect resting on trials this small is where the smallest studies dominate, and the authors caution against over-reading it.

Whether cannabidiol eases anxiety is still open. A large pooled effect for cannabidiol alone and scarce benefit across all medicinal cannabinoids, mostly THC trials, answer different questions, so neither settles it.

There was no control group — the comparison is to historical population figures, not a concurrent placebo arm. Blood was drawn once at 30 days, so a delayed signal would not appear. Participants were recruited through twelve cannabidiol companies, and three of four authors declare industry ties.

The prescription label records liver-enzyme rises that a study of people buying cannabidiol for themselves did not find. Doses differ more than tenfold and so do the people, so the split is about who was studied, not the compound.