Cannabidiol (CBD)

the one approved use, and what the fda has said about the rest

the fda has approved one cbd product: epidiolex, a purified plant-derived cannabidiol oral solution, for seizures associated with lennox-gastaut syndrome and dravet syndrome (approved june 2018) and tuberous sclerosis complex (added july 2020), in patients one year and older [label 1; fda-regulation]. the fda’s own words: that approval "means fda has concluded that this particular drug product is safe and effective for its intended use". there is no fda-approved cbd for anxiety, depression, insomnia, ptsd or any other mental health condition [status; fda-regulation]. epidiolex is not a controlled substance [label 9.1].

everything else sold as cbd is an unapproved consumer product. the 2018 farm bill removed hemp, cannabis with no more than 0.3% delta-9 thc by dry weight, from the controlled substances act, but the fda states that this "explicitly preserved fda’s authority" over products containing cannabis compounds, and that "it is currently illegal to market cbd by adding it to a food or labeling it as a dietary supplement" [fda-regulation; fda-consumer]. on january 26, 2023, the fda concluded "that a new regulatory pathway for cbd is needed", denied three citizen petitions asking it to allow cbd as a dietary supplement, and wrote that "it is not apparent how cbd products could meet safety standards for dietary supplements or food additives", because "we have not found adequate evidence to determine how much cbd can be consumed, and for how long, before causing harm" [fda-2023]. in practice that means retail cbd is sold with no premarket review of potency, purity or claims [status].

the fda’s enforcement has been warning letters, not a ban. its running list of letters to firms selling cannabis-derived products, including cbd and delta-8 thc, runs from 2015 to 2025 and held about 160 entries when read on the verification date (page dated september 3, 2025) [fda-letters]. the fda describes the letters as going to companies that claimed cbd products "prevent, diagnose, treat, mitigate, or cure serious diseases such as cancer, alzheimer’s disease, psychiatric disorders and diabetes", and its news releases list letters for claims about opioid addiction, pain relief, and infant teething and ear pain [fda-consumer; fda-regulation]. a warning letter is a regulator’s finding about a company’s marketing claim; it is not a finding that cbd does or does not work, and this page draws no inference from the letters beyond what the fda states.

what the epidiolex label found: the liver, sleepiness, suicidal thoughts and interactions

the label is the only place a regulator has written down what cbd does to the body at measured doses, and it is read here for that reason [label; risks], with the caveat the fda itself gives: the trials were in people with severe epilepsy, almost all on other anti-seizure drugs, at 10 to 25 mg per kg a day [label 6.1; fda-consumer]. the label’s own abuse-potential study maps doses to adults: 750, 1,500 and 4,500 mg correspond to 10, 20 and 60 mg per kg in a 75 kg adult [label 9.2], so the lowest trial dose is roughly 750 mg a day for an average adult.

• liver [5.1]: "epidiolex can cause dose-related elevations of liver transaminases." alt rose above three times the upper limit of normal in 13% of patients on 10 or 20 mg/kg and 12% on 25 mg/kg, against 1% on placebo; fewer than 1% exceeded twenty times the limit; some were hospitalised. the risk was 17% at 20 mg/kg against 1% at 10 mg/kg. it was concentrated in people also taking valproate: 30% with valproate and clobazam, 21% with valproate alone, 4% with clobazam alone, 3% with neither. most elevations came in the first two months, some up to 18 months in; about two thirds resolved after stopping or reducing cbd or valproate and about a third resolved during continued treatment. the label requires liver tests before starting and at one, three and six months [2.1, 5.1]. the fda’s consumer page adds that liver injury "was seen also in patients not on those drugs" and that it is "less clear how it might be managed when cbd is used far more widely, without medical supervision" [fda-consumer].
• sleepiness [5.2]: somnolence and sedation in 32% of patients on 10 or 20 mg/kg against 11% on placebo, 46% in those also taking clobazam. the label says not to drive or operate machinery until the effect is known, and that alcohol and other cns depressants add to it [5.2, 7.4]. in the 25 mg/kg trial the figure was 19% against 17%.
• suicidal thoughts [5.3]: the class warning carried by every anti-seizure drug, including epidiolex: in a pooled analysis of 199 placebo-controlled trials of 11 such drugs, 0.43% of drug-treated patients had suicidal thoughts or behaviour against 0.24% on placebo, about twice the risk and one extra case per 530 people treated. it is a class finding, not a cbd finding, and the label applies it to cbd because cbd is an anti-seizure drug.
• other drugs [7]: cbd is cleared by and interferes with several liver enzymes. the label says it raises the active metabolite of clobazam, raises blood levels of everolimus about 2.5-fold, and can raise or lower levels of drugs cleared by cyp2c19, cyp1a2, cyp2b6, cyp2c8 and ugt1a9, including, by the label’s named example, lowering the active form of clopidogrel. rifampin lowered cbd levels by about a third [7.1]. the library row on this label notes that some of the enzymes named also clear citalopram, escitalopram, clozapine, olanzapine and some benzodiazepines, and that nobody has measured the effect at retail doses [epidiolex-row]. the valproate label carries the mirror warning: with cannabidiol, raised liver enzymes, and liver tests during the combination [valproate-row].
• the rest [6.1]: in the lennox-gastaut and dravet trials, decreased appetite in 16% to 22% against 5% on placebo, diarrhoea 9% to 20% against 9%, raised transaminases 8% to 16% against 3%, rash 7% to 13% against 3%, and 16% lost at least 5% of body weight against 8%. dropouts for side effects were 2.7% at 10 mg/kg and 11.8% at 20 mg/kg against 1.3%. about 30% developed a new laboratory-defined anaemia against 13%. hypersensitivity reactions including angioedema occurred, and the product contains sesame oil [5.4].
• addiction [9.2, 9.3]: in a human abuse-potential study, cbd at 750 to 4,500 mg produced "drug liking" scores within the placebo range, while 2 mg of alprazolam and 10 to 30 mg of synthetic thc produced large increases; 28 days at 1,500 mg a day produced no withdrawal signs over six weeks. the label concludes cbd "likely does not produce physical dependence". this is one of the few places where the label is reassuring, and it is also one of the few findings that plausibly carries down to retail doses.

what the anxiety trials measured, in whom, and for how long

the trials that people cite for cbd and anxiety are laboratory challenge studies, not treatment trials; the companion page on this site walks through the same five records in more detail [topic-page]. in the first, 24 never-treated people with social anxiety disorder received a single 600 mg dose of cbd or placebo 90 minutes before a simulated public-speaking test; the cbd group reported less anxiety, cognitive impairment and discomfort during the speech, ending close to an untreated healthy comparison group [bergamaschi-2011]. in the second, 57 healthy men received 150, 300 or 600 mg or placebo before the same kind of test; only the 300 mg dose reduced anxiety, and 150 and 600 mg were no different from placebo, an inverted-u that means more is not better [linares-2019]. both came from the same university research group, several of whose members are co-inventors on a cbd patent the university has licensed to pharmaceutical companies, and the 2019 trial’s cbd was supplied free by a manufacturer [bergamaschi-2011; linares-2019].

the most-cited review of cbd for anxiety, from 2015, concluded that the animal evidence is strong but the human evidence is limited to single acute doses, with very few studies in people who have an anxiety disorder and essentially none on daily dosing; ten years on, that gap has not been filled [blessing-2015]. the harvest’s own note is blunter: every controlled trial showing an anxiolytic signal used a single oral dose of 300 or 600 mg of pharmaceutical-grade cbd, no controlled trial has tested the far smaller amounts in retail oils, gummies or vapes, and no controlled trial has tested repeated daily dosing for an anxiety or mood disorder [exclusions].

the largest review to date, published in 2026, pooled 54 randomised trials of cannabinoids as a primary treatment for mental and substance-use disorders in 2,477 people, searched to may 2025. it found no significant effect on anxiety outcomes and no randomised evidence at all for depression. cannabinoids raised the odds of any adverse event (odds ratio 1.75; one extra person harmed for every seven treated) without raising serious adverse events, and the authors concluded that routine use for mental disorders "is currently rarely justified" [wilson-2026]. that review covers cannabinoids broadly, cbd among them; it is the pooled answer to the single-dose signals above.

sleep: what the records on this page can and cannot say

no study record on this page is a trial of cbd for insomnia, and the harvest found no controlled trial of daily cbd for a sleep disorder to include [exclusions]. what exists is indirect. the epidiolex label counts somnolence and sedation as the most common nervous-system side effect, in 32% of patients against 11% on placebo, with a dose relationship [label 5.2]. the fda’s consumer page lists drowsiness as the most common change in alertness and warns that cbd with alcohol or with "other drugs that slow brain activity, such as those used to treat anxiety, panic, stress, or sleep disorders, increases the risk of sedation and drowsiness, which can lead to injuries" [fda-consumer]. the same page lists insomnia as a possible side effect too. so the honest statement is that cbd at prescription doses makes a third of people sleepy, that nobody has tested whether retail doses improve a sleep disorder, and that the combination with sleeping tablets or benzodiazepines is one the fda warns about.

what is in the bottle

laboratory analysis of 84 cbd products bought online found 26% contained less cbd than labelled and 43% more, roughly seven in ten mislabelled, and delta-9 thc was detected in 21% [bonn-miller-2017]. the study was funded by a cannabinoid research nonprofit on whose board three authors sit, a tie that cuts against the result rather than for it [bonn-miller-2017]. a 2023 analysis of 25 melatonin gummies found one product with no detectable melatonin and 31.3 mg of cbd instead [cohen-2023]. the fda reports testing products itself and finding that "many were found to not contain the levels of cbd they claimed", and that it is investigating reports of contaminants including pesticides, heavy metals and thc [fda-consumer]. whatever the trials found applies to the product and dose they used; a retail bottle cannot be assumed to contain the dose on its label.

what the research says, in both directions

in favour. a single 600 mg dose reduced anxiety in 24 people with social anxiety disorder in a laboratory test [bergamaschi-2011]; a single 300 mg dose did the same in healthy men [linares-2019]; the animal literature is consistent [blessing-2015]. at prescription doses cbd showed no abuse potential and no withdrawal, which the fda label states plainly [label 9.2, 9.3]. the fda has approved one form of it as safe and effective for its intended use, which is more than can be said of most supplements on this site [fda-regulation].

against, or smaller than it sounds. the two positive trials total 81 people, one dose each, in a laboratory, from one research group with patent interests, and the dose that worked was ten to a hundred times what retail products deliver [bergamaschi-2011; linares-2019; exclusions]. the pooled randomised evidence finds no effect on anxiety and none at all for depression, with more adverse events [wilson-2026]. the one label shows that at measured doses cbd raises liver enzymes in about one in eight people and interacts with common psychiatric drugs [label 5.1, 7]. the fda has said it cannot determine how much cbd can be consumed, and for how long, before causing harm [fda-2023], has flagged possible harm to the male reproductive system seen in animal studies [fda-consumer], and seven in ten retail products tested did not contain what the label said [bonn-miller-2017].

questions people ask

does cbd work for anxiety?

not shown in a treatment trial. two laboratory studies found that a single dose of 300 mg (in 57 healthy men) or 600 mg (in 24 people with social anxiety disorder) of pharmaceutical cbd reduced anxiety during a simulated public-speaking test, with 150 and 600 mg failing in the dose-ranging study. no controlled trial has tested daily cbd for an anxiety disorder, and none has tested the much smaller doses in retail products. the largest review of randomised trials, 54 trials in 2,477 people published in 2026, found no significant effect of cannabinoids on anxiety and more adverse events.

does cbd help you sleep?

no trial on this page tested it for insomnia, and the harvest found no controlled trial of daily cbd for a sleep disorder. what is documented is that at prescription doses cbd causes drowsiness in about a third of people (32% vs 11% on placebo in the epidiolex trials), and that the fda warns that cbd with alcohol or with medicines for anxiety, panic or sleep increases sedation and the risk of injury. sleepiness as a side effect is not the same as a trial showing better sleep.

is cbd fda approved or legal?

one cbd product is approved: epidiolex, a purified cannabidiol oral solution, for seizures in lennox-gastaut syndrome, dravet syndrome and tuberous sclerosis complex from age one. no cbd is approved for anxiety, depression, insomnia or any mental health condition. the 2018 farm bill removed hemp with no more than 0.3% thc from the controlled substances act, but the fda states it is illegal to add cbd to food or label it as a dietary supplement, and in january 2023 it denied three petitions to allow that and said a new pathway from congress is needed. the fda has sent about 160 warning letters since 2015 to companies selling cannabis-derived products with disease claims.

is cbd bad for your liver?

at prescription doses it can be. in the epidiolex trials, 12% to 13% of patients had the liver enzyme alt rise above three times the normal limit against 1% on placebo, with some hospitalised; the rate was 17% at 20 mg/kg a day against 1% at 10 mg/kg, and 30% in people also taking valproate and clobazam against 3% in people on neither. the label requires liver tests before starting and at one, three and six months. the fda says liver injury was also seen in patients not on those drugs and that it is less clear how the risk can be managed when cbd is used without medical supervision. whether retail doses carry the same risk has not been measured.

does cbd interact with medications?

yes, by the only label that exists. epidiolex raises the active metabolite of clobazam, raises everolimus about 2.5-fold, and can raise or lower drugs cleared by cyp2c19, cyp1a2, cyp2b6, cyp2c8 and ugt1a9, including lowering the active form of clopidogrel; with valproate it raises the rate of liver enzyme elevations, and the valproate label carries the mirror warning. some of those enzymes also clear citalopram, escitalopram, clozapine, olanzapine and some benzodiazepines. the fda says taking cbd with other medicines "may increase or decrease the effects of the other medications". nobody has measured the effect at retail doses, which is a reason to tell the prescriber, not to assume it is nothing.

is cbd addictive?

the epidiolex label says no, on the best evidence there is. in a human abuse-potential study, cbd at 750 to 4,500 mg produced "drug liking" scores within the placebo range while alprazolam and synthetic thc produced large increases, and 28 days at 1,500 mg a day produced no withdrawal signs. epidiolex is not a controlled substance. this does not transfer to products that also contain thc, which one in five online products did.

sources for this section

the library rows link to our one-page summary of each study, with the original paper, its funding and our evidence grade on that page.

1. the legal-status paragraph on this page (epidiolex as the one approved product; no psychiatric indication; the 2018 farm bill; the fda’s position on food and supplements; the january 2023 petition denials), written from the regulator’s documents in the 2026-08-18 study harvest and quoted whole above
2. the risks paragraph on this page (the 2026-08-18 study harvest’s known-risks record for cbd: the liver numbers, enzyme interactions, sedation, the suicidality warning, mislabelling, the fda’s flagged concerns)
3. the "what is not known" notes on this page: the epilepsy trials excluded as off-topic for mental health, open-label anxiety-clinic series excluded, the dose gap stated, no controlled trial of daily dosing for an anxiety or mood disorder; retraction check 2026-08-18
4. epidiolex (cannabidiol) oral solution prescribing information, fda label on dailymed (set id 8bf27097-4870-43fb-94f0-f3d0871d1eec, spl version 35, effective may 29, 2026; jazz pharmaceuticals, nda 210365): sections 1, 2.1, 4, 5.1 to 5.5, 6.1 tables 3 and 4, 7.1 to 7.4, 9.1 to 9.3
5. the epidiolex label read for the library: it raises active clobazam, raises liver enzymes with valproate, and adds to alcohol (dailymed spl version 35) (library row)
6. the valproate (depakote) label read for the library: carbapenem antibiotics, topiramate, cannabidiol and lamotrigine each interact in a different way (dailymed spl version 1621, april 2026) (library row)
7. us food and drug administration. fda regulation of cannabis and cannabis-derived products, including cannabidiol (cbd): questions and answers, news releases and statements 2017 to 2024; read on the verification date
8. us food and drug administration. what you need to know (and what we’re working to find out) about products containing cannabis or cannabis-derived compounds, including cbd (consumer update); read on the verification date
9. us food and drug administration. fda concludes that existing regulatory frameworks for foods and supplements are not appropriate for cannabidiol, will work with congress on a new way forward (statement, january 26, 2023)
10. us food and drug administration. warning letters for cannabis-derived products (the running list, 2015 to 2025; page dated september 3, 2025); read on the verification date
11. wilson j, dobson o, langcake a, et al. the efficacy and safety of cannabinoids for the treatment of mental disorders and substance use disorders: a systematic review and meta-analysis. lancet psychiatry 2026 (an erratum is recorded on the row; no retraction) (library row)
12. bergamaschi mm, queiroz rh, chagas mh, et al. cannabidiol reduces the anxiety induced by simulated public speaking in treatment-naïve social phobia patients. neuropsychopharmacology 2011 (university of são paulo cbd group; patent interests recorded on the row) (library row)
13. linares im, zuardi aw, pereira lc, et al. cannabidiol presents an inverted u-shaped dose-response curve in a simulated public speaking test. brazilian journal of psychiatry 2019 (study drug supplied free by a manufacturer; four authors co-inventors on a cbd patent) (library row)
14. blessing em, steenkamp mm, manzanares j, marmar cr. cannabidiol as a potential treatment for anxiety disorders. neurotherapeutics 2015 (narrative review) (library row)
15. bonn-miller mo, loflin mje, thomas bf, marcu jp, hyke t, vandrey r. labeling accuracy of cannabidiol extracts sold online. jama 2017 (library row)
16. cohen pa, avula b, wang yh, katragunta k, khan i. quantity of melatonin and cbd in melatonin gummies sold in the us. jama 2023 (library row)
17. does cbd help anxiety? the dose-response trials, the companion page on this site, which walks through the same five study records in more detail