CBD Safety and Drug Interactions: Liver, Medicines, Pregnancy and Driving

A bottle of CBD oil can sit on the shelf between the vitamin D and the herbal tea and look like the most polite guest in the cupboard. In your body it behaves less politely. CBD is handled by the same liver enzymes that process many prescription medicines, and the best-documented problems with it are liver enzyme rises, sleepiness, diarrhoea and interactions with certain drugs.

Let's be fair to CBD and to the evidence. Most of the hard data comes from prescription CBD given to people with severe epilepsy at high weight-based doses, plus a handful of dedicated studies in healthy volunteers. That tells us what CBD can do. It doesn't tell us that a small amount of retail oil is harmless, and it doesn't tell us that a given combination with your medicine is fine. Below is what each source shows, how strong it is, and what it leaves open.

What actually goes wrong in trials

The most useful single summary is a 2020 systematic review and meta-analysis of randomised placebo-controlled trials lasting at least seven days (Chesney and colleagues, Neuropsychopharmacology): 12 trials, 803 participants. Compared with placebo, CBD was linked to more withdrawals because of adverse events, and to more decreased appetite, diarrhoea, somnolence and sedation. Serious adverse events linked to abnormal liver function tests were also more common (odds ratio 11.2, but with a very wide confidence interval of 2.1 to 60.0, because the events were few). Their own conclusion was that the trial data suggest CBD is "well tolerated and has relatively few serious adverse effects", provided interactions with other medicines are monitored carefully. The authors point out that liver, somnolence, sedation and pneumonia signals were limited to trials in childhood epilepsy, where CBD may have interacted with clobazam or sodium valproate. Once those trials were left out, diarrhoea was the only adverse outcome that stood out. They also said that data from trials outside childhood epilepsy and from over-the-counter products were still needed.

The prescribing information for the prescription CBD medicine Epidiolex gives frequencies from its own controlled trials (children and adults with Lennox-Gastaut or Dravet syndrome). The European product information lists the most common reactions at 10 to 25 mg/kg/day as somnolence (23 per cent), decreased appetite (21 per cent), diarrhoea (20 per cent), fever (16 per cent), vomiting (12 per cent) and fatigue (10 per cent). European product information, section 4.8. In the US label, somnolence and sedation occurred in 32 per cent of treated patients versus 11 per cent on placebo, and in 46 per cent of those who also took clobazam versus 16 per cent of those who didn't (section 5.2).

Those percentages describe specific people at specific, high, weight-based doses. They don't predict what a retail product will do to you, and a lack of events in a small study can't show that a risk is absent.

The liver: the signal with the best evidence

In plain terms: in prescription trials, liver enzyme rises climb with dose and are most common alongside valproate, and one trial in healthy adults found them without any epilepsy medicine involved. The evidence is stronger than most "side effects" lists suggest. The detail:

• Prescription trials. The US label says CBD causes dose-related rises in ALT and AST. In controlled trials, ALT above three times the upper limit of normal occurred in 13 per cent of patients on 10 or 20 mg/kg/day and 12 per cent on 25 mg/kg/day, compared with 1 per cent on placebo. At 10 mg/kg/day alone it was 1 per cent; at 20 and 25 mg/kg/day it was 17 and 12 per cent. Among patients with Lennox-Gastaut or Dravet syndrome at 10 and 20 mg/kg/day, ALT rises of this size occurred in 30 per cent of those also taking valproate and clobazam, 21 per cent with valproate alone, 4 per cent with clobazam alone and 3 per cent with neither. The label says most cases arose in the first two months, with some up to 18 months in, particularly with valproate. (Sections 5.1 and 7.3.)
• A healthy-volunteer trial. In a randomised, double-blind, placebo-controlled trial of 201 healthy adults (Florian and colleagues, 2025, JAMA Internal Medicine), participants took an oral CBD solution at 5 mg/kg a day (about 400 mg a day for an 80 kg adult, in two doses) or placebo for 28 days. The authors describe that dose as within the range of reported consumer use; most people who self-report their dose say 0 to 100 mg a day, so it sits above the usual end of that range. Eight people on CBD (5.6 per cent) and none on placebo had ALT or AST above three times the upper limit of normal; seven met criteria for stopping because of possible drug-induced liver injury. Rises appeared from about three weeks, and, the authors report, returned to baseline within one to two weeks of stopping; most of those with rises also showed raised eosinophils, a type of white blood cell. Only one participant reported symptoms (temporary stomach discomfort). The trial was funded by the US FDA. The authors concluded that lower doses, longer use and other populations need study.
• A regulatory-style review. The German Senate Commission on Food Safety (SKLM) reviewed CBD in foods and supplements (Engeli and colleagues, 2025, Nutrients) and identified 4.3 mg/kg a day for three to four weeks as the lowest dose with raised liver enzymes in healthy volunteers. Because lower doses hadn't been tested, it couldn't set a no-effect level.
• EFSA. In its 9 February 2026 provisional safe-level announcement, EFSA lists liver effects among the data gaps, alongside the endocrine, nervous and reproductive systems.

You can't feel a rising liver enzyme. The US label tells clinicians to act on symptoms such as unexplained nausea, vomiting, pain in the upper right abdomen, unusual tiredness, loss of appetite, yellowing of the skin or eyes and dark urine. If you notice those while using CBD, speak to a doctor or pharmacist promptly and say what you've been taking. Severe or rapidly worsening symptoms need urgent medical help.

Medicine interactions, and how much to trust each one

CBD is broken down mainly by the liver enzymes CYP2C19 and CYP3A4 (plus UGT enzymes) and, in turn, can slow down other enzymes and transporters. The US label describes CBD as a moderate inhibitor of CYP2C19, a weak inhibitor of CYP1A2 and an inhibitor of UGT1A9, and says orally taken CBD can raise exposure to P-glycoprotein substrates. A systematic review of reported interactions between narrow-therapeutic-index drugs and THC, CBD or cannabis (Nachnani and colleagues, 2024, Frontiers in Pharmacology; the review covers all three together, not CBD alone) found 31 reports; warfarin, valproate, tacrolimus and sirolimus were the most frequently reported combinations. A mechanism says an interaction is plausible. A measured change in blood levels says it happens. A clinical event says it can matter. The table keeps those apart.

• Medicine or class: Clobazam (epilepsy); What the source shows: The active metabolite N-desmethylclobazam rose about 3-fold (Cmax and AUC) in healthy volunteers given Epidiolex (US label); the European text says 3- to 4-fold. In 13 children with epilepsy (Geffrey and colleagues, 2015), mean norclobazam rose about 500 per cent at 4 weeks, clobazam doses were reduced in 10 of 13, and side effects reported by 10 were relieved by reducing clobazam; Type of evidence: Dedicated healthy-volunteer study plus a small uncontrolled clinical series; What it leaves open: Doses were medicinal; effect with low retail doses isn't measured
• Medicine or class: Valproate (epilepsy); What the source shows: More liver enzyme elevations, diarrhoea and reduced appetite (labels). Valproate levels fell only slightly in patients (about 17 to 21 per cent), so the interaction is mainly about liver effects; Type of evidence: Trial data in prescribing information: strong for liver signal; What it leaves open: Findings apply to the doses and patients in the trials
• Medicine or class: Stiripentol; What the source shows: Stiripentol exposure rose by 28 to 55 per cent in healthy volunteers (17 to 30 per cent in patients); Type of evidence: Dedicated studies; What it leaves open: Clinical significance varies
• Medicine or class: Tacrolimus, everolimus, sirolimus, digoxin (P-gp substrates); What the source shows: Everolimus exposure rose about 2.5-fold with CBD 12.5 mg/kg twice daily. In a phase I study of 12 healthy participants (So and colleagues, 2025), tacrolimus peak concentration rose 4.2-fold and total exposure 3.1-fold after 14 days of CBD at up to 5 mg/kg twice daily. A participant in an epilepsy trial who also took tacrolimus, on 2,000 to 2,900 mg of CBD a day, had about 3-fold higher dose-normalised tacrolimus levels (case report); Type of evidence: Dedicated studies for everolimus and tacrolimus; case report for a patient; sirolimus and digoxin are label-level cautions by analogy; What it leaves open: Doses were high; transplant medicines have narrow margins, so any use needs the transplant team
• Medicine or class: Warfarin; What the source shows: Two case reports describe warfarin dose reductions after starting Epidiolex: one patient needed a reduction of nearly 20 per cent; another had an INR of 6.86 and needed a 30 per cent weekly reduction; Type of evidence: Case reports (weak for frequency, serious for outcome); among the most frequently reported combinations in the 2024 review (which covered THC, CBD and cannabis together); What it leaves open: No trial; size of risk unknown. Both cases involved Epidiolex rather than retail oil
• Medicine or class: Citalopram; What the source shows: In a study with two cohorts of 20 healthy adults each (Salcedo and colleagues, 2026), the citalopram cohort took CBD for 12 days at 2.5 mg/kg twice daily, and their citalopram total exposure was about 43 per cent higher (ratio 1.43); Type of evidence: Open-label, sequential study, no placebo group; What it leaves open: Healthy volunteers, single citalopram dose, high weight-based CBD dose
• Medicine or class: Amitriptyline; What the source shows: A single 30 mg CBD dose raised amitriptyline exposure by 13 per cent and peak by 17 per cent in 12 participants who completed the study (Gorbenko and colleagues, 2026); Type of evidence: Small open-label crossover; What it leaves open: The authors suggest that regular use could have a bigger effect; this is a single dose
• Medicine or class: Morphine; What the source shows: Little change in total exposure; peak up 11 to 19 per cent; Type of evidence: Open-label study (Salcedo, 2026); What it leaves open: Healthy volunteers
• Medicine or class: Caffeine; What the source shows: Exposure (AUC) up 95 per cent, peak up 15 per cent with Epidiolex 750 mg twice daily and a single 200 mg caffeine dose (US label); Type of evidence: Dedicated study; What it leaves open: Single caffeine dose; high CBD dose
• Medicine or class: Alcohol and other sedatives; What the source shows: Alcohol raised CBD's peak by 93 per cent and exposure by 63 per cent; the label says sedatives and alcohol may add to sleepiness; Type of evidence: Dedicated study (alcohol); label warning; What it leaves open: Not measured for retail doses
• Medicine or class: Clopidogrel and other CYP2C19-activated drugs; What the source shows: The label says CBD may reduce the active metabolite and therefore the effect; Type of evidence: Mechanism-based statement; the label gives no clinical data; What it leaves open: Whether any effect is meaningful in practice isn't shown here
• Medicine or class: Strong enzyme inducers (for example rifampicin); What the source shows: Rifampicin lowered CBD by about 32 per cent and one active metabolite by about 63 per cent; Type of evidence: Dedicated study; What it leaves open: Means a medicine can change CBD exposure too

Sources for the table: US prescribing information, sections 7 and 12.3; the European product information, sections 4.4 and 4.5; Geffrey and colleagues, 2015; So and colleagues, 2025; Leino and colleagues, 2019; Cortopassi, 2020; Salcedo and colleagues, 2026; Gorbenko and colleagues, 2026.

Three things follow from the table. First, the strongest evidence involves antiepileptic drugs and transplant medicines, which are high-stakes and often prescribed together with specialist monitoring. Second, almost all quantified findings use prescription CBD or a weight-based dose (2.5 to 12.5 mg/kg twice a day) well above a single drop of oil, so they show direction and mechanism more than a threshold. Third, a medicine missing from the table isn't cleared. A 2021 review of CBD interactions (Balachandran and colleagues, Journal of General Internal Medicine) found interactions reported with antiepileptics, antidepressants, opioid painkillers, alcohol and even paracetamol, while stressing the shortage of controlled trials, and another review (Brown and Winterstein, 2019) judged the potential for interactions with commonly used medicines to be high.

The Dutch national formulary (Farmacotherapeutisch Kompas, undated page read on 8 October 2026) covers prescription cannabidiol in similar terms: a 3- to 4-fold rise in the active clobazam metabolite, more diarrhoea, reduced appetite and rising transaminases with valproate, an interaction expected with CYP2C19 substrates on in-vitro data, and a recommendation to consider monitoring and dose reduction for P-glycoprotein substrates such as sirolimus, tacrolimus and digoxin.

Pregnancy and breastfeeding

For prescription CBD, the European product information says there are only limited data in pregnant women, animal studies have shown reproductive toxicity, and it shouldn't be used in pregnancy unless the benefit clearly outweighs the risk. It adds that there are no clinical data on CBD in human milk, that CBD is highly protein bound and likely passes into milk, and that breastfeeding should be stopped during treatment. The US label likewise reports no adequate data in pregnant women, developmental toxicity in rats and rabbits, and no human data on milk. The Dutch Farmacotherapeutisch Kompas advises against use in pregnancy unless strictly necessary and advises against either using the medicine or breastfeeding. EFSA said in February 2026 that safety couldn't be established for pregnant or breastfeeding women. Those statements are about CBD in general: a retail product that isn't a medicine hasn't been assessed for these groups at all.

Driving, sleepiness and alcohol

Prescription CBD can cause somnolence and sedation; the European text says it can have a major influence on the ability to drive and use machines, and both the US and European texts tell patients to be careful until they know how it affects them. Alcohol can add to sleepiness and also raised CBD exposure in a dedicated study.

A Dutch randomised trial at Maastricht University (Arkell and colleagues, 2020, JAMA) looked at on-road driving after vaporised CBD-dominant cannabis (13.75 mg CBD) in 26 healthy occasional cannabis users. Lane weaving didn't differ significantly from placebo, whereas THC-dominant and THC/CBD-equivalent cannabis impaired it at 40 to 100 minutes. The authors noted that the effect size for CBD-dominant cannabis couldn't exclude a clinically important impairment, and that the doses may not reflect common use. It was vaporised cannabis, not oral oil at higher doses. If a product leaves you drowsy, treat that as a reason not to drive that day and to tell a pharmacist or doctor.

Does the bottle match the label?

CBD products can differ from their labels. A US analysis of 202 products (Gidal and colleagues, 2024) found 74 per cent deviated from labelled potency by at least 10 per cent, and European analyses of retail oils have found THC isomers that the label didn't mention, which matters for anyone who drives for work or is drug-tested. Those samples say why batch evidence matters, not how often a particular brand fails. A seller's reassurance ("tested", "THC-free", "safe for everyone") is a claim: ask for the batch report, check that it matches the pack, and remember that a clean report doesn't tell you whether the product suits your medicines.

Who should ask before using CBD

Speak to a pharmacist or doctor before starting if any of these apply:

• You take prescription medicines, particularly for epilepsy, blood thinning, transplant, depression or pain, or you take several medicines.
• You have liver disease or have been told your liver enzymes are raised. The US label reports higher rates of enzyme rises in people with raised baseline levels and says exposure was 2.5 to 5.2 times higher in moderate or severe hepatic impairment.
• You're pregnant, trying to conceive or breastfeeding.
• You're under 25. EFSA said in February 2026 that safety couldn't be established for people under 25.
• You drive or do safety-critical work, or you're subject to drug testing.

Bring the pack, the batch number, the actual amount and timing, and the full list of everything else you take, including occasional painkillers and sleep aids. If you think a product has caused a problem, keep the packaging and tell your pharmacist or doctor, who can advise on reporting it.

Our guide to how supplements and medicines interact explains how to prepare for that conversation, and how to read supplement labels helps you pull the right details from a pack.

Where to read next

This guide stays with the medicines question. For the rest of the CBD picture:

• CBD: the ingredient, the labels and the questions, for what CBD is, how to read an oil label and where it stands in the EU and the Netherlands.
• CBD oil, capsules and gummies, for what each format changes, the label arithmetic in mg per drop and how to read a batch report.
• CBN and CBG: how do they differ from CBD?, for CBN and CBG, which this guide doesn't cover.
• Is CBD Legal in Europe? EU Law and the Dutch Position, for how EU novel food status, EFSA's 2 mg figure, THC numbers and the Dutch position fit together.

Written by Smart Supplements with AI assistance from the sources linked in the text. It has not been reviewed by a clinician.

Last updated: 12 October 2026

This article is for informational purposes only and is not intended as medical advice. Always consult a healthcare professional before starting any new supplement, especially if you take prescription medication.