CBD or THC: Sorting the Claimed Advantages by What Randomized Trials Have Shown

The most cited evidence base for cannabinoid benefit is the 2015 JAMA systematic review by Whiting and colleagues, covering 79 randomized trials and 6,462 participants. It found moderate-quality evidence supporting cannabinoids for chronic pain and for spasticity, and low-quality evidence for chemotherapy-induced nausea and vomiting, weight gain in HIV infection, sleep disorders, and Tourette syndrome. Almost all of the products in those trials contained THC, either alone or combined with CBD.

CBD’s strongest result is in a different place entirely. Pharmaceutical cannabidiol reduced convulsive seizures in Dravet syndrome in a 2017 New England Journal of Medicine trial, at 20 mg per kilogram per day, a dose roughly one hundred times what a typical consumer tincture delivers.

Most patients choose a product before they choose a target. Someone who believes CBD treats pain, anxiety, sleep, inflammation, and epilepsy equally well will buy a low-dose CBD tincture for arthritis and conclude that cannabis does not work when it fails. A trial of CBD at 20 to 30 mg daily for hand osteoarthritis found a pain difference of 0.23 mm on a 100 mm scale, which is precisely nothing.

The reverse error is also common. Patients avoid THC entirely because of the intoxication, and in doing so avoid the cannabinoid with the strongest evidence for the symptoms they actually have. Low, carefully titrated THC is a legitimate option that a blanket rule against it removes from the table.

The clearest CBD result in medicine is in rare epilepsy. In the 2017 trial by Devinsky and colleagues, 120 children and young adults with Dravet syndrome received pharmaceutical cannabidiol at 20 mg per kilogram per day or placebo added to standard therapy. Median monthly convulsive seizures fell from 12.4 to 5.9 on cannabidiol against 14.9 to 14.1 on placebo, an adjusted median difference of 22.8 percentage points. That result is the basis of a regulatory approval and it is genuinely strong.

It is also specific in a way that consumer writing usually omits. The dose was weight-based and very large, the product was a pharmaceutical-grade solution, the population was a defined epilepsy syndrome, and adverse events included diarrhoea, vomiting, fatigue, somnolence, and abnormal liver function tests, with more withdrawals in the cannabidiol group than in placebo.

For anxiety and mood, the evidence is thinner than the reputation. A 2019 systematic review in Neurotoxicology by Bonaccorso and colleagues screened 1,301 papers, included 27, and concluded that evidence for CBD in psychiatric disorders is currently limited, with potential signals in substance use disorders, chronic psychosis, and anxiety that need large-scale trials. The mechanism most often quoted for CBD and mood, action at serotonin 5-HT1A receptors, comes from preclinical pharmacology rather than from human outcome trials.

Negative CBD trials get far less circulation than positive ones, and one of them is directly relevant to the most common consumer use. Vela and colleagues, publishing in Pain in 2022, randomized 136 patients with hand osteoarthritis or psoriatic arthritis and moderate pain despite therapy to synthetic CBD at 20 to 30 mg daily or placebo for 12 weeks. The between-group difference in pain intensity at 12 weeks was 0.23 mm on a 100 mm scale, with 22 percent on CBD and 21 percent on placebo achieving a 30 mm reduction.

The trial also measured sleep quality, anxiety, depression, and pain catastrophizing as exploratory outcomes. None of them differed significantly from placebo either. That is a well-conducted 12-week trial at a dose range matching what most consumer products deliver, testing the exact indication people buy CBD for, and it found nothing.

The 2019 Lancet Psychiatry review by Black and colleagues reached a similar verdict for mental health. Across 83 studies including 40 randomized trials, the authors concluded there is scarce evidence that cannabinoids improve depressive disorders, anxiety disorders, attention-deficit hyperactivity disorder, Tourette syndrome, post-traumatic stress disorder, or psychosis, and noted that few randomized trials examined pharmaceutical CBD at all.

THC-containing products carry the bulk of the positive evidence. The Whiting review found cannabinoids associated with a greater proportion of patients achieving complete control of chemotherapy-induced nausea and vomiting, 47 percent against 20 percent with an odds ratio of 3.82, an average reduction of 0.46 points on a 10-point pain scale, and an average reduction of 0.36 points on the Ashworth spasticity scale.

The 2021 BMJ meta-analysis by Wang and colleagues, covering 32 trials and 5,174 patients with chronic pain, refined that picture with modern methods. Non-inhaled medical cannabis produced a small improvement in pain relief, with a modelled 10 percent risk difference for achieving the minimally important difference, a very small improvement in physical functioning, and a small improvement in sleep quality. It did not improve emotional, role, or social functioning, and that was a high-certainty finding.

The costs are quantified in the same review. Oral cannabis increased transient cognitive impairment, vomiting, drowsiness, impaired attention, and nausea, each by a few percentage points, and increased dizziness by 9 percent in trials under three months and by 28 percent in trials of three months or longer. Black and colleagues found that pharmaceutical THC with or without CBD increased adverse events with an odds ratio of 1.99 and withdrawals due to adverse events with an odds ratio of 2.78.

The question most patients ask, whether CBD or THC is better for their condition, cannot be answered from most of this literature, because most of the trials did not separate them. Nabiximols, the most studied cannabis-based medicine in pain and spasticity, is a roughly equal combination of THC and CBD delivered as an oromucosal spray. A positive nabiximols trial tells you that the combination works. It does not tell you which component did the work, or whether either alone would have.

Where a trial has separated them, the results are informative. The 2016 Diabetes Care trial by Jadoon and colleagues ran five arms including CBD alone, THCV alone, and two combinations, and found that neither combination arm produced a significant effect on any endpoint while single agents did. The 2023 sleep trial by Bonn-Miller and colleagues tested CBN alone against CBN plus three doses of CBD and found that adding CBD improved nothing.

Those are two of the very few controlled tests of the entourage idea available in humans, and both came out against it. That does not disprove cannabinoid interaction as a phenomenon. It does mean the assumption that adding CBD makes a formulation better is currently unsupported.

The standard framing is that CBD is safe and THC is risky. The direction is right and the magnitude is overstated. CBD inhibits cytochrome P450 enzymes and therefore interacts with a substantial list of medications, and the pivotal epilepsy trial recorded abnormal liver function tests as a treatment-emergent adverse event with more withdrawals in the active arm than in placebo. A 2024 US Food and Drug Administration laboratory study found CBD among the more potent cannabinoids for hepatocyte transcriptional and cytotoxic effects.

THC’s risks are better characterized and largely dose-dependent and reversible: intoxication, cognitive impairment, dizziness, tachycardia, anxiety at higher doses, and in a minority of people an unpleasant acute reaction. The dizziness figure of 28 percent in longer trials is not trivial in an older patient at risk of falls.

It is also worth correcting a terminology error that appears in nearly every consumer comparison. CBD is not non-psychoactive. A compound that reduces anxiety is by definition psychoactive. CBD is non-intoxicating, which is a different and more accurate claim, and preserving that distinction keeps the conversation honest.

Several claims recur in CBD and THC comparisons that have no place in clinical writing. The first is that THC is useful for nausea and vomiting of pregnancy. There is no trial evidence supporting cannabis in pregnancy for any indication, observational data raise concerns about fetal growth and neurodevelopment, and every major obstetric body advises against it. Recommending it is a serious error regardless of how it is hedged.

The second is glaucoma. THC does lower intraocular pressure briefly, but the effect is short and requires dosing frequent enough to be impractical, and glaucoma did not appear among the conditions with supportive evidence in the Whiting review’s conclusions. It is not a treatment.

The third is the euphoric effect of THC framed as an advantage in depression. Black and colleagues found no benefit of cannabinoids for depressive disorders across 23 randomized trials and 2,551 participants, and found that pharmaceutical THC worsened negative symptoms of psychosis in the one study that measured them. Mood elevation lasting a few hours is not an antidepressant effect.

For THC-containing products in chronic pain and spasticity, the evidence is moderate quality and reproducible across two large independent syntheses a decade apart, with effect sizes that are consistently small. Small and real is the accurate description, and it is the description most likely to leave a patient satisfied rather than disappointed.

For CBD, the evidence is bimodal. In defined epilepsy syndromes at pharmaceutical doses it is strong enough to have supported regulatory approval. In chronic pain at consumer doses it has a clean negative trial. In psychiatric indications it is described by two independent systematic reviews as scarce or limited. Treating those three situations as a single body of CBD evidence is the most common error in this area.

For the direct comparison between them, the evidence is weak by design. Most trials used combination products, and the handful of studies that isolated components found no advantage to combining.

The Whiting review is now more than a decade old, and it judged only four of its 79 trials to be at low risk of bias. Its effect estimates should be read as the best available synthesis of a flawed literature rather than as precise values.

Placebo response in cannabinoid pain trials is substantial, and patients often know which arm they are in because THC produces recognizable effects. Unblinding pushes estimates in the direction of benefit, which is a reason to prefer the smaller effect sizes in these reviews over the larger ones reported in open-label and registry work.

Consumer product identity is a separate problem sitting underneath all of it. Trial results describe characterized pharmaceutical preparations. Independent testing of the retail CBD market has repeatedly found label content that does not match the bottle, so a trial dose and a shelf dose are not interchangeable.

Nothing here shows that CBD is useless. It shows that CBD at consumer doses failed in one well-run arthritis pain trial and that psychiatric evidence remains thin. Those are specific findings, not a verdict on the molecule.

Nothing here establishes an optimal ratio. No adequately powered trial has compared several defined THC to CBD ratios head to head for a single indication, which is the study the field most obviously needs.

And none of this addresses long-term outcomes. The pain meta-analysis included trials with follow-up from one to five and a half months. What daily cannabinoid use does over years, in either direction, is not established by this literature.

The CBD versus THC framing took hold during a period when CBD was legally accessible in places THC was not, which gave commercial reasons to describe CBD as capable of nearly everything THC does without the drawbacks. The clinical literature never supported that, and the strongest CBD result turned out to be in a condition almost nobody was buying CBD for.

In practice the useful question is rarely which cannabinoid. It is which symptom, at what dose, by what route, and measured how. A patient with nocturnal spasticity, a patient with chemotherapy nausea, and a patient with generalized anxiety are three different problems, and the cannabinoid answer differs across them in ways that a two-column comparison cannot capture.

The field would also benefit from retiring the entourage assumption as a default. Two controlled human tests of adding CBD to another cannabinoid both found no added benefit. That is not a large evidence base, but it is more than the assumption itself has.

The conversation I have most often starts with a patient telling me they want to try CBD because they do not want to feel high. I understand the instinct completely, and I usually agree with the caution. What I tell them is that the CBD evidence for what they came in with, which is frequently arthritis pain or anxiety, is weaker than they have been led to believe, and that a very small amount of THC may do more for them than a large amount of CBD.

Very small is the part people miss. The doses that work in clinic are often far below what patients expect, and the difference between a helpful dose and an unpleasant one can be a couple of milligrams. That is a titration problem, not an argument for avoiding THC.

I also correct the safety framing every time. CBD interacts with a long list of medications through the same liver enzymes that handle much of what my patients take. It is not a harmless supplement, and treating it as one is how people get into trouble with their anticoagulants and their antiepileptics.

THC-containing products carry moderate-quality evidence for chronic pain and spasticity and lower-quality evidence for nausea, appetite, and sleep, with small effect sizes and real dose-dependent side effects. CBD has strong evidence in specific epilepsy syndromes at pharmaceutical doses and a clean negative trial in arthritis pain at consumer doses. Choose by symptom target and dose rather than by which cannabinoid has the better reputation, and tell your prescriber, because CBD in particular interacts with common medications.

When a comparison page lists advantages, check whether each one names a trial, a population, and a dose. CBD for epilepsy names all three. CBD for inflammation, skin health, or depression usually names none. The difference between those two kinds of entry is the entire value of a comparison, and most published comparisons do not mark it.