cbdlovers
EN

CBD for pain: two 2024 reviews, opposite answers

cbdlovers editorial team 14 min

In April 2024, two systematic reviews of cannabidiol and pain were published. One reported pain reduction of 42 % to 66 %. The other reported that 15 of 16 randomised trials showed no benefit over placebo.

Same question. Same month. Same publisher, as it happens. Opposite conclusions.

Google’s own AI answer for this search cites both of them, side by side, without mentioning that they disagree.

Most pages on this subject pick one and ignore the other. That is the wrong move, because the two results are not actually in conflict — and the reason they diverge is the single most useful thing anyone can learn about pain research. Once you see it, you can read every pain claim you will ever encounter, about any product at all.

The two papers

The negative one

Cannabidiol (CBD) Products for Pain: Ineffective, Expensive, and With Potential Harms, published in The Journal of Pain in April 2024 by Moore, Straube, Fisher and Eccleston, with the Centre for Pain Research at the University of Bath among the affiliations.

Its findings, in its own terms:

Its stated perspective is that there is no good reason to think cannabidiol relieves pain, and good reasons to doubt what is in the products.

The positive one

Effectiveness of Cannabidiol to Manage Chronic Pain: A Systematic Review, published in Pain Management Nursing in April 2024 by Mohammed and colleagues at Adamson University in Manila, with a co-author at the University of Connecticut School of Medicine.

Its findings:

Both papers are real systematic reviews. Neither is fringe. Neither author group is selling anything.

Why they disagree, in one sentence

Forty-two to sixty-six per cent is how much people’s pain scores fell. Fifteen of sixteen is how often that fall was bigger than the fall in the placebo group.

Both numbers can be true at the same time, and in this literature they are. People taking cannabidiol do report substantially less pain than before they started. So do the people taking the placebo.

The comparator is everything

A within-person change asks: how much better is this person than they were? A controlled comparison asks: how much better is this person than an identical person who got nothing?

The first question is easy to answer and almost always produces an encouraging number. The second is expensive, slow, and frequently produces nothing. Almost every impressive statistic in consumer health is an answer to the first question presented as though it answered the second.

Which studies each review let in

The Bath-affiliated review restricted itself to randomised trials using verified, pharmaceutical grade cannabidiol. The Manila review used narrative synthesis across “various study designs” and measured outcomes “mostly through self-reporting” on visual analogue and verbal numerical scales.

Those are different inclusion criteria, honestly applied, producing different answers. Neither team did anything wrong. They asked adjacent questions and got the answers those questions deserve.

And the THC problem, again

The Manila review states plainly that its pain reduction figures come from studies of cannabidiol alone and cannabidiol with tetrahydrocannabinol. This is the same pattern that shows up in the sleep literature: the evidence base for “cannabinoids” is being read as evidence about cannabidiol, and the products people can legally buy contain a THC ceiling designed to make THC irrelevant.

What the sixteen trials actually covered

The negative finding is stronger than a single study and weaker than a settled verdict, and the detail of what was tested explains both halves.

Twelve different pain states

The sixteen trials were spread across twelve pain conditions. That is close to one condition per trial, and pain conditions are not variations on a theme — neuropathic pain arises from damaged nerves, inflammatory pain from an immune response, nociplastic pain from altered central processing. They respond to entirely different medicines.

A compound could work well in one of those and do nothing in the others, and a set of trials spread this thinly would struggle to detect it either way.

Three routes and a hundredfold range of quantity

The trials used oral, topical and buccal or sublingual administration, with quantities from 6 mg to 1,600 mg. That range spans more than two orders of magnitude, and the routes differ in how much reaches the bloodstream by a factor of five or more.

Six milligrams swallowed and 1,600 mg sublingually are not the same intervention in any meaningful sense. Grouping them is not sloppiness — it is what you do when sixteen trials is all there is.

Single administrations up to twelve weeks

Durations ranged from one administration to three months. A single administration answers a question about acute effect; twelve weeks answers a question about a chronic condition. Both are legitimate, and they are not the same study.

The one that was positive

One of the sixteen did show a benefit. The review does not identify it in the abstract, and it would be dishonest to guess which. What can be said is that one positive result in sixteen is roughly what chance produces when a series of trials tests something inert at conventional significance thresholds — which is precisely why the count matters more than any single trial in it.

Why this is still the better evidence

Every one of those weaknesses applies with greater force to the uncontrolled studies. Thin, heterogeneous randomised evidence beats abundant uncontrolled evidence, because the control group is the part that does the work. That is the argument for taking the sixteen seriously even while acknowledging how thin they are.

Where cannabinoid pain evidence does exist

Being fair here means saying what has cleared a regulatory bar, because something has — and its composition is instructive.

Nabiximols contains THC

Nabiximols, sold as Sativex, is an oromucosal spray authorised in a number of countries for spasticity in multiple sclerosis. It contains roughly equal quantities of THC and cannabidiol, extracted from cannabis, and it is a prescription medicine.

It is the cannabinoid preparation that has been through regulatory files in the pain-adjacent space, and it is not a cannabidiol product. The compound doing the regulated work is present alongside one that is capped out of every legal consumer product in Europe.

What the IASP said in 2021

The International Association for the Study of Pain convened a task force on cannabinoids and pain and, on cannabidiol specifically, found no trials to assess. The organisation declined to endorse general use of cannabinoids for pain on the evidence available.

That is the professional body for pain research declining to recommend, three years before the sixteen trials existed.

Why the distinction keeps mattering

Across sleep, anxiety and now pain, the same structure appears: the studies with the more convincing results tend to contain THC, and the products people can legally buy are defined by a THC ceiling. Any argument that leans on the cannabis literature to sell a cannabidiol product has to explain that gap, and none of them do.

Chronic pain has the largest placebo response in medicine

This is why the distinction above matters more here than almost anywhere else.

What the placebo arm does

In chronic pain trials, placebo arms routinely show large improvements — often clinically meaningful ones, sustained over weeks. This is not participants being fooled. Pain is generated by a nervous system that responds to context, expectation, attention and reassurance, and a trial supplies all four.

Why the response has been growing

There is a documented tendency for placebo responses in pain trials to increase over time, particularly in North America, which has made it progressively harder for genuinely effective analgesics to demonstrate superiority. This is a known headache for pain researchers and it has nothing to do with cannabidiol specifically.

What that does to an uncontrolled study

If the placebo arm improves by 50 %, then an uncontrolled study of anything at all will report roughly 50 % improvement. Water would. Which is exactly why a 42–66 % figure without a comparison group tells you about the setting, not about the compound.

What “no benefit over placebo” does not mean

Being fair to the positive review requires being precise about what the negative one establishes.

It does not mean nothing happened

People in those trials genuinely felt better. Their pain scores moved. That relief was real and it was experienced by real bodies. What the trials show is that it was not caused by the cannabidiol.

It does not mean the question is closed

Sixteen trials across twelve different pain states is thin. Neuropathic pain, inflammatory pain, fibromyalgia and post-surgical pain are different mechanisms, and pooling them is a compromise made because there is not enough data to do otherwise. A properly powered trial in one specific pain state could still change the picture.

It does mean the burden has moved

Sixteen attempts with pharmaceutical grade material, and fifteen negative, is a meaningful body of evidence. Anyone claiming an effect now has to explain why fifteen trials missed it.

The other half of the Bath paper: what is in the bottle

The review’s most concrete finding is not about efficacy at all, and it survives regardless of which side of the efficacy argument you land on.

Content from none to much more than advertised

Testing of products on sale in North America and Europe found cannabidiol content ranging from zero to substantially more than the label claimed. Not a rounding error — products with none of the compound they were named after.

Other chemicals present

The review notes potentially harmful additional chemicals in products, and serious harm reported in children, adults and the elderly associated with them. Synthetic cannabinoids and residual solvents are the usual culprits in this category.

The regulatory acknowledgement

In January 2023 the US Food and Drug Administration announced that a new regulatory pathway for cannabidiol was needed, which is an agency saying out loud that the existing framework does not cover what is being sold.

What this means for the efficacy debate

It undermines a large part of the positive literature too. A study of a retail product that turns out to contain no cannabidiol is not a study of cannabidiol, whatever its result. This is the strongest argument for taking the verified-material trials more seriously than the rest.

The safety picture, stated fairly

The reassuring part

The WHO’s critical review concluded that cannabidiol shows no effects indicative of abuse or dependence potential in humans, and small clinical trials using verified material suggest the compound itself is largely benign.

The less reassuring part

The Bath review notes growing evidence linking cannabidiol to increased rates of serious adverse events and hepatotoxicity — liver injury. This is not speculative: liver enzyme monitoring is a standard requirement in the prescription cannabidiol medicine authorised in Europe.

And the food regulator’s position

EFSA’s provisional safe intake for cannabidiol as a novel food, published on 9 February 2026, is 0.0275 mg per kilogram of body weight per day, roughly 2 mg a day for a 70 kg adult, and it explicitly does not apply to anyone under 25, pregnant, breastfeeding, or taking medication.

Almost every product sold for pain exceeds that in a single serving.

Topicals are a separate question

Why creams are not covered by the above

Most of the trials in both reviews involve oral or sublingual administration. A cream applied to a knee is a different pharmacological event, and the systemic evidence does not transfer to it either positively or negatively.

What is plausible locally

Cannabinoid receptors exist in skin, and a topical could in principle act where it is applied without meaningful systemic absorption. That is a reasonable hypothesis with limited human evidence behind it.

What is doing the work in most creams

Menthol, camphor and capsaicin produce immediate, reliable sensory effects and appear in a large share of cannabidiol topicals. As with melatonin in sleep gummies, if the cream contains menthol, what you feel in the first minute is the menthol.

What this means if you are in pain

The honest summary

There is no good randomised evidence that cannabidiol reduces pain more than placebo. There is consistent evidence that people who take it report feeling better, which is worth something, and which you can obtain more cheaply and with less risk.

It is expensive

The Bath paper puts cost in its title deliberately. A monthly habit at typical retail prices is a meaningful sum for a benefit that fifteen of sixteen trials could not detect.

It can delay other things

The real cost of an ineffective option is often not the money but the months. Chronic pain has treatments with genuine evidence behind them, several of which are not drugs at all, and a pain clinic is a better destination than a checkout.

If you try it anyway

Buy something with a batch certificate you can actually read, given that content has been found to range from none to far more than advertised. Use a numerical scale weekly rather than a memory. Define in advance what would count as failure. And tell whoever prescribes your other medication, because cannabidiol inhibits CYP3A4 and CYP2C19 and this population is usually taking something else.

How to use this page on any other claim

The two questions

Whenever you meet a health statistic, ask: compared with what? and who was in the comparison group? Those two questions resolve most disagreements in consumer health, and they are the entire content of the disagreement between these two reviews.

The pattern to recognise

A large percentage improvement with no control group is not evidence of effect. It is evidence that people who seek help tend to improve — which is true, important, and says nothing about the product they bought on the way.

Where to look on any study you are shown

Three lines answer nearly everything. What was the comparison group — placebo, another medicine, or nothing at all? How many people — forty, or four hundred? How long — one administration, or long enough to matter for a condition measured in years?

Those three are usually in the abstract, which is free to read on PubMed for essentially every paper cited in consumer health writing. Fifteen seconds of checking is enough to sort most claims, and it costs nothing.

Why we wrote it this way

There was an easier version of this page. Pick the review that suits a commercial interest, cite it confidently, and leave the other one out — nobody would have noticed, because the two papers rarely appear together.

The problem is that a reader who follows that advice and gets nothing has been misled by omission, and a reader who reads both and understands the reason for the gap has learned something they can use for the rest of their life. The second is worth more, including to us, and it is the only version worth publishing.