CBD and arthritis: what the trials actually found

What the research has measured

The strongest human evidence in arthritis specifically is a Danish randomised, double-blind, placebo-controlled trial published in Pain in 2022. One hundred and thirty-six people with hand osteoarthritis or psoriatic arthritis, all still in pain despite their existing therapy, were randomised to CBD or placebo for twelve weeks. One hundred and twenty-nine were included in the primary analysis.

The difference in pain intensity between the two groups at twelve weeks was 0.23 mm on a 100 mm scale, with a 95% confidence interval from −9.41 to 9.90 and a p-value of 0.96. In plain terms: the CBD group and the placebo group were indistinguishable. The trial also found no significant difference in sleep quality, anxiety, depression or pain catastrophising scores.

That is one trial, in two specific conditions, at one dose schedule. It is not the last word on the subject. It is, however, the most rigorous thing anyone has published about CBD and arthritis, and it points the opposite way to most of what is written about the topic online.

Where animals and humans part company

The optimistic case for CBD in arthritis rests largely on animal work, and the most cited piece is a 2016 study in the European Journal of Pain. Rats with induced arthritis in one knee were given transdermal CBD gel for four days. Joint swelling fell, limb posture scores improved, immune cell infiltration and synovial membrane thickening reduced, and inflammatory markers in the spinal cord and dorsal root ganglia dropped — all dose-dependently, with 6.2 mg and 62 mg per day the doses that did it.

That is a real result and it is why researchers keep looking. It is also a rat, with a chemically induced arthritis, dosed through the skin at the joint, over four days. Three things make it a poor guide to what a capsule does in a person:

• The model is not the disease. Chemically induced monoarthritis in a rat knee runs a different course from rheumatoid arthritis or osteoarthritis in a human joint.
• The route is not the same. A gel applied over the joint puts CBD where it is wanted. A swallowed capsule goes through the liver first, and only a fraction of the dose reaches circulation.
• The doses do not translate. Milligrams per rat per day do not convert to milligrams per person by simple arithmetic.

This is the ordinary shape of drug research, not a scandal. Most things that look promising in animals do not survive contact with a randomised human trial. What is unusual here is how rarely the distinction is made in the material sold alongside the products.

What the Arthritis Foundation says

In September 2019 the Arthritis Foundation published the first CBD guidance aimed at adults with arthritis, after consulting rheumatologists. Their summary is blunt: there is no good evidence that CBD helps arthritis of any type, the material suggesting it does is anecdotal, and CBD should not be used in place of a disease-modifying antirheumatic drug.

Their practical advice, which we would repeat:

• Talk to your rheumatologist before starting, and tell them if you have already started.
• Do not stop or reduce an existing therapy in favour of CBD.
• Buy from a company that will show you a batch certificate of analysis.
• Start low, go slow, and keep a written note of what you took and what changed.

That last point is worth more than it sounds. Arthritis pain fluctuates on its own. Without a record kept over weeks, it is very hard to tell a change caused by something you took from the ordinary movement of the condition — which is precisely what a placebo-controlled trial exists to do, and precisely what an individual cannot do from memory.

Reading a label without being taken in

There is a second problem sitting underneath the evidence question: a CBD product may not contain what its label says. A 2017 analysis published in JAMA bought 84 CBD extracts online and tested them. Around seven in ten were inaccurately labelled — some containing substantially less CBD than stated, some substantially more, and some containing measurable THC that the label did not mention.

Four things on a label are worth more than the rest of the packaging combined:

That last row matters more than people expect. A product that pairs CBD with something else, and then works, tells you nothing about which ingredient did it.

What this page will not tell you

It will not give you a dose. There is no established dose of CBD for arthritis, the trials that exist used schedules chosen for the trial rather than derived from evidence of what works, and suggesting a number here would give a false impression that one is known.

It will not tell you CBD is a substitute for anything your rheumatologist has prescribed. The Arthritis Foundation is explicit on that point and so are we.

It will not tell you it is inert either. People report that it helps them, in numbers large enough that researchers keep running trials. Reported experience is real information; it is simply a weaker kind of information than a controlled trial, and it belongs in a different column.

Trying it properly

Plenty of people will read the trial result and want to try it regardless. That is a reasonable position — a null result in a group of 129 does not mean nothing happens in any individual. If that is you, the way to get an answer worth having is to borrow the structure of the trial you just read about.

• Change one thing. Not a new supplement the same week, not a new exercise routine. One variable, or the result tells you nothing.
• Write the number down daily. Pain out of ten, at roughly the same time of day, starting two weeks before you begin. Without that baseline there is nothing to compare against, and arthritis fluctuates enough that memory is not a reliable substitute.
• Give it long enough. The trial ran twelve weeks. A fortnight tells you about this fortnight.
• Stop, and watch. The most informative thing most people never do. If the numbers drift back after stopping and recover when restarting, that is worth far more than any amount of remembering.

Done that way, a few months produces something a rheumatologist can read. Done any other way, it produces an impression, and an impression is what the whole field is already drowning in.

The products this page covers