Relief Is Not the Same as Slowing a Disease: What Cannabis Trials Have Actually Measured

Two different claims get blurred together in cannabis conversations. One is symptom relief: pain, spasticity, nausea, sleep, appetite. The other is disease modification: changing the biological course of an illness so that the underlying damage accumulates more slowly. These are separate outcomes, tested with separate measures, and cannabis has a far better record on the first than on the second.

The clearest test came from the CUPID trial, which gave oral dronabinol or placebo to 493 adults with progressive multiple sclerosis for up to three years and asked whether the drug slowed disability progression. It did not. The hazard ratio for progression on the Expanded Disability Status Scale was 0.92, with a confidence interval spanning 0.68 to 1.23.

Patients make real decisions on this distinction. Someone who believes cannabis is slowing their multiple sclerosis may reason differently about starting or continuing a disease-modifying therapy that has actually been shown to do that. That is a consequential misunderstanding, and it is one the cannabis field has not been careful enough about.

The honest version of the message is still useful. Symptom relief is a legitimate clinical goal. Sleeping through the night, eating a meal, or moving without spasm changes a person’s life. It is simply a different claim from altering the disease, and patients deserve to know which one the evidence supports.

In clinical research, a symptomatic treatment changes how a patient feels or functions while the treatment is present. A disease-modifying treatment changes the biology underneath, so that the damage accumulates more slowly even after accounting for symptom control. The two are tested differently. Symptom trials run weeks and measure pain scores, spasticity scales, or seizure counts. Disease-modification trials run years and measure disability accrual, imaging change, tissue healing, or organ damage.

This is not a technicality. In multiple sclerosis, the disease-modifying therapies are a specific drug class held to a specific evidentiary standard: relapse rates, lesion burden on MRI, confirmed disability progression. In rheumatoid arthritis, the equivalent standard includes radiographic joint damage. A drug that reduces pain without touching those endpoints is a good analgesic, not a disease-modifying agent.

When a cannabis article says a cannabinoid may slow a disease, the fair question is which endpoint was measured. Most of the time the answer is that no such endpoint was measured at all.

The Cannabinoid Use in Progressive Inflammatory brain Disease trial, known as CUPID, was designed specifically around the neuroprotection hypothesis. Laboratory work had suggested cannabinoids might protect neurons, so investigators at 27 UK centres randomized 498 adults with primary or secondary progressive multiple sclerosis, in a 2 to 1 ratio, to oral dronabinol at up to 28 mg per day or matching placebo, for 36 months. The primary outcomes were time to confirmed progression on the Expanded Disability Status Scale and change in the physical impact subscale of the 29-item Multiple Sclerosis Impact Scale.

The result, published in Lancet Neurology in 2013 by Zajicek and colleagues, was negative on both. The hazard ratio for disability progression was 0.92 with a 95 percent confidence interval of 0.68 to 1.23, and the between-group difference on the physical impact subscale was 0.9 points in favour of dronabinol with a confidence interval crossing zero. Secondary clinical and MRI outcomes showed no significant treatment effect either.

The authors of the fuller Health Technology Assessment report, published in 2015, put it plainly: there was no clear symptomatic or disease-modifying treatment effect. They noted an important caveat, that participants deteriorated more slowly than expected, which reduces the power to detect a difference. That caveat is a reason to be humble about a strong negative conclusion. It is not a reason to claim a positive one.

Inflammatory bowel disease offers the cleanest illustration, because gastroenterology routinely measures both how a patient feels and what the bowel looks like. In a 2021 randomized, double-blind, placebo-controlled trial published in the Journal of Crohn’s and Colitis, Naftali and colleagues gave 56 patients with active Crohn’s disease either CBD-rich cannabis oil at 160 mg CBD and 40 mg THC per millilitre, or placebo, for eight weeks.

Symptoms improved substantially. The Crohn’s Disease Activity Index fell from a median of 282 to 166 in the cannabis group, against 264 to 237 on placebo. Quality of life scores rose from 74 to 91 with cannabis and barely moved on placebo. Those are meaningful patient-level gains.

The endoscopic picture did not follow. The Simple Endoscopic Score for Crohn’s Disease was 10 before and 7 after in the cannabis arm, and 11 before and 8 after on placebo, with no significant difference between groups. C-reactive protein and calprotectin, the standard inflammatory markers, were unchanged. The authors concluded that cannabis treatment in Crohn’s disease should be used only in the context of clinical trials until more is known. An earlier 2013 trial by the same group, using THC-rich cannabis cigarettes, produced the same shape of result: clinical response in 10 of 11 patients, but the remission endpoint was not met.

A 2018 randomized trial of a CBD-rich botanical extract in ulcerative colitis, reported by Irving and colleagues in Inflammatory Bowel Diseases, also missed its primary endpoint, with remission rates of 28 percent on the extract and 26 percent on placebo.

Rheumatoid arthritis is often cited as a case where cannabis reduces inflammation and therefore joint damage. The evidence behind that claim is a single 58-patient, five-week trial of nabiximols published by Blake and colleagues in Rheumatology in 2006. It found significant improvement in pain on movement, pain at rest, sleep quality, and the DAS28 composite disease activity score. The authors themselves described the differences as small and variable across the population and called for more detailed investigation.

No radiographic outcome was measured, which is the standard endpoint for structural joint damage, and in the two decades since, no adequately powered follow-up trial has tested whether cannabinoids change joint destruction in rheumatoid arthritis. A 2026 real-world study by Kruger and colleagues in Clinical Therapeutics found symptom improvement across fibromyalgia, rheumatoid arthritis, and osteoarthritis with several cannabinoid combinations, but it was a self-report study of symptoms, not of joint structure.

Huntington’s disease is a useful counterexample because the endocannabinoid rationale there was strong on paper. A double-blind, randomized, placebo-controlled crossover trial of nabiximols in 26 patients, reported by Lopez-Sendon Moreno and colleagues in the Journal of Neurology in 2016, found the treatment safe and well tolerated, with no significant effect on motor, cognitive, behavioural, or functional scores, and no significant change in the molecular biomarkers the investigators tracked.

There is one place where a cannabinoid moves a hard disease-activity endpoint rather than a subjective one. In the 2017 New England Journal of Medicine 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. 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.

Seizure frequency is a directly observed disease-activity measure, not a symptom rating, and the finding supported a regulatory approval. It is still worth being precise about what it shows. Reducing seizures is not the same as correcting the underlying sodium channel mutation that causes Dravet syndrome, and the trial did not test whether cannabidiol alters the long-term neurological trajectory of the condition.

That is the most favourable case in the literature, and even there the honest description is strong control of disease activity rather than modification of the disease itself.

The evidence that cannabis relieves specific symptoms is moderate and reasonably consistent. The 2015 JAMA systematic review by Whiting and colleagues, covering 79 trials and 6,462 participants, found moderate-quality evidence for chronic pain and spasticity, with an average reduction of 0.46 points on a 10-point pain scale and 0.36 points on the Ashworth spasticity scale. The 2021 BMJ meta-analysis by Wang and colleagues, covering 32 trials and 5,174 patients, reached a similar conclusion: a small improvement in pain, a very small improvement in physical function, and a small improvement in sleep.

The evidence on disease modification is a different matter. It consists mainly of one adequately powered negative trial, several small trials that measured symptoms only, and a large volume of laboratory and animal work that has not been translated. Preclinical neuroprotection findings are the reason CUPID was funded and run. The trial is what turns a hypothesis into an answer, and the answer was negative.

CUPID has a genuine limitation that its own authors flagged. Participants were more disabled at entry than in earlier studies and progressed more slowly than the sample size calculation assumed, which reduces the ability to detect a real but modest effect. A cleanly negative trial with reduced power is weaker evidence of absence than a cleanly negative trial with full power.

The Crohn’s disease trials are small, single-centre, and eight weeks long. Eight weeks is a short window in which to expect mucosal healing, and a negative endoscopic result over that period does not prove that longer exposure would do nothing. The honest reading is that the trials as designed did not show it, not that the question is closed.

There is also publication asymmetry to consider. Negative cannabinoid trials are reported, but they circulate far less in patient-facing writing than positive preclinical findings do, which distorts the impression a reader gets from a casual search.

None of this shows that cannabis accelerates any disease, and none of it argues against using cannabis for symptom control where symptom control is the goal. The CUPID trial recorded serious adverse events in 35 percent of the dronabinol group and 28 percent of the placebo group, with no major safety signal identified.

It also does not show that cannabinoids can never be disease-modifying in any condition. Most diseases have never been studied this way with cannabinoids at all. Absence of a trial is not a negative trial, and both should be described as what they are.

Finally, the Dravet syndrome result does not generalize. Pharmaceutical cannabidiol at 20 mg per kilogram per day in a specific epilepsy syndrome tells you very little about a CBD tincture taken for arthritis.

The disease-modification claim has a particular hold in cannabis writing because the endocannabinoid system touches inflammation, neuronal signalling, and immune regulation, so a mechanistic story is easy to construct for almost any condition. Mechanism is a reason to run a trial. It is not a substitute for one.

The broader clinical conversation has been moving toward precision on this point. Systematic reviews now consistently describe cannabinoids in terms of symptom outcomes, effect sizes, and adverse events rather than in terms of disease course. That is a maturation of the field, not a retreat from it.

There is also a practical reason to hold the line. Cannabis medicine gains credibility with colleagues by being accurate about its limits. Overclaiming on disease modification costs the field more than the honest, narrower claim ever would.

This question comes up constantly, and I understand why. When something helps you this much, it is natural to hope it is doing more than helping you feel better. I have watched patients sleep again, eat again, and move again on a carefully titrated regimen. That is real and it matters.

What I will not tell a patient is that cannabis is slowing their multiple sclerosis or repairing their joints, because the trials that looked for that did not find it. The most dangerous version of this misunderstanding is the patient who quietly stops a proven disease-modifying therapy because the cannabis makes them feel better. Feeling better and getting better are not the same thing, and in some diseases the difference is measured in years of function.

So I say it directly. We are treating your symptoms. We are treating them well, and we are going to keep measuring whether it is working. Your neurologist is treating your disease. Both jobs matter, and neither one replaces the other.

Cannabis has reasonable evidence for relieving pain, spasticity, nausea, and sleep disturbance, and poor evidence for slowing the course of any disease. The one adequately powered trial designed to test progression, CUPID in progressive multiple sclerosis, was negative. Use cannabis for the symptom target it has been shown to help, keep disease-modifying therapy separate, and tell your prescribing clinicians about both.

When you read that a cannabinoid may slow a condition, look for the endpoint. Disability scales measured over years, imaging, endoscopy, radiographs, and biomarkers are progression measures. Pain scores, symptom indices, and quality of life questionnaires are not. If the article does not name a progression endpoint, the study almost certainly did not measure one.