Parkinson’s, Alzheimer’s, Huntington’s and ALS: What Cannabis Has Actually Been Tested For

Parkinson’s disease, Alzheimer’s disease, Huntington’s disease and amyotrophic lateral sclerosis get grouped together as neurodegenerative conditions, but they have almost nothing in common at the level of what cannabis has been asked to do. Each has its own small set of randomized trials, and those trials point in different directions.

The honest summary across the four is this: one reasonably clean positive randomized signal, in spasticity in motor neuron disease, and a series of null results everywhere else on the endpoints that were pre-specified.

When a family hears that cannabis is neuroprotective, they are usually hearing a summary of cell culture and rodent work. That work is real, and it is why these trials were funded. It is not the same as evidence that a person with Parkinson’s disease will move better or a person with Alzheimer’s disease will be calmer.

Knowing which endpoint a trial measured changes the conversation. A drug can fail on motor scores and still be worth discussing for sleep or anxiety. A drug can improve a spasticity scale and still do nothing for the underlying disease. Patients deserve that level of precision before they spend money on products.

Neurodegenerative disease is a category, not a condition. Alzheimer’s disease is a disorder of cortical networks and memory. Parkinson’s disease is a disorder of dopaminergic signaling and movement. Huntington’s disease is a single-gene disorder producing chorea, cognitive change, and psychiatric symptoms. Amyotrophic lateral sclerosis destroys motor neurons and spares much of cognition in most patients. Grouping them under one heading makes sense for a research funder and very little sense for a patient asking what cannabis might do for them.

The trials reflect that. In Parkinson’s disease, cannabinoids have mostly been tested against motor scores, dyskinesia, and non-motor symptoms such as pain. In dementia, against agitation and neuropsychiatric symptoms. In Huntington’s disease, against the Unified Huntington’s Disease Rating Scale. In motor neuron disease, against spasticity. These are four different questions with four different answers.

None of these trials was designed to test whether cannabis slows the disease itself. That is a separate claim with a separate evidence base, covered in detail in a companion piece on this site.

Parkinson’s disease has the longest run of randomized cannabis trials, and the pattern is consistent. In 2004, Carroll and colleagues published a randomized, double-blind, placebo-controlled crossover trial of oral cannabis extract for levodopa-induced dyskinesia in Neurology. Nineteen patients were randomized and seventeen completed. The extract was well tolerated and produced no improvement on the dyskinesia items of the Unified Parkinson’s Disease Rating Scale, and no improvement on any secondary measure, including the Rush scale, the Bain scale, and patient-reported daily living scores.

In 2014, Chagas and colleagues published an exploratory double-blind trial in the Journal of Psychopharmacology assigning 21 patients to placebo, cannabidiol 75 mg per day, or cannabidiol 300 mg per day. There were no significant differences in UPDRS scores, plasma brain-derived neurotrophic factor, or magnetic resonance spectroscopy measures. Total scores on a quality of life questionnaire differed between placebo and the 300 mg group at a p value of 0.05, which the authors themselves framed as a possible effect requiring larger and more targeted study.

The most recent and largest entry is a 2026 phase II randomized placebo-controlled trial published in Movement Disorders by Kubota and colleagues, testing an oral cannabis extract in people with Parkinson’s disease and chronic pain over nine weeks. One hundred and one participants enrolled and 87 completed. The primary endpoint, the Parkinson’s Disease Pain Classification System score, showed no difference between groups, with a p value of 0.757. No other non-motor scale separated either. No serious adverse events were reported.

Set against that, an open-label dose-escalation study of pharmaceutical cannabidiol in Parkinson’s disease by Leehey and colleagues in Cannabis and Cannabinoid Research reported improvement in total and motor UPDRS scores in the ten participants who completed. That study had no control group, thirteen participants, and a meaningful safety finding: liver enzyme elevations in five of the participants taking 20 to 25 mg per kilogram per day, and three withdrawals for intolerance.

No cannabis trial in dementia has ever targeted memory or cognitive decline as a treatment endpoint. The trials target neuropsychiatric symptoms, principally agitation, which is a leading reason families seek help and a leading driver of nursing home placement.

The clearest randomized result comes from van den Elsen and colleagues in Neurology in 2015. Fifty patients with dementia and clinically relevant neuropsychiatric symptoms received oral tetrahydrocannabinol 1.5 mg three times daily or matched placebo for three weeks. Symptoms improved in both arms. The difference between them on the Neuropsychiatric Inventory was 3.2 points, with a 95 percent confidence interval from minus 3.6 to 10.0, which crosses zero. Agitation, quality of life, and activities of daily living showed no separation either. The paper was graded as Class I evidence that low-dose oral THC does not significantly reduce neuropsychiatric symptoms at 21 days.

The authors were careful to note what the trial did establish: 4.5 mg of THC daily was well tolerated in a frail older population, with no effect on vital signs, weight, or episodic memory. That is a real finding. It is the basis for arguing that higher doses deserve testing, not the basis for claiming efficacy at this one.

The broader dementia literature, including later trials of THC and CBD oil and a recent meta-analysis, is covered separately on this site and does not change the direction of this summary.

Huntington’s disease has strong preclinical rationale for cannabinoid intervention. Cannabinoid receptor loss in the striatum is one of the earliest measurable changes in the disease, which made receptor stimulation an obvious thing to try.

The trial that tried it was published by Lopez-Sendon Moreno and colleagues in the Journal of Neurology in 2016. Twenty-six patients were randomized in a double-blind, placebo-controlled crossover design to nabiximols oromucosal spray, up to twelve sprays per day, or placebo for twelve weeks. Twenty-four completed. The primary objective was safety, and on that count the drug did well: no serious adverse events and no clinical worsening.

On every efficacy measure, the trial was flat. Motor scores differed at p equal to 0.286, cognitive at 0.824, behavioral at 1.0, and functional at 0.581. The biomarker analysis showed no significant molecular changes. The authors concluded that future designs should consider higher doses, longer exposure, or different cannabinoid combinations. That is a fair reading of a pilot trial that ruled nothing in.

No larger randomized trial has replaced it. Anyone telling a Huntington’s family that cannabis helps chorea is going beyond what has been shown.

The CANALS trial, published by Riva and colleagues in The Lancet Neurology in 2019, is the strongest randomized evidence in this entire field. It was an investigator-initiated, multicenter, double-blind, placebo-controlled phase 2 trial at four Italian motor neuron disease centers. Sixty participants with amyotrophic lateral sclerosis or primary lateral sclerosis and established spasticity were randomized to nabiximols oromucosal spray or placebo for six weeks, on top of a stable antispasticity regimen.

Fifty-nine were analyzed. Modified Ashworth Scale scores improved by a mean of 0.11 in the nabiximols group and worsened by a mean of 0.16 on placebo, giving an adjusted effect estimate of minus 0.32, with a 95 percent confidence interval from minus 0.57 to minus 0.069 and a p value of 0.013. No participant withdrew during the double-blind phase and no serious adverse events occurred.

Read this carefully, because the framing matters. This was a six-week phase 2 trial in 59 people, testing an add-on treatment for a symptom, with an effect size that is modest on a scale that is itself imperfect. The authors described it as a proof-of-concept result requiring confirmation in larger trials. It is not evidence that cannabinoids affect the course of ALS, and the trial made no such claim.

Whatever the trial data say, the practical constraints in this population are real and they cut in one direction: caution. People with neurodegenerative disease are usually older, frequently on several central nervous system medications, and often already at risk of falls, orthostatic hypotension, confusion, and swallowing difficulty. Adding a sedating compound to that picture is not a neutral act.

Hepatic monitoring deserves specific mention. The open-label Parkinson’s study using pharmaceutical cannabidiol at 20 to 25 mg per kilogram per day found liver enzyme elevations in five of thirteen participants, mostly cholestatic in pattern. High-dose cannabidiol has a known interaction profile with hepatic enzymes and with commonly used anticonvulsants, and the doses used in research are far above what most retail products deliver.

Cognitive effects are the other constraint. In a patient with existing cognitive impairment, even modest THC-related slowing may be more consequential than it would be in a healthy adult. That is a reason to favor the lowest workable dose, to change one variable at a time, and to involve the treating neurologist rather than working around them.

These are mostly small trials, and small trials are the wrong instrument for detecting modest effects. The Parkinson’s dyskinesia trial had 17 completers. The Huntington’s trial had 24. The dementia trial had 50. A null result in a trial of that size does not prove absence of effect; it proves that any effect present was not large enough to emerge from that sample.

What raises confidence is the consistency. Independent teams, in different countries, across different diseases, using different preparations and different endpoints, mostly did not find what they were looking for. The 2026 Parkinson’s trial is the largest of the set and reported the flattest result. Repeated nulls in the same direction carry more weight than any one of them alone.

The CANALS spasticity result is the exception and it is a legitimate one: multicenter, double-blind, placebo-controlled, with a pre-specified primary endpoint that was met. It is also a phase 2 trial of 59 people over six weeks, which is exactly why its own authors asked for larger confirmation.

Dose is the most defensible criticism of the null trials. Low-dose THC at 4.5 mg daily in the dementia trial, twelve sprays daily in the Huntington’s trial, and a final daily dose of roughly 44 mg of cannabidiol with under 1 mg of THC in the 2026 Parkinson’s trial are all modest exposures. Investigators chose them for tolerability in frail populations, which is defensible, but a negative trial at a low dose does not settle the question at a higher one.

Endpoint choice is the second. The Unified Parkinson’s Disease Rating Scale and the Unified Huntington’s Disease Rating Scale were built to track disease, not to detect symptomatic comfort. A patient who sleeps better, feels less anxious, or has less pain may not register on either. The 2026 Parkinson’s trial did use a pain-specific instrument and still found nothing, which partly answers this objection for that trial but not for the older ones.

Duration is the third. Three weeks in dementia, six weeks in motor neuron disease, nine weeks in Parkinson’s, twelve weeks in Huntington’s. These are short exposures for chronic, slowly changing conditions, and none of them can speak to what happens over a year.

None of these trials shows that cannabis or any cannabinoid slows, halts, or reverses neurodegeneration. None of them was designed to. Neuroprotection in these diseases remains a laboratory finding in cell and animal models, and laboratory neuroprotection has failed to translate for many drug classes before this one.

They also do not show that cannabis is useless for people with these conditions. Symptom targets such as sleep, appetite, anxiety, and pain sit outside most of these pre-specified endpoints. Absence of evidence on those targets in this particular set of trials is not evidence of absence, and it is not a reason to dismiss what a patient reports.

Finally, they say almost nothing about the products most patients actually buy. These trials used pharmaceutical-grade cannabidiol, a standardized oromucosal spray, and characterized extracts. A dispensary product of unknown terpene content and variable potency is a different exposure.

The pattern here is familiar across neurology. A compound shows anti-inflammatory and antioxidant effects in a disease model, generates enthusiasm, and then produces flat human trials. That has happened with numerous candidate neuroprotective agents that had nothing to do with cannabis. It is a statement about how hard these diseases are, not a statement about cannabis specifically.

The spasticity finding fits a different and more coherent story. Cannabinoids have the most consistent clinical track record in spasticity, which is where nabiximols earned regulatory approval in multiple countries for multiple sclerosis. The motor neuron disease result extends that same mechanism to a new population rather than opening a new one.

That distinction is worth holding onto. Where cannabinoids work in neurology, they tend to work on the symptom layer, through mechanisms that are reasonably well described. Claims that outrun that layer should be treated as hypotheses.

The families who come to me after a Parkinson’s or ALS diagnosis have usually read a headline about neuroprotection before they reach my office. I understand the pull of it. Someone has just told them there is nothing to be done, and here is a plant that laboratory studies say protects neurons. I will not take that hope away, but I will not pretend the trials say something they do not.

What I can say honestly is narrower and still useful. If spasticity is the problem, there is real randomized evidence behind cannabinoids and it is worth a conversation with the neurologist. If sleep, appetite, or anxiety are wrecking a household, those are reasonable targets to try carefully, at low doses, with the neurology team informed. If the hope is that this changes the trajectory of the disease, I say plainly that nothing in the human data supports that yet.

The other thing I watch for is the money. These are families under enormous financial and emotional strain, and the neurodegenerative space attracts sellers who know it. A product that costs hundreds of dollars a month and was never tested in the condition on the label is not a therapy. It is a purchase.

Across randomized trials in Parkinson’s disease, dementia, Huntington’s disease, and motor neuron disease, cannabinoids have met one pre-specified primary endpoint: spasticity in motor neuron disease, in a 59-person phase 2 trial. Everything else that was pre-specified came back null. Safety was consistently acceptable. Discuss cannabis with the treating neurologist as a possible symptom tool, not as a disease treatment.

Ask which disease, which symptom, and which endpoint before accepting any claim about cannabis and neurodegeneration. A result in motor neuron disease spasticity does not transfer to Parkinson’s tremor, and a laboratory finding about neuroprotection does not transfer to a person. The trials that exist are small, short, and mostly negative on what they set out to measure, and they were run by people who expected a different answer.