| Audience | Geriatricians, long-term care physicians, palliative care clinicians, dementia caregivers, and cannabis-medicine clinicians treating behavioral and psychological symptoms of dementia (BPSD) |
| Primary Topic | A THC/CBD oil (the MedCanDem trial) tested against placebo for agitation in patients with severe dementia in long-term care |
| Source | Read the full source |
THC/CBD Oil for Dementia Agitation: What the MedCanDem Trial Found
A multicenter randomized crossover trial of a titrated THC/CBD oil in 25 severely demented nursing home residents found no significant reduction in agitation over placebo, but active treatment was linked to a significant drop in as-needed psychotropic medication use, and the oil was safe and well tolerated across fourteen months of real-world care.
| Study Type | Multicenter, randomized, double-blind, placebo-controlled AB/BA crossover trial (MedCanDem) |
| Setting | Five long-term care facilities in the Canton of Geneva, Switzerland; September 2023 to November 2024 |
| Participants | 25 residents randomized (68% female), average age 83, all with a Clinical Dementia Rating score of 3 (severe dementia) and a Neuropsychiatric Inventory score above 10 |
| Intervention | Standardized cannabis oil, 1% THC and 2% CBD, titrated from 2.5 mg THC/5 mg CBD up to a maximum of 20 mg THC/40 mg CBD daily, versus matched placebo oil |
| Design Detail | Two 8-week treatment periods separated by a 1-week washout, with random assignment to an active-then-placebo or placebo-then-active sequence |
| Primary Outcome | Change in Cohen-Mansfield Agitation Inventory (CMAI) score: mean cross-over difference 5.0 (SD 33.8), P = .53, not statistically significant |
| Key Secondary Finding | As-needed (PRN) psychotropic medication administrations were significantly lower during active treatment (0.06 vs. 0.14 per subject per day; P < .0001) |
| Safety | 23 non-blood-pressure adverse events and 5 serious adverse events recorded, all judged unrelated to study drug and attributed to frailty or comorbidities; no exclusions due to a drug-related adverse event |
| Registration | ClinicalTrials.gov NCT05432206; Swiss National Clinical Trials Portal SNCTP 000005168 |
| Journal / Publication | Age and Ageing, published online August 3, 2026 |
| PMID / DOI | 42613673 / 10.1093/ageing/afag239 |
The primary endpoint, change in Cohen-Mansfield Agitation Inventory (CMAI) score, did not separate the active THC/CBD oil from placebo. The mean cross-over difference was 5.0 points (SD 33.8, P = .53), and neither Period 1 (P = .87) nor Period 2 (P = .38) showed a statistically significant between-group difference.
A covariance analysis adjusted for baseline imbalance did reach statistical significance (ANCOVA P = .04), but the authors attribute this to the fact that patients assigned to the active-first sequence started with meaningfully higher baseline CMAI scores, not to a genuine treatment effect. Secondary behavioral scales, including the Neuropsychiatric Inventory (P = .87) and the Doloplus-2 pain score (P = .27), told the same null story.
While agitation ratings did not differ between arms, the number of as-needed psychotropic medication administrations did: 0.06 per subject per day during active treatment versus 0.14 during placebo, a difference the authors report as statistically significant by chi-square testing (P < .0001).
In practical terms, patients received roughly 64 total PRN psychotropic administrations across both treatment sequences during active periods, compared with 153 during placebo periods. Because PRN psychotropics in this population typically mean antipsychotics and benzodiazepines, both carrying real risk in frail older adults, a reduction in their use is a clinically relevant outcome even without a corresponding change on the primary rating scale.
MedCanDem was powered to detect a 20-point reduction in CMAI score based on a prior 2-year observational study, but the observed cross-over difference was only 5 points, well below that threshold, and the trial’s 24% attrition further reduced statistical power.
The authors also point to a likely-too-short, one-week washout: plasma testing showed active-compound metabolite residues persisting in some patients at the start of the placebo period, a carryover effect that the slower drug metabolism of older adults was not anticipated to produce. Baseline CMAI scores were also unevenly distributed between the two randomized sequences despite correct randomization procedures, a chance imbalance that likely masked part of any true effect.
Unlike many pivotal dementia-agitation trials, MedCanDem allowed all concomitant medications, including patients’ existing psychotropics, rather than requiring physicians to taper them before enrollment as the phase 3 brexpiprazole trial did. Participants averaged eight concomitant medications and 2.5 psychotropic drugs each, and 68% already had standing PRN psychotropic prescriptions at baseline.
That real-world design plausibly created a ceiling effect that made a treatment benefit on top of existing medications harder to detect, but it also means the trial’s safety and tolerability findings, including the absence of clinically significant drug interactions on cytochrome P450 phenotyping, apply directly to the multimorbid, polymedicated patients clinicians actually treat.
Cannabinoid-based medications for BPSD remain a thinly studied area: despite a roughly 250% increase in cannabinoid clinical trials globally between 2013 and 2023, only a handful have focused on BPSD, and fewer still have enrolled severely demented, institutionalized patients with the comorbidity burden MedCanDem’s population carried.
The trial’s improvement in both active and placebo groups echoes a pattern documented in other dementia-agitation and psychiatric trials, including the phase 3 brexpiprazole program, and is a recurring methodological challenge for this entire field, not a flaw unique to cannabinoid research.
What I take from MedCanDem is not disappointment that the primary endpoint was negative. It’s respect for a study team that ran a genuinely difficult trial, in real nursing homes, in patients most research excludes, and reported the null result as clearly as the encouraging one. That kind of transparency is rare and valuable, and it is exactly what this field needs more of before anyone makes confident claims about cannabinoids and dementia agitation.
The PRN psychotropic finding is the part I would want a caregiver or geriatrician to sit with. A significant reduction in as-needed antipsychotic and benzodiazepine use, in a population where those drugs carry real mortality and fall risk, is clinically meaningful on its own even without a CMAI difference. I would not use this single trial to start THC/CBD oil as a frontline agitation treatment, but I do think it justifies a larger, adequately powered follow-up trial, ideally with a longer washout and stratification by baseline severity.
How to Read a Negative Primary Endpoint With a Positive Secondary Signal
Clinical trials rarely deliver a single, unambiguous verdict. MedCanDem is a useful teaching case precisely because its two headline results point in different directions: a clearly null primary outcome and a statistically robust secondary finding on medication use.
Learning to hold both results in view, without either dismissing the null primary endpoint or ignoring the medication-sparing signal, is the core skill this paper demands of a careful reader.
Four distinctions that protect interpretation
Primary endpoint versus secondary endpoint
The CMAI agitation score was the trial’s prespecified, powered primary outcome, and it was negative. PRN medication use was a secondary endpoint; it can be genuinely important and still carry less evidentiary weight than a well-powered primary result.
Underpowered versus disproven
A 5-point observed effect against a 20-point powered target, combined with 24% attrition, means this trial could not reliably detect a smaller true effect if one exists. Null and underpowered are not the same as disproven.
Statistical significance versus baseline imbalance
The significant ANCOVA result was explained by the authors as reflecting unequal baseline severity between randomized sequences, not a treatment effect, a distinction the paper draws out explicitly rather than leaving the ANCOVA number unexplained.
Symptom relief versus medication-sparing
A reduction in PRN psychotropic administrations can matter clinically, by reducing drug-related risk, even when it does not correspond to a measurable difference on the agitation rating scale used to assess symptom severity directly.
A Null Primary Endpoint That Still Changed How Patients Were Medicated
Eight perspectives on a randomized crossover trial that failed to prove THC/CBD oil reduces dementia agitation, yet found real, patient-relevant differences in medication use and safety in one of the most vulnerable populations in cannabis-medicine research.
A Trial That Didn't Prove Calmer Days, But Did Show Fewer Rescue Medications
If you are caring for someone with severe dementia, the honest headline here is that this oil did not measurably reduce agitation compared with placebo on the standard rating scale researchers used in this trial. That is worth knowing clearly before hoping for a dramatic behavioral change from a bottle of oil.
What the trial did find is that residents on active treatment needed fewer as-needed doses of antipsychotics or sedatives, drugs that carry real risks such as falls and stroke in older adults. That is a meaningful, if different, kind of benefit, and one worth raising directly with your loved one’s physician.
Weighing a Negative Primary Endpoint Against a Real Medication-Sparing Effect
For clinicians managing behavioral and psychological symptoms of dementia, MedCanDem’s honest message is that a titrated THC/CBD oil did not outperform placebo on the CMAI, the field’s standard agitation measure, in a rigorously randomized, blinded, real-world trial conducted across five long-term care facilities.
The medication-use finding is harder to set aside: a statistically robust drop in as-needed antipsychotic and benzodiazepine administrations, in a polymedicated population where those drugs carry real risk, deserves clinical attention even though it was analyzed as a secondary, not primary, endpoint in this trial.
Twenty-Five Patients and a Missed Power Target Limit What This Trial Can Prove
MedCanDem was designed to detect a 20-point CMAI improvement but observed only a 5-point cross-over difference, in a trial of just 25 randomized patients that lost 24% to death, withdrawal, or hospitalization over its fourteen-month course.
Those numbers mean the study was structurally underpowered to detect anything short of a large treatment effect, and the single significant ANCOVA result, driven by baseline imbalance between randomized sequences rather than treatment, should not be read as evidence the drug worked.
The Authors Name Their Own Limitations Instead of Hiding Them
This paper’s strongest feature may be its self-critique: the authors directly attribute their null result to baseline imbalance, an underpowered 5-point observed effect against a 20-point target, a likely-too-short one-week washout, and 24% attrition, rather than quietly downplaying any of them.
They also flag that dementia etiology was never systematically classified, meaning Alzheimer’s, vascular, and mixed dementia patients were analyzed together despite potentially different cannabinoid responsiveness, an important caveat for anyone extrapolating this trial’s findings to a specific dementia subtype.
One of the First Randomized Trials in This Specific, Underserved Population
Despite a roughly 250% rise in cannabinoid clinical trials worldwide since 2013, only a handful have targeted BPSD, and MedCanDem is among the very few randomized, placebo-controlled trials conducted specifically in severely demented, institutionalized, multimorbid older adults rather than healthier or community-dwelling populations.
That population choice is itself a contribution: prior encouraging signals came from a 2-year observational study in similar patients, and MedCanDem is the first attempt to test that hypothesis under randomized, blinded, placebo-controlled conditions in the same real-world care setting.
What This Means for Families and Facilities Weighing Cannabinoid Options Today
For a facility or family already considering THC/CBD oil for a resident with severe dementia agitation, this trial does not provide evidence that it will outperform standard care on symptom control, but it does offer real dosing, titration, and safety data drawn from fourteen months of institutional use.
The trial’s titration schema, starting at 2.5 mg THC and 5 mg CBD and increasing gradually to a maximum of 20 mg THC and 40 mg CBD daily, along with its daily blood pressure monitoring protocol, gives prescribing physicians a concrete, tested framework rather than an untested starting point.
What a Definitive Follow-Up Trial Would Need to Include
A stronger successor trial would need a larger sample sized to detect an effect closer to what MedCanDem actually observed, stratified randomization to prevent the baseline CMAI imbalance seen here, and a washout period long enough to account for slower cannabinoid metabolism in older adults.
Systematic classification of dementia etiology, prespecified analysis of PRN medication use as a co-primary rather than secondary endpoint, and replication beyond a single Swiss canton would all strengthen confidence in whether the medication-sparing signal seen here is real and reproducible.
A Safety Case Study for a Population Regulators and Facilities Underserve
Long-term care facilities and regulators grappling with how to manage BPSD safely should note that MedCanDem found no clinically significant drug-interaction signal on cytochrome P450 phenotyping, stable vital signs, and no serious adverse events attributed to the study drug across fourteen months of real-world, polymedicated use.
That safety profile, generated in one of the most complex, high-risk populations in medicine, is itself policy-relevant: it supports continued, carefully monitored research into cannabinoid alternatives to antipsychotics and benzodiazepines, medications whose risks in this population are already well documented in the literature.
Join the Conversation
Have a question about how this applies to your situation? Ask Dr. Caplan
Want to discuss this topic with other patients and caregivers? Join the forum discussion
Frequently Asked Questions
What is the MedCanDem trial?
MedCanDem is a multicenter, randomized, double-blind, placebo-controlled crossover trial that tested a standardized THC/CBD oil against placebo for behavioral and psychological symptoms in 25 residents with severe dementia across five long-term care facilities in Geneva, Switzerland.
Did the THC/CBD oil reduce agitation?
No. The primary endpoint, change in Cohen-Mansfield Agitation Inventory (CMAI) score, showed no statistically significant difference between active treatment and placebo (mean cross-over difference 5.0 points, P = .53).
What dose of THC and CBD was used?
Dosing was titrated from 2.5 mg THC and 5 mg CBD up to a maximum of 20 mg THC and 40 mg CBD daily, given as a 1:2 THC-to-CBD ratio oil, divided into two daily oral administrations.
Did anything improve significantly with the active treatment?
Yes. As-needed (PRN) psychotropic medication administrations, primarily antipsychotics and benzodiazepines, were significantly lower during active treatment (0.06 per subject per day) than during placebo (0.14 per subject per day, P < .0001).
Was the THC/CBD oil safe?
The oil was well tolerated. Five serious adverse events and 23 other adverse events occurred, but all were judged unrelated to the study drug and attributed to participants' frailty and comorbidities; no participant was excluded due to a drug-related adverse event.
Why might the trial have missed a true treatment effect?
The observed 5-point effect was far below the 20-point effect the trial was powered to detect, attrition reached 24%, baseline agitation scores were unevenly distributed between randomized sequences, and a one-week washout may have been too short given slower cannabinoid metabolism in older adults.
Who was included in the trial?
Twenty-five residents of five Geneva long-term care facilities were randomized; participants averaged 83 years old, were 68% female, and all had a Clinical Dementia Rating score of 3, indicating severe dementia, along with an average of eight concomitant medications.
Is this the first study of its kind?
It is one of the first randomized, placebo-controlled trials of a cannabinoid-based medication conducted specifically in severely demented, institutionalized, multimorbid older adults, building on an earlier 2-year observational study in a similar population.
Was the trial registered?
Yes. MedCanDem was registered on ClinicalTrials.gov (NCT05432206) and the Swiss National Clinical Trials Portal (SNCTP 000005168), and its protocol was published before the trial's results were reported.
Should families or facilities change dementia care based on this trial?
Not based on this trial alone. It did not show an agitation benefit over placebo, though it did show a significant reduction in as-needed psychotropic medication use and a favorable safety profile; any decision should involve the resident's treating physician.
